_id
stringlengths 36
36
| bug_id
int64 1
425k
| description
stringlengths 1
373k
⌀ | bug_severity
stringclasses 7
values | bug_status
stringclasses 3
values | resolution
stringclasses 8
values |
---|---|---|---|---|---|
{'$oid': '52e9b44754dc1c25ebdb1ee5'} | 3 | KM (10/2/2001 5:55:18 PM)
This PR about the deletion indicator in the sync viewer being too subtle.
This would be even more true for someone with vision problems.
NE has good suggestion of using the label of the compare text viewer.
We may also consider a better icon.
On to the PR....
- delete file from project
- synchronize
- there is no indication that the file has been deleted from the workspace
- it looks the same as if the file is in the workspace but is empty
NOTES:
KM (8/15/2001 3:37:53 PM)
It should appear as a rightfacing '-' arrow in the sync view.
Sent screen cap to NE requesting clarification.
NE (8/20/01 9:13:38 AM)
The - in the arrow is clearer when there are other arrows present too (the + and change ones).
Otherwise, it's easy to miss the distinction, especially when the user's attention is focussed
on the text compare panes.
I'd expect the Workspace file: label to indicate that it has been deleted.
The Repository file: label does this (if I remember correctly) when it has been deleted from the repository. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44754dc1c25ebdb1ee6'} | 1 | - Setup a project that contains a *.gif resource
- release project to CVS
- edit the *.gif resource with an external editor, e.g. PaintShop
- save and close external editor
- in Navigator open the icon resource and verify that your changes are there
- release project
-> nothing to release!
- in Navigator open the icon resource and verify that your changes are still there
Problem: because I never "Refreshed from local", the workspace hasn't changed so "Release" didn't find anything.
However opening the resource with an external editor found the modified file on disk and showed the changes.
The real problem occurs if "Release" actually finds something to release but you don't spot that some resources are missing.
This is extremely error prone: one of my changes didn't made it into build 110 because of this!
NOTES:
EG (5/23/01 3:00:33 PM)
Release should do a refresh from local before doing the release.
Moving to VCM
KM (05/27/01 5:10:19 PM)
Comments from JM in related email:
Should not do this for free. Could have a setting which made it optoinal but should nt be mandatory. Default setting could be to have it on.
Consider the SWT team who keep their workspaces on network drives. This will be slow.
Side effects will be that a build runs when the refresh is completed unless you somehow do it in a workspace runnable and don't end the
runnable until after the release. This would be less than optimal as some builders may be responsible for maintaining some invariants and deriving resources which are releasable. If you don't run the builders before releasing, the invariants will not be maintained and you will release inconsistent state.
Summary: Offer to "ensure local consistency" before releasing.
KM (5/31/01 1:30:35 PM)
See also 1GEAG1A: ITPVCM:WINNT - Internal error comparing with a document
which failed with an error. Never got log from Tod though. | normal | CLOSED | FIXED |
{'$oid': '52e9b44754dc1c25ebdb1ee7'} | 8 | With the current VCM API, a repository adapter can either be pessimistic or
optimistic.
If it is pessimistic, all check-outs are "reserved", i.e. no other user
is allowed to check
out the same resource. If it is optimistic, no check-out is "reserved".
In other VCM systems (e.g. CC and CVS), you can decide on a resource-by-
resource basis
if a check-out should be "reserved" or not.
Usually, a check-out is not "reserved" if a merge is not difficult,
e.g. if there are merge tools.
If there are no merge tools, and a merge by hand is not trivial, the
check-out should be a
"reserved" check-out.
NOTES:
BB (3/1/2001 9:36:23 AM)
Proposed change to the VCM API: checkout() takes an additional boolean
parameter
"reserved" and the corresponding activity event is changed accordingly.
Does not seem relevant for
CVS. Moving to Active_MarkH for comment.
SK (4/11/2001 6:10:17 PM)
This will be solved via use of ICCActivity to do checkout and unlock.
SK (4/25/2001 3:49:50 PM)
Moving back to active, as this work is only a release candidate.
KM (4/25/01 4:42:08 PM)
From note from Mark H:
I am not exposed to the whole development
community, so I am not aware of what people are complaining
about/asking
for. It is needed if we want to support serial development.
Estimated development time is 1 week by Steve, so I would think
that puts
it in the future bucket by necessity.
Jean-Michel (11/06/2001 9:37:22 AM)
Our current internal VCM API does not support a pessimistic model. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44754dc1c25ebdb1ee8'} | 4 | - become synchronized with some project in a repository
- use a different Eclipse to make a change to a file resource within this project and release it to the repository
- in the original Eclipse mark this file resource as being read-only (select it, right-click -> Properties, change, OK)
- select the file resource, right-click -> Team -> Synchronize with Stream
- in the subsequent comparison view select the file resource, right-click -> Catchup
- since it has been marked as read-only there are inevitable problems. However the error dialog that is shown
does not offer much assistance ("An IO error occurred: IO Error")
NOTES: | normal | RESOLVED | FIXED |
{'$oid': '52e9b44754dc1c25ebdb1ee9'} | 6 | IResource.setLocal has problems. This method (which replaces ensureLocal
and scrubLocal) requires some form of connection to the VCM provider,
since the VCM provider is the only thing that could make a resource local.
Indeed, the whole notion of "local" resources requires VCM support and needs
to be reviewed.
NOTES:
JM (12/11/2000 10:24:05 PM)
The only scenario in which non-local resources makes sense is one
involving a VCM.
setLocal does not replace ensureLocal. It is there to support the VCM
fetching content
and then telling the workspace that it should expect to see the content
in the file system.
Similarly, with scrubLocal. It is expected that the VCM itself is the
one that does the
dirty work of adding/deleting the content in the file system.
jeem (12/13/00 7:01:26 PM) - Where are ensureLocal and scrubLocal? They don't
appear
in the API any more (they were there in tech preview).
JM (01/02/2001 11:32:24 AM)
These notions should be related to the VCM.
Moving to BB for consideration.
KM (4/6/01 3:49:13 PM)
Only sender that calls it with 'false' is ScrubLocalAction, which looks
like it can be deleted.
We are not making use of this core feature since neither CVS nor CC
benefits from it.
I am concerned then it may not even work given there are no senders
of 'false'.
Nothing to be done about this right now, moving to future.
JM (06/04/2001 6:58:28 PM)
There likely are test suites which test the (false) case.
This is both legacy and future expansion. The VCM model does not
require one to
have all resources local. CVS and CC appear to have a model where
everything is local.
Things like VSS do not use this model. I believe that PVCS and
Perforce do not require you
to have everything local but that is fuzzy for me. DAV does not
specify either way as
it is a protocol. It is easy to see both modes of operation (i.e.,
both "I GET the whole project" and
"I GET only what I need").
Regardless, the setLocal functionality relates to modeling resource
existance
without actualy being forced to fetch the resource.
So, this is core API which is related to part of the VCM model which is
not yet fully implemented.
If you have a particular concern about the API being there, we can
remove the it until VCM
implementations for VSS, DAV, ... are being done and determine that
they need the behaviour. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44754dc1c25ebdb1eea'} | 7 | The platform is able to notify people that a resource was moved using resource
change events (deltas). It does not track copies.
In neither case is it easy for a VCM or target to retain the semantics of a move
or copy since by the time it is run, all it has it the current state (i.e., it can tell
that a file was deleted and another added but not that this was a move).
This is unfortunate since the VCM/target sync code may well be able to do
optimizations if this information was available.
NOTES:
JM (12/10/2000 2:49:42 PM)
This is related to
1G5RAC5: ITPCORE:WINNT - move/copy semantics not preserved for VCM/targets
It is only relevant to address non-local move/copy if the above PR is addressed.
otherwise, there is nothing to do at release time since there is no local copy and all
we can do is put/delete.
JM (4/15/2001 9:54:43 PM)
There is nothing we can do about this now. Moving to VCM for logging as a future
issue. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44754dc1c25ebdb1eeb'} | 5 | For getting/setting the managed state of a resource, the methods are:
isManaged(resource) and
manage(resource)
for getting/setting the ignore state, the methods are:
getIgnored(resource)
setIgnored(resource)
These should be made more consistent. I suggest renaming ignore methods:
isIgnored(resource) and
ignore(resource).
I think it's good practice to not use "get" and "set" in method names unless it's really
just a field accessor.
NOTES: | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44754dc1c25ebdb1eec'} | 9 | The implementation has to be changed because a root resource might be passed
where an IResource is expected. See also 1G8435U
NOTES:
KM (4/26/01 3:54:24 PM)
This PR is very old. It should either be fixed, moved to future, or obsolete. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44754dc1c25ebdb1eed'} | 10 | There seems to be a need for a VCM Event notification model, along the lines
of the activity notification and resource change notification. This would only
notify listeners when local VCM API had been manipulated, it wouldn't notify
changes that have occurred on the server. For example, notification of the
following events (and possibly others) would greatly improve my ability to
maintain a well refreshed UI:
- repository creation/deletion
- team stream creation/deletion
- team stream lock/unlock
- setSharing of project to team stream
- project version creation/deletion
- catchup/release to/from team stream
Without such a notification method, we're left with a LotusNotesian "Press F9
to refresh" UI model. For example, the user has a view open that shows a list
of the currently known team streams. The user then creates a new team stream.
S/he will then have to press F9 on the team stream view to make it include the
TS they've just created. We cannot link the team stream creation wizard and the
team stream view at the UI level, because they might well have been created by
different plugins that don't know about each other.
Also, repository deletion is not currently supported. By deletion I mean removal
of the repository from the list of the workspace's known repositories. Once the
known repositories list is persistent, there will be a need to remove items from
the list to prevent it from growing indefinitely.
NOTES:
BB (2/1/01 11:57:12 AM)
For a first cut, see
>IVCMEvent 0.4 000
>IVCMListener 0.4 000
>IVCMProvider 0.3 010
JA 2/1/01
Some possible additional event types:
- change in resource transience.
-create/delete project version
As a side note, you might consider using IVCMListener.handleEvent
instead of IVCMListener.vcmEvent. It is more in line with other event/listener
pairs in Eclipse.
BB (2/2/01 2:51:14 PM)
We should change IActivityListener.activityEvent too. (-> IActivityListener.handleEvent)
KM (4/6/01 12:13:55 PM)
Prioritization:
Important:
- update labels (navigator) and property sheets when loading/versioning (explicit refresh)
- update sync info on resource change (vcm provider becomes a delta listener)
- update prop sheet when change in sharing
Nice to have but can be dropped:
- show creation of new repo connections, streams
- update repo stream children when releasing
- update version list when versioning
BB (5/11/2001 12:29:33 PM)
Implemented events for creation/deletion of repositories/team streams.
JohnA (5/11/2001 6:24:49 PM)
Implemented event listening for creation/deletion of repositories/team streams. | normal | VERIFIED | FIXED |
{'$oid': '52e9b44754dc1c25ebdb1eee'} | 14 | KM (4/5/01 8:53:30 PM)
This is arguably not stop ship and could be dropped from June release.
These appear to be the important file:
CVS/Entries (and possible friends)
CVS/Repository
CVS/Root
CVS/Tag
See
http://cvsbook.red-bean.com/cvsbook.html#Working_Copy_Files
for purpose and contents of these files.
If we don't have them, you won't be able to run a CVS tool against our local files in the workbench.
However, maintaining these files is a problem since it duplicates information we are already
storing using Core mechanisms (eg. version/sync info).
Problems:
1. Keeping these files up to date due to changes in Eclipse
2. Reading them in order to keep up with changes by an external tool
3. Ensuring we are using the correct format (testing)
Q: what happens with migration to/from other repository types (eg. CC <-> CVS)
NOTES:
KM (4/10/01 11:52:24 AM)
Moving to Future. For June delivereable we will not support 3rd party CVS tools over our workbench files.
Discussed with both John W. and Dave T.
Jean-Michel (14/06/2001 4:08:20 PM)
Add to readme. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44754dc1c25ebdb1eef'} | 12 | Currently every new resource added to a shared (version controlled) project is
managed. In
the real world this won't work since most derived artifacts from an active
developed project
will never be managed (e.g. class files, object files, user configuration
files...).
The VCM API has ISharingManager::manage() and isManaged() operations. The UI
should
provide support for managing resources. By default resources would be
unmanaged, even
if parent is managed and project is shared. A policy could be added to allow
new resources
created with the UI (e.g. and not generated by the build process) to be managed
automatically.
NOTES:
KM (4/12/01 6:37:32 PM)
From discussion with Jean-Michel, still feels like we need to close on
whether they are hints or more. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44754dc1c25ebdb1ef0'} | 13 | Would it make sense to have a (the?) project's "Version" menu on its popup in a
stream
(in the repo explorer)?
NOTES:
Jean-Michel (22/03/2001 8:29:49 AM)
"Tagging affects the repository, not the working copy. That begs
the question: Why require a working copy at all when tagging? The only
purpose that it serves is to designate which project and which
revisions of the various files in the project are being tagged. If you could
specify the project and revisions independently of the working copy,
no working copy would be necessary. " from CVS manual.
CVS can tag (take workspace revisions and tag then in the repo) and
rtag (no workspace required, take repo revisions and tag. This command
is mostly always used when branching 'cvs rtag -b -r version1 version1_bugfix'
so you don't tag the team stream but ). Currently we version using
rtag just because rtag records a tag event in the history and allows
us to query for existing tags. But using rtag on the team stream (HEAD) is
dangerous because you don't really know what you are tagging.
Notes:
- in our cvs adapter we must use tag where we currently use rtag.
- to find the versions created with tag in the repo the only thing to do (as
recommended by the CVS docs):
"Until then, there are workarounds. You can run cvs log -h and read the
sections of the output following the header symbolic names:."
but how do we chose which file to look into, it could be a file that has just
been added to the repo?
- we must allow the user to enter his own tag name where ever he can select a
version.
- we could use rtag internally when branching (creating a new team stream)
based on a version.
BB (26/03/2001 10:31:40 AM)
I think we should provide two operations: Create a project version from
a team stream,
and create a project version from the workspace. The API has both
operations:
ISharingManager.createProjectVersion()
ITeamStream.createProjectVersion()
The first one should be on the context menu for projects in the
navigator, the second one
on the context menu for projects in team streams (repository explorer).
CVS has both operations too.
from Jean-Michel's comment: "- in our cvs adapter we must use tag where we
currently use rtag."
I disagree. I think we need both because CVS provides both too.
KM (3/26/01 2:23:40 PM)
Supporting both makes for a very confusing user model. Versioning from
the workspace is closer
to VAJ and is arguably better since you have a fixed state that you are
versioning
(as opposed to trying to version a non-locking changing stream).
KM (4/4/01 4:36:51 PM)
Decision is to support both. Versioning stream is helpful for
baselining before releasing.
Must make the distinction clear in the UI. As a bonus, informing the
user that he his
workspace differs from the stream when versioning the stream may be
helpful.
JohnA (4/6/01 5:20:27 PM)
Added team stream versioning to the UI. Moving the PR to active so we
don't
loose Mark's useful comments above. | normal | VERIFIED | FIXED |
{'$oid': '52e9b44754dc1c25ebdb1ef1'} | 11 | Jean-Michel (08/02/2001 1:38:48 PM)
The ISharingManager::load operation does not support the following scenario:
- Create project in workbench called TomCat3.3 and share a teamstream with a
vcm project called TomCat with version 3.3.
- Create another project in workbench called TomCat4.0 and share
with same tream stream and same project TomCat but with version 4.0 of the
project.
This issue can be more generally seen as being about the semantics of "projects"
in the workbench versus those in vcm. Projects in the workbench should be
allowed
open mapping to any vcm project, regardless of the name of the local workbench
project
that is chosen. This is similar to ClearCase's and CVS's notion of having any
number
of local work areas for a project configuration. In ClearCase these are called
views, and you
can have as many as you like. Likewise, in CVS you can use the -d option with
the
update command to load a specific module version to a user named directory (the
default
is to use the module's name).
The load API mentions that it is responsible for renames. If this is the case
then if
projects are to be threated differently, an alternative method should be
available for
renaming projects on the server. Local projects, or workbench projects, could be
renamed without affecting the repository project to which it is mapped.
NOTES:
JM (08/02/2001 7:56:47 PM)
Yes. Currently there is an (at least underlying) assumption that the
name of a
project in the workspace is the same as its name in the versioning
system. This
should be relaxed.
Moving to BB for consideration. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44754dc1c25ebdb1ef2'} | 15 | In support of some of the VCM CVS work we need an SSH implementation that we can use. There are
two basic options, Java client implementation, native C implementation. The native C story has
the standard drawbacks of having to spawn an external process and manage the standard in/standard
out. The Java approach would be more closely integrated but may suffer in performance and
flexibility due to Java's inferiour socket implementation.
NOTES:
JM (18/04/2001 5:21:19 PM)
There are also licensing issues. many of the public implemenations are GPL. This may or may not
be an issue. A good place to start looking is http://www.openssh.org. The SSH Plugin Java implementation
is likely to be the most promising of the Java guys.
To test/investigate this area we need an SSH enabled Linux box.
Moving to JEP for investigation
JEP (4/24/01 3:49:53 PM)
Here is what I've done so far:
1. Installed openssh-2.5.2p2 (includes linux ssh client and server) on dj.
2. Logged on to dj from dj using linux ssh client.
3. Installed the Java Telnet Application 2.0 (includes java ssh client plugin) on pico1.
4. Logged on to dj from pico1 using the JTA ssh client plugin.
I used the default options and configurations whenever possible.
JohnA (10/2/2001 5:02:53 PM)
Assuming this is resolved. Moving to VCM for confirmation/close. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44854dc1c25ebdb1ef3'} | 17 | We need to look at gender change cases, where at a given path
- the workspace has a file but the team stream has a folder
- the workspace has a folder but the team stream has a file
NOTES:
BB (06/04/2001 10:03:58 AM)
Currently, the implementation does not cover these cases. It assumes that the resource types
match.
KM (4/6/01 5:11:46 PM)
This case occurs when on machine A you release folder Foo and on machine B you try to release resource Foo.
Unlikely and CVS doesn't handle well anyway.
We should test to ensure we are not failing bad (eg. Null Ptr Ex). | normal | RESOLVED | FIXED |
{'$oid': '52e9b44854dc1c25ebdb1ef4'} | 18 | Merge managers store sync info in the resource tree. Because this is persisted, we need some
form of life cycle management for merge managers. When a merge manager is no longer used,
we can then flush its sync info.
NOTES:
KM (21/05/01 6:58:10 PM)
Since this is solved by taking a new workbench, moving to Future.
BB (6/4/2001 7:47:20 AM)
Note that there is rudimentary life cycle management for merge managers that we have put in;
sync info of merge managers is cleared at startup. This should be changed to support ongoing merges.
See 1GEO006. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44854dc1c25ebdb1ef5'} | 19 | I should be able to pick a stream from the respository and catchup to the stream.
Instead I must select projects within the stream and "add to workspace".
This is annoying because it doesn't merge it bashes, and besides it warns you
about every project that you already have loaded.
Basically, it should work the same as picking the individual projects from the navigator
and catching up, except instead I'm picking the stream to denote the set of projects.
NOTES:
BB (17/04/2001 9:57:55 PM)
Note that the lineup of projects in a stream (at least for CVS) is not necessarily
meaningful, because projects added to one stream will show up in all other
streams in the same repository. A new project in a stream is also not an incoming
creation of the project; you would have to use project references for transitive
loading or transitive catching up.
It would be useful, though, to have a global toolbar button which does a catchup
on all shared projects.
Jean-Michel (4/25/01 9:14:46 AM)
1GCQK5Y: ITPVCM:WINNT - Should be able to add HEAD to workspace
Moving to time permitting. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44854dc1c25ebdb1ef6'} | 21 | Some CVS users are used to using the release command to delete their working checkout (e.g. in Eclipse
terminology this means to delete the working project). In Eclipse you can workaround this:
1. release the project
2. delete project from workspace (simply by deleting from within Eclipse)
Note: this won't trigger a note added to the cvs history file.
Cancels a checkout (indicates that a working copy is no longer in use). Unlike most CVS commands that
operate on a working copy, this one is not invoked from within the working copy but from directly above
it (in its parent directory). You either have to set your CVSROOT environment variable or use the -d global
option, as CVS will not be able to find out the repository from the working copy.
Using release is never necessary. Because CVS doesn't normally do locking, you can just remove your working
copy. However, if you have uncommitted changes in your working copy or you want your cessation of work to
be noted in the CVSROOT/history file (see the history command), you should use release. CVS first checks for
any uncommitted changes; if there are any, it warns you and prompts for continuation. Once the working copy is
actually released, that fact is recorded in the repository's CVSROOT/history file.
NOTES: | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44854dc1c25ebdb1ef7'} | 2 | Opening repository resource always open the default text editor and doesn't
honor any mapping between resource types and editors.
As a result it is not possible to view the contents of an image (*.gif file) in a sensible way.
NOTES:
EG (5/24/2001 5:13:45 AM)
Agreed that this is needed eventually but not critical for June.
There should also be a Java outliner when editing a java resource from the repository.
Moving to VCM.
KM (5/24/01 2:09:21 PM)
In general you cannot open a specialized editor on a remote resource because they can sometimes
be dependent on other resources and expect these to be local. For example, an HTML page which
references a graphic, the HTML editor is going to try to look for the graphic to display it.
You can't be sure that the editor was written to accomodate remote or 'missing' resources.
We discussed this some time ago with Dave T. and felt it was safer to just use vanilla editors for now.
KM (5/24/01 4:42:09 PM)
See also 1GE8YMJ: ITPVCM:ALL - 'Repositories View': all file types open to the text editor | normal | RESOLVED | FIXED |
{'$oid': '52e9b44854dc1c25ebdb1ef8'} | 22 | Project descriptions don't store sharing recommendations and project version references yet, but they should.
NOTES:
Jean-Michel (4/6/01 12:39:12 PM)
We will live with unqualified project references for now. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44854dc1c25ebdb1ef9'} | 16 | In the merge view, I'd like to have an auto-merge button with the semantics:
"Copy all non-conflicting incoming changes from the server resource to the local resource".
After clicking on that button, I would only have to look at the red stuff (conflicting incoming changes)
and decide what to do with them.
CVS users are used to auto-merge in the "trivial" case where the changes on the source code (line) level don't overlap.
When I used CVS, the normal mode for me was to not look at changes at all unless they are non-overlapping.
I don't know what the existing buttons "Copy all from left to right" and "Copy all from right to left" are good for.
NOTES: | normal | RESOLVED | FIXED |
{'$oid': '52e9b44854dc1c25ebdb1efa'} | 20 | Thought it would be useful if the set of repo connections could be stored somewhere
external in the file system instead of everyone having their own list cashed in
their workspace (.metadata). Thus if you get a new workspace you don't need to reconstruct
this information, you can just configure your workspace to point to that list.
Also, it means a team such as ours could share a set of repo connections by
all pointing to the same file.
NOTES: | normal | RESOLVED | FIXED |
{'$oid': '52e9b44854dc1c25ebdb1efb'} | 23 | Not sure if anyone is using ScrubLocalAction.
This is related to the PR
1G5TC8L: ITPCORE:ALL - API - IResource.setLocal has problems
Its possible nobody in practice calls IResource.setLocal(false)
NOTES:
JM (06/04/2001 6:55:52 PM)
Don't know what ScrubLocalAction is.
Moving to VCM
See referenced PR for comments on setLocal | normal | RESOLVED | INVALID |
{'$oid': '52e9b44854dc1c25ebdb1efc'} | 24 | Since refresh() is called before the catchup/release view opens, it should be as fast as possible.
NOTES:
BB (12/04/2001 11:32:05 AM)
The most promising optimization idea is: Don't return an IServerResourceTree from fetchRemoteTree
but instead use a call-back to interleave the server communication with updating the sync info.
BB (12/04/2001 11:30:31 AM)
See this PR for other performance tuning opportunities: 1GAT2V2: ITPVCM:WINNT - Releasing very slow
MichaelV (19/09/2001 2:14:23 PM)
Here are some percentages on a refresh from OTT to Paris with auto-build turned OFF
Total Time: 97776 ms
(57026) 58.3% in Subscriber.refresh()
(40704) 41.6% CVSAdapter.fetchRemoteTree()
(14918) 15.3% Subscriber.recursiveRefresh()
(13242) 13.5% in SharingManager.manage() from org.eclipse.vcm.ui.sync.Repository.autoManage()
What is autoManage?
(12234) 12.5% in org.eclipse.vcm.ui.sync.Repository.collectResourceChanges() (recursive)
11% in JavaModelManager.resourceChanged()
4.3% in ResourceDeltaFactory.computeDelta()
Hot-spots are
14% socket read
13% socket write
12% String#substring
4.6% String#indexOf
3.6% Path#segments
3.2% String#compareTo
3.2% AccumulatingProgressMonitor#subtask
3% Path#append
3.2% String#concat
Auto manage time (13.5%) seems to be spent determining what is ignored. There seems to be many ways to improve this!
MichaelV (20/09/2001 8:51:55 AM)
There seems to be a lot of room for improvement in Subscriber#recursiveRefresh
However, the general approach used by VCM to determine what changes there are seems to be
inherently ineficient.
In CVS clients, the following steps take place on an update
1. Client sends information about all loaded files (versions and what has changed locally)
2. Server makes one pass through to determine update reponse.
3. Client uses response to update only as necessary
This results in 2 to 3 passes through the project
In VCM, the following occurs
1. client sends information about known directories
2. Server responds with information about new directories and files
3. client requests file information (status)
4. server responds + client must place file in resulting tree
5. resulting tree + local structure traversed to update sync info
6. local structure traversed to determine if their are changes
My guess is that this is roughly equivalent to 5 passes
Considering that some projects may be very large, this leads to slow syncing
which takes the same time whether there are changes or not.
MichaelV (21/09/2001 2:06:02 PM)
I have optimized the recursive refresh itself but this has bought little (13000). The problem is
that we do a get and write of the sync info at each node and this is costly. Furthermore,
getting and setting get slower as the distance from the root increases. This is probably due
to the fact that there are lots of files at the leaves.
Within getWorkspaces, getSharing and getProject are the main culprits. Can these be passed in?
The project should be OK but I'm unsure of the sharing. getRepository also does a getProject.
I have optimized SharingManager.manage to avoid internalGetSyncInfo and getProject. (from 13000 to 9000)
However, we should not need a separate traversal to ensure that all non-ignored resources are managed!
I have included change collection as part of the refresh.
(Much speed improvement!! from 12000 to virtualy 0)
MichaelV (24/09/2001 12:49:57 PM)
Timings of local sync with auto-manage refresh and collection in same pass.
100% = 30290 ms
82.73% (25059) in Subscriber.refresh
49.95% (15130) in CVSAdapter.fetchRemoteTree
27.25% (8257) in Subscriber.refreshChildren
5% (1512) in Policy.countResources
15.34% (4649) in NotificationManager.getDelta()
This is a savings of over 75% on VCM operations but no improvement in fetchRemoteTree.
Also, getDelta timings result from all the storeSyncInfo calls. One solution is
to request only the delta information from the server as the CVS command-line
client does.
Test Cases
Sync for first time
- on unshared project
- on folder/file of unshared project
- on folder/file of shared project
Sync with outgoing, incomming and both
Merges with outgoing, incomming and both | normal | VERIFIED | FIXED |
{'$oid': '52e9b44854dc1c25ebdb1efd'} | 25 | It would be helpful if there was a notion of default repository connection so that when you
connected a project you wouldn't have to walk through the list everytime.
Not sure how this goes into prefs if its never been created before. Maybe when you create them you can mark one
as the default.
At a minimum having a default repository TYPE (cvs/cc) would be helpful (which then the pulldown in the repo connection
wizard has a default).
NOTES: | normal | RESOLVED | DUPLICATE |
{'$oid': '52e9b44854dc1c25ebdb1efe'} | 29 | VCM is not NSL ready! Need resource bundles.
Policy has bind() methods ready for use.
NOTES:
JohnA (4/17/01 2:03:02 PM)
Painfully, all six VCM plugins will have to have a Policy class,
and duplicate the code for generating strings from bundles.
Isn't core supposed to be providing support and a story for this?
JEM (5/28/01 3:45:45 PM)
Internationalization complete.
Each plugin currently has getResourceString(String) and
getResourceString(String, Object[]). Core has no support for us.
In the future we should shorten the names (e.g. bind) because they're
used so much. Moving to Inactive for future consideration. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44854dc1c25ebdb1eff'} | 28 | Currently, the properties/Managing page for files shows the creation date and author of
the base version, but not of the common or the remote version. It should either display
this information for all three versions or for none. It could be like this:
Status: versioned [one of: versioned / changed / unmanaged]
Local version: 1.1 author: jlemieux created: Mon Jul 01 2001 19:35
Common version: 1.2 author: johna created: Tue .....
Released version: 1.4 author: jlemieux created: ....
NOTES:
BB (4/25/01 12:39:00 PM)
Ideally, the property page would first display information that's available without contacting the
server. Currently, it does a refresh() on the resource; this doesn't work when you're disconnected.
The property page could have a "More info >>" button that does the fetchStateInfo() required to get
the author, date etc. and then displays the more detailed information.
Displaying the version comments would be nice too.
Note that if you do a sharingManager.getRemoteResource(resource).fetchStateInfo(), you don't
need the refresh as this will give you the version number already.
JohnA (4/25/01 2:37:46 PM)
The property page now only shows all the information it can get without contacting
the server (no author, comment, etc). We can add the "More Info" button later time permitting. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44854dc1c25ebdb1f00'} | 30 | We currently ignore the cvsignore file on the server, this may of been configured for the
whole project by the administrator. Our implementation of the cvs adapter should read this
file.
NOTES: | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44854dc1c25ebdb1f01'} | 27 | CVS tags (a.k.a. project versions) that have not been created using rtag don't
show up in the repository explorer because the history file on the server does
not reference them.
Current workaround: define a team stream using the same tag name, then you can
browse the project version. This is less than optimal because it does not behave
like a team stream - you cannot release anything to it...
NOTES:
KM (5/22/01 10:03:38 PM)
This can occur for open source repos which have history turned off.
Jean-Michel (11/06/2001 4:44:02 PM)
Include the cvs error that is returned when trying to release to a
stream that is really a version :)
cvs server: sticky tag 'v1' for file a.txt is not a branch tag.
Jean-Michel (6/13/01 9:59:17 PM)
We should describe in the CVS docs that in order to browse or add a version
of an existing CVS project that was not created using rtag users should create
rtag version using the cvs command line.
We should probably not mention anything about the creating a stream hack?
Jean-Michel (8/23/01 10:24:09 AM)
Many Eclipse VCM operations rely on project versions being available. For example,
without them you cannot merge, compare/replace project versions, and create a new
stream (copy version to stream). That means that any open source project that has
disabled history would not be able to use much of the functionality in Eclipse VCM. I suggest
the following CVS specific features be added to the CVS adapter:
1. A mechanism for defining a project version by name on CVS projects. This would be
a name not validated until an operation is performed (e.g. load, browsing). The
names would be workspace persistent, and allow using all Eclipse features which rely
on version names. This resembles CVS' support for providing -r <version name> on
the command line, we simply cache a set of -r parameters.
2. Add a CVS action to IFileEditions which allow creating project versions automatically
based on the version tags already defined on the CVS file. This would be easy to do.
Jean-Michel (9/14/01 1:36:14 PM)
Fixed, released into HEAD. | normal | VERIFIED | FIXED |
{'$oid': '52e9b44854dc1c25ebdb1f02'} | 26 | Its very awkward that you cannot create a new repository connection when you go
to associate a project with a repo.
In particular, if you walk through a first-time user's scenario, as soon as he
does VCM operations
he will presented with an empty prompter with no indication of how to get
anything in it.
NOTES: | normal | RESOLVED | INVALID |
{'$oid': '52e9b44854dc1c25ebdb1f03'} | 31 | We have two repositories, one for binaries and one for source.
I switched from one repository to the other (using "Managing" properties page).
Project context menu --> Replace With --> Team Stream Resource
==> com.ibm.eclipse.vcm.ui 1 com.ibm.dt.core does not belong to a team stream.
NOTES:
Jean-Michel (4/25/01 12:08:31 AM)
When changing the sharing of a project to another stream you lose
sync info from the last sharing and don't have any new sharing based on the
new sharing that was specified. This is what causes this PR:
ReplaceWithTeamStream::getEditionToLoad()
...
try {
IResourceEdition edition = getSharingManager().getRemoteResource(resource);
if (edition == null) {
throw new WorkbenchVCMException(IStatus.WARNING, resource.getName() + " does not belong to a team stream.", null);
}
return edition;
} catch (CoreException e) {
throw new WorkbenchVCMException(e.getStatus());
}
Since a load or refresh does not occur when the sharing is changed, getRemoteResource() will always return
null.
Thus changing the sharing from the properties page is not recommended until we help the user
out a little more. After changing the sharing, we should ask:
"do you want to load (add to workspace) team stream" or
"refresh from new team stream" | normal | VERIFIED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f04'} | 32 | The VCM API still has methods for getting (persistent server) properties. This is not implemented.
Should we remove the property-related methods from the VCM API?
NOTES:
KM (4/23/01 3:36:40 PM)
Yes. | normal | VERIFIED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f05'} | 33 | The VCM API has methods for locking team streams (preventing others from releasing). This is not implemented,
and in the case of CVS, not implementable. Should we delete this from the API?
NOTES:
KM (4/23/01 3:37:26 PM)
Yes. | normal | VERIFIED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f06'} | 34 | In the repo explorer, under Project Versions, there are project names which have no versions.
Perhaps we should filter these? When do we find out if there are any versions to display?
(ie. can we do the filtering efficiently without having to get the children version list for all project names).
NOTES: | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44954dc1c25ebdb1f07'} | 36 | Since I'm allowed to release on a resource level I also want
to version it (i.e. give it a CVS tag).
NOTES:
KM (4/23/01 3:16:38 PM)
See discussion in
1GCN7TR: ITPVCM:WINNT - Term "version" confusing when working with CVS
moving as duplicate.
Jean-Michel (26/06/2001 11:05:29 AM)
Should close this PR. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44954dc1c25ebdb1f08'} | 38 | Our use of ISharingManager::getIgnored(IResource[]) does not make use of the fact that
the method takes an array. Calling getIgnored(new IResource[] {resource})[0] is slow.
This is especialy true with the cvs implementation of getIgnored(). We must parse files
in each directory. If an array of resources are passed, we can parse the files only once.
NOTES: | normal | VERIFIED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f09'} | 35 | Whenever I get files from CVS (e.g. as a result of a Catchup or Replace operation)
their modification timestamp shows the time when CVS updated the file on my disk, not the time when it was modified the last time.
So if I keep a local copy of my modified file somewhere (because eclipse is not yet robust enough... :-)
I have a hard time finding out what files were really modified.
I think our CVS implementation should have the same behavior as "normal CVS" would have.
NOTES:
KM (4/26/01 3:13:48 PM)
1. The last modified time of the resource is provided by core and is shown in the Info property page.
2. What we are missing is some kind of indicator that the resource has been modified since the base version.
This was in the plan but we've been waiting for support from UI to decorate elements in the navigator.
AW (27.04.01 10:00:08)
I'm not asking for 1&2) since we have Catchup and Release and other operations to show whether a resource
has been modified since the base version.
I'm asking for normal CVS behavior outside of eclipse, that is replacing a file with a file from the repository
should set back its modification date to the date maintained by the server not the time when
the file was locally written (because that time is of no value).
JohnA (4/27/01 3:51:15 PM)
Moving to inbox for comment from Boris. What he's interested in is the timestamp
stored in the resource handle IStateInfo objects.
KM (5/22/01 10:41:15 PM)
Boris, I thought you changed this?
BB (5/23/2001 5:42:02 PM GMT+1)
Sorry, I forgot this. Will fix.
BB (5/28/2001 11:38:55 AM)
CVSAdapter.load and .loadShallow refactored.
Jean-Michel (02/06/2001 2:30:48 PM)
(1GCOIFN: ITPVCM:WINNT - CVS and modification timestamps)
Sorry, I wasn't able to put this in. However, my confidence level would have been low anyway.
Here's why:
To set the server's time as the modification timestamp, we would have to install a handler (ModTimeHandler) and read the
timestamp from the context provided by the cvs client. We would be using some of Dirk's old code that we haven't touched
at all for a very long time (some of which is currently not used). I hope you don't mind if I make the necessary changes on a
separate branch Monday or Tuesday next week.
Thanks,
Boris. | normal | VERIFIED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f0a'} | 37 | The term "version" is confusing when CVS is used
as repository.
Users expect another behavior when versioning current work.
They don't expect that what they have in the workspace must be released first.
Also CVS experienced users might be confused. They expect to be able to
tag a resource.
NOTES:
JohnA (4/23/01 2:16:27 PM)
Versioning == tagging. I expect CVS users will not have too much
difficulty adapting to the terminology change. They may not expect
that they have to release before versioning, but since a warning
dialog tells you this if you try, it quickly becomes clear.
BB (4/23/2001 2:45:40 PM)
You can version a project (both in the workspace, this is like "cvs tag", and in
the repo explorer on a team stream, which is like "cvs rtag"). You cannot version
one individual resource; you have to release it to the server ("cvs commit") which
will generate a version number for you.
If you want to attach meaning to specific file versions, use release comments.
This will be in tomorrow's build.
Why would you want to tag an individual resource? Do you have a usecase for this?
DM (4/24/01 9:28:02 AM)
1) Not every single person has its own project. JSearch for example is a package inside a
project. We don't version the whole project whenever someone versions or releases his
work. When I self-host on Eclipse I have to move code back and forth. I don't want to write
down what version in Leapfrog or VA/J corresponds to some technical CVS number.
2) Not only top level i.e. project versions and released resources are of interest. Sometimes
you try something out on a class or package and want to keep it for further reuse. Giving
those resources a meaningful name is crucial to find them again.
If I can directly see the release comments directly in the list of CVS versions then this is an
alternative If I have to click through to see the comment then it is not a replacement for the
desired tag on a resource.
As far as I know this is not a CVS restriction - correct me if I'm wrong. Other CVS users would
probably also not understand Eclipses' restriction in this area.
DB (24.04.2001 11:28:49)
I think the use of the word version really confuses CVS users. See Version Management with
CSV document chapter 4 Revisions for a discussion about Versions, revisions, releases and
tags in CVS.
I opt to use the corresponding CVS names whenever we trigger a CVS action instead of using
a different name that means something different in CVS.
KM (4/25/01 12:32:13 PM)
This isn't CVS. Its a front end to support a unified usage model across multiple repository types.
We try to make it as CVS as possible but sometimes we can't.
WRT DM's usage scenario:
We only support explicit versioning at the project level, not at the folder level.
Releasing resources versions them. That's it.
For your usage case, I would suggest separate projects, or branching and merging.
We cannot change terminology on the fly since we must then change the popup menu
in the explorer to not say "Version".
We tried this and its even more confusing when you select a CC project vs. a CVS project
(not to mention what do you put in it if its not yet connected to a repository).
We decided to use a single terminology for all repository types.
Moving to closed.
DB (25.04.2001 19:08:34)
I don't fully agree with your arguments. It is very seldom that a user connects projects to CVS and
CC in the same workspace. Since user either use CVS or CC it would be better to have the corresponding
repository naming instead of coming up with names that mean something different in that respository
system.
Regarding: version only allowed at project see new PR
1GCR0O1: ITPVCM:WIN2000 - Need way to tag source folders, packages and single files in CVS | normal | CLOSED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f0b'} | 39 | I may have logged this previously.
It would be nice to have a preference for the default repository type. This way when for example
creating repository connections the default for the type in the pulldown would be what you want.
Or, remember the last one someone picked.
This is a minor UI nicety, time permitting only.
NOTES:
JEM (5/18/2001 1:28:02 PM)
This is no longer relevant with the new UI
KM (05/26/01 6:51:21 PM)
It still is, we just happen to only be shipping one <g> | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44954dc1c25ebdb1f0c'} | 40 | I would like to be able to connect to a team stream and directly add its content to the
workspace. This could be added to the "New repository wizard".
Currently I need at least 5 steps to populate a new workspace with a team stream:
1. Add repository
2. Open repository browser
3. Expand HEAD (see PR: 1GCQK5Y: ITPVCM:WINNT - Should be able to add HEAD to workspace)
4. Select all children of HEAD
5. Add to workspace
NOTES:
KM (4/25/01 2:02:28 PM)
Moving to future.
BB (4/25/01 12:27:47 PM)
At least for CVS, a team stream cannot be used as a container of projects, so this operation would not
be meaningful for a CVS repository. Add to workspace is not a frequent operation (unless you're self-hosting),
I don't think 5 steps is too bad for this. | normal | RESOLVED | DUPLICATE |
{'$oid': '52e9b44954dc1c25ebdb1f0d'} | 41 | Whenever you checkin something into CVS, the repository assigns a revision number to that file. The revision
number used by CVS is normally not under control of the user (e.g. CVS deceides what revision number to
use). Most of the time the user isn't even aware of the revision number used by CVS.
CVS uses tags to make single files or a collection of files identifyable via a name. Since it isn't always appropriate
to fully tag a project (for example the Java Core project contains 1600 files and we don't want to tag 1600 files to
indentify a revision of a single file) it should be possible to tag packages, files and source folders too.
NOTES:
KM (4/25/01 3:05:26 PM)
Moving to future.
Jean-Michel (14/06/2001 4:18:08 PM)
Use version from workspace after removing the resources you don't want to version. This
operation will only tag the base version you have in the workspace :) | normal | VERIFIED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f0e'} | 42 | In the navigator, I selected A.java and B.java, both of which were versioned.
Right-click->Replace with->Version...
Select a version of A.java. Hit OK.
A.java is replaced with the version.
Now select a version of B.java. Hit OK.
B.java is replaced with the version.
I was expecting to select the versions of both A.java and B.java FIRST,
then have the replacement actually occur. In the case where I am replacing
a large number of files, some of which are very large, it can be
time-consuming to have to baby-sit it while the replacement happens; it would
be more convenient to select the versions I want and then leave it alone
while it loads them.
NOTES: | normal | RESOLVED | INVALID |
{'$oid': '52e9b44954dc1c25ebdb1f0f'} | 45 | We should remove the repo dropdown chooser from the New Stream wizard
when you have the repo already chosen.
The dropdown should only appear when you start the wizard from the "New" menu.
When in the repo explorer, it should be an operation on the repo and then not
show the dropdown.
NOTES:
JEM (10/2/2001 5:03:42 PM)
Right now we can't distinguish these two cases. I.e. the user could select a repo
in the repo explorer, then select New->Other from the file menu, and open the
stream wizard. In that case, should we remove the combo box or not?
JEM (10/10/2001 1:57:19 PM)
I think that attempting to hide the combo in this case would be inconsistent with
other usages. In every other creation wizard, even if the wizard takes its initial
configuration from the selection (e.g. new folder knows the container from the selection),
you still have the ability to change it. This includes when you right-click on an item
and say, for example, New->Folder. It invokes exactly the same wizard. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44954dc1c25ebdb1f10'} | 44 | Some dialogs only show the items that are versioned (i.e. the tag)
and some dialogs show the revision number.
NOTES:
KM (28/04/01 11:13:16 AM)
We'll do a sweep through
and make sure we are using the correct terms everywhere.
KM (21/05/01 10:42:00 PM)
We are just calling everything a version for June. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44954dc1c25ebdb1f11'} | 43 | I have a project (Junk) that has been released to a teamstream.
I rename the project and attempt to release it to the same teamstream.
Compare Failed: Junk does not exist.
Project must be renamed back to its old name and the artifacts left by the first rename that are left must be deleted before the
project can be released. When the project IS released, it shows conflicts on all elements after the compare phase.
NOTES:
KM (5/22/01 9:42:15 PM)
We need to flush sync info on rename.
Jean-Michel (31/05/2001 6:35:07 PM)
Time-permitting. | major | VERIFIED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f12'} | 46 | Meta PR for catchup/release issues.
1GD5YE9: ITPVCM:WINNT - Direction needed from Julian on Catchup/Release usability
1GD2FP9: ITPVCM:WIN2000 - Merge: You don't have to 'save' text for it to be saved
1GD4KC1: ITPVCM:WINNT - Can't revert a change
1GB1UF2: ITPVCM:WINNT - Compare icons, discussion
1GAT32K: ITPVCM:WINNT - Need a better release UI
1GD80I4: ITPVCM:ALL - Should prompt for unresolved changes
Additional notes:
- "Synchronize" instead of catchup/release?
NOTES:
JohnA (5/9/2001 10:04:01 AM)
Following PRs fixed:
1GCE4DM: ITPVCM:ALL - Multiple empty release viewers
1GCWG9F: ITPVCM:WINNT - UI: indicators for items discarded from release
KM (21/05/01 11:16:56 PM)
Decision is as follows:
1. Menu is Navigator->Synchronize
2. Have buttons in view for "Catchup all non-conflicts", "Release all non-conflicts" in that order.
3. Perhaps the later is grayed if there are any conflicts. You can still release individual resources, we just discourage it.
This may be confusing though and people will think its a bug.
4. You cannot release individual conflicts. Instead we put down in the title for the compare on each side a "Take this one" button
(need better name). Idea here is that you encourage/force people to examine the changes and ensure that they are not
picking the wrong one by sheer accident, as can happen now because the menu items were close together.
5. You can only edit the workspace side of the compare, not the repo side.
6. The buttons in 4 are disabled if there are unsaved modifications to the file in the compare. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44954dc1c25ebdb1f13'} | 47 | Subscriber.refresh does beginTask with an empty message (""). This causes it to have
the generic "Operation in Progress" message in the UI. Please provide a more descriptive
message.
NOTES: | normal | VERIFIED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f14'} | 49 | The Subscriber#snapshot method doesn't provide progress monitoring, and
swallows CoreExceptions without logging them. Callers of snapshot seem
to pass in their progress monitors, but don't allocate any "work" for it. They
should pass in a new SubProgressMonitor.
NOTES: | normal | CLOSED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f15'} | 51 | If I select two projects, I cannot compare them to their team stream resources.
I would typically like to perform this operation on multiple projects at a time.
NOTES: | normal | RESOLVED | FIXED |
{'$oid': '52e9b44954dc1c25ebdb1f16'} | 48 | Project references come in three flavours:
1. project name
2. project name + version (+ repo)
3. project name + stream (+ repo)
We currently store #1. While noone makes use of form 2 or 3, we should ensure that being able to store them at some
future time will not require changing our .vcm_meta file format.
This could be solved, for example, by having distinct xml tags for each of these forms.
NOTES:
BB (5/14/2001 4:13:08 PM)
Right now, project references of type 1 are stored as: <project-reference name="..."/>
If we use different tag names for 2. and 3. we should be fine. For example:
<project-version-reference name="projectname" versionname="v1" repository=":pserver:user@cvslnx:/home/cvs/vcmrepo"/>
Moving to Active-TimePermitting as a reminder to add project references of type 2 and 3. | normal | RESOLVED | DUPLICATE |
{'$oid': '52e9b44954dc1c25ebdb1f17'} | 54 | The replace with version dialog shows revision numbers to identify a "version". Since revision number are under control of CVS this
isn't very helpful for the user to pick a version. For a user it would be better to see tags and comments, since this information is provided
by the user.
NOTES:
JEM (5/14/01 2:55:26 PM)
I agree that tags and comments would be helpful. As an aside, you can specify the revision number; it isn't
under the control of CVS. (Managing->Version->One Name).
JohnA (5/14/2001 3:18:51 PM)
James, revision numbers in CVS are not under user control. Those are the labels on files that start
at "1.1" and increment automatically. The version name specified in Managing->Version->One Name
are called CVS "tags".
KM (5/16/01 8:20:47 PM)
I don't understand, it is showing the tags. This is in 107. Has this been fixed since 105?
JohnA (5/16/2001 9:48:18 PM)
It shows tags for versions, but not for files. I plan to change to other compare dialog for files
(the one used when you do file compare with version). | normal | RESOLVED | INVALID |
{'$oid': '52e9b44954dc1c25ebdb1f18'} | 50 | How can I move multiples Project from One repository to another, or from one
Stream to another
Currently I am developping 12 simple plugins examples and I need to 'drop' them
to the OTI
CVS repos.... should I open the property page for each and edit the Stream ???
NOTES:
BB (5/9/2001 4:48:19 PM)
If you want to put project versions into a stream, you can use "Copy
Version To Stream..."
on the stream's context menu. (This action used to be called something
like "Add Version...")
This only works within one repository though, and you cannot add a
project version to the
"HEAD" stream.
Here's how you release a project version created in repo1 into
repo2/HEAD:
If the project doesn't exist in repo2/HEAD, just change the sharing
from repo1 to
repo2/HEAD using the Managing page in the project properties.
Everything will be
an outgoing creation now, and you can release to repo2/HEAD.
If the project does exist in repo2/HEAD, do an "Add To Workspace" from
the
"old" project from repo2/HEAD. After this, you will be in sync with
repo2/HEAD.
Now, "Add To Workspace" the *version* you created in repo1, this will
not change
the sharing relationship with repo2/HEAD. After this, you will have
outgoing changes
you can release into repo2/HEAD. Create a version in repo2 with the
same name as
in repo1 if you want.
CE (5/10/01 10:23:05 AM)
Too much work <g> Isn't there an 'export' kind of action (like in VaJ)
We, in TOR, are going to work on our own repository. Then for the build, we'll
drop into OTT
repository...
JohnA (5/11/2001 12:22:12 PM)
Sorry, the above steps are your only choice. Maybe we'll have export
in version 2.0.
Here is a simplified version of the steps:
1) Version your project
2) Connect to OTT repository, find your project in repo view,
select "Add to Workspace"
3) Select your project in the navigator, choose "Replace With ->
Version"
4) Choose your version from step 1.
5) Select your project, choose "Manage->Release" | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44954dc1c25ebdb1f19'} | 55 | This wizard needs more descriptive text explaining what a
stream is and when to make it - it's too tempting to create them.
This will be very important for people starting (which in part is the purpose of the wizard)
NOTES: | normal | RESOLVED | INVALID |
{'$oid': '52e9b44a54dc1c25ebdb1f1a'} | 52 | To replace a version with a different one from the repsoitory you have to do the following:
- open compare with
- find the right version
- remember the CVS revision number
- open the replace with dialog
- pick the right revision number
==> It would be easier if the replace with version dialog would contain a compare viewer.
NOTES:
KM (5/16/01 6:01:56 PM)
Bonus but nice to have for usability.
Better if you could:
- open compare from history view
- load from history view
- load from comparer (but its a dialog so would need to dismiss dialog, which people would not like). | normal | RESOLVED | INVALID |
{'$oid': '52e9b44a54dc1c25ebdb1f1b'} | 53 | Since the info of whether or not a file has been modified locally is
available (Properties -> Managing then see version with "modified"
following it), we should have a Management Query view (a la VAJ)
which lets us get a list of resources which have been locally
changed.
NOTES:
JohnA (5/13/2001 11:16:38 AM)
Perhaps that would be nice, but I don't think it's going to happen for this release.
Too much extra code to debug and support, too little time. It would also be
difficult to support, because users would expect to be able to multi-select
and release, but that release could span multiple repositories, which we don't
currently have support for.
If you open a release view on your projects, and filter so only outgoing changes
appear, you've got what you're looking for.
JEM (9/10/2001 12:30:35 PM)
Jean-Michel mentioned the idea of having the sync view always show you the
current outgoing changes, as this information is avaiable locally. The view would
update as changes were made to the sync info of the resources. This would
function as a management query view, as it would always show you which
resources have been locally changed, but doesn't provide the same sort of
searching/locating flexibility as items in the search results window, for example. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44a54dc1c25ebdb1f1c'} | 56 | (minor)
Progress dialogs should have useful titles instead of "Progress Information"
NOTES:
JohnA (5/22/2001 4:48:32 PM)
Don't see anywhere that this happens...
JEM (10/1/2001 2:45:56 PM)
It happens everywhere, but we are consistent with the UI, which has "Progress Information"
for the title of all of its progress dialogs. Moving to Inactive. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44a54dc1c25ebdb1f1d'} | 57 | * pserver:[email protected]:/home/cvs/scratch1
* had head stream expanded
* select DaveHelloWorld project
* expand the project
(this is slow)
* go into (DaveHelloWorld)
* SLOW -- Still very slow even though we already know the information because we showed it when we expanded it
* select bin
* go into bin
* go back
* SLOW -- it's really slow showing the DaveHelloWorld even though we have already been there
aside: it's also losing all of my tree expansion etc when going back
NOTES:
GDA (5/14/01 7:56:00 PM)
It's going back to the server too much. This is annoying especially given that there is a refresh
button I could use to force it to update in the event that I thought the data was out of date.
JohnA (5/15/2001 10:45:21 AM)
This is how the drill down adapter works. It sets a new tree input when the user drills. This
loses all child information and expansion state. I suspect it is too late to improve this...
Moving to ITPUI since it's their piece...
KH (5/28/2001 6:54:39 AM)
Moving back to ITPVCM. We will not be changing this piece. You should write your own optimized
code for this case.
KM (5/28/01 10:40:48 AM)
Moving to future | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44a54dc1c25ebdb1f1e'} | 59 | When creating a new repository connection , the wizard should offer to open the repository view
once the wizard finishes
This will help a user discover the various vcm functionality.
NOTES:
KM (5/23/01 8:36:55 PM)
Will add Repo Explorer as sibling to Navigator. This should help.
Moving to future. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44a54dc1c25ebdb1f1f'} | 61 | User 1 & User 2
- install drop 105 into d:\eclipse\105\
User 1
-Java perspective
-Create Java project called GregJProject1
-Create HelloWorld example class with main method that prints to the console
- run the project
-Create repo connection to cvslnxtest /home/cvs/scratch1
-Connect to the repo by expanding the HEAD stream (enter your passwd)
-Select the project. Manage, Release
-Go to Catchup/Release view
- select the project
- choose release (it correctly releases everything)
User 2
- connect to the above repository
- open the repository view
- select the GregJProject1 and add it to the workspace
- run the java main class
- in the navigator select GregJProject1
- Manage -> Release
BUG
- Observe that there are several changes to be release that the user had nothing to do with including
* .vcm_meta
NOTES:
GDA (5/14/01 9:06:11 PM)
This is severe because it forces the users to be releasing files they had nothing to do with
(they are basically system files).
In addition - these files are typically hidden in navigator and packages views.
JohnA (5/15/2001 10:33:47 AM)
Only a bug if it's an outgoing change and there isn't really a change.
Kevinm (5/15/01 7:33:39 PM)
1. there is an existing PR about us doing the write only on resource delta, since its written too often at present
2. Greg doesn't have a .cvsignore so he's also getting all the classfiles.
3. We need our own .vcm_meta compare | normal | RESOLVED | DUPLICATE |
{'$oid': '52e9b44a54dc1c25ebdb1f20'} | 58 | * Create a java project & add a main to it
* Connect to the scratch1 repo on cvslnxtest
* Choose Manage -> Release on the popup for the project
The catchup/release view opens but it is too small to be of use.
This forces a user to always have to resize it.
NOTES:
GDA(5/14/01 8:03:17 PM)
This is very frustrating for users & will need to be changed.
JohNA (5/15/2001 10:37:23 AM)
There is now a managing perpspective, and a user option to automatically switch to the managing
perspective on catchup/release. The managing perspective allocates a large area to the catchup/release view.
Do you think this is a good model? Other than through perspectives, plugins have no say about
where their views first appear, and how much space they're allocated.
KM (5/16/01 10:41:28 PM)
Conclusion from usability was that it should be a fast view by default.
This raised the question of whether we then needed our own perspective.
Switching perspectives on release mess people up.
JohnA (5/17/2001 9:36:18 AM)
Using it as a fast view turns out to be not that great, unless it's a really simple release case.
As soon as you catchup or release one file, a progress dialog pops up, causing the fast view
to close. This is annoying if you want to release mutiple groups of changes, one at a time.
I admit some people are confused at first by it switching to the managing perspective on catchup/release.
However, people quickly seem to get used to it, and find it useful thereafter. If we're just concerned about
initial confusion for new users, the option could be off by default. More advanced users could turn it
on if they prefer that style.
GDA (5/22/01 12:59:44 AM)
This is a bug in the fast view technology. It should be fixed. Check with kevinh.
It is important to make sure problems like this which prevent your view from functioning
as a fast view get reported to the workbench team rather than making an alternative
perhaps less ideal decision.
A separate issue is that the desktop is still missing support to declare this as a fast
view by default. Connect with kevinm. We are trying to push workbench team to add this support.
JohnA (6/1/2001 4:24:44 PM)
Desktop team will not be supplying this functionality. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44a54dc1c25ebdb1f21'} | 60 | User 1 & User 2
- install drop 105 into d:\eclipse\105\
User 1
-Java perspective
-Create Java project called GregJProject1
-Create HelloWorld example class with main method that prints to the console
- run the project
-Create repo connection to cvslnxtest /home/cvs/scratch1
-Connect to the repo by expanding the HEAD stream (enter your passwd)
-Select the project. Manage, Release
-Go to Catchup/Release view
- select the project
- choose release (it correctly releases everything)
User 2
- connect to the above repository
- open the repository view
- select the GregJProject1 and add it to the workspace
- run the java main class
- in the navigator select GregJProject1
- Manage -> Release
BUG
- Observe that there are several changes to be release that the user had nothing to do with
* .classpath
NOTES:
GDA (5/14/01 9:06:11 PM)
This is severe because it forces the users to perpetually be releasing changes
for things they actually did not change
EG (5/15/01 12:49:05 PM)
class path is a system file that captures the shared build path information.
It has to be released so that others can catch-up with it. As for any other
change there is a potentional for conflicting changes to the class path.
This means the .classpath file cannot be just hidden from the user.
The catchup/release view could provide a filter for changes to incoming or
outgoing system files (same as the navigator and packages view provide).
It is important that the filter is overruled in the case of conflicts so that the
user can resolve them.
For now the merging has to be done textually, but eventually the JDT will
provide a class path merge viewer.
Moving to ITPVCM.
Jean-Michel (26/06/2001 11:08:04 AM)
I think that this has been resolved. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44a54dc1c25ebdb1f22'} | 62 | If you have an empty workspace and add a project version to it, the project
version contents are added, but the sharing is not saved (we don't even know
from which repository it came from). This causes the following usability problems:
1. The user cannot determine his project lineup. For example, in VAJ the user
could see the version name in brackets beside the project, even if it was a
version or an open edition: ProjectFoo (v102). Also in CVS, if you update from a
version, the local cvs workspace tracks that you have a version loaded and
not a team stream.
2. To update your workspace with another version or with a team stream contents,
you have to guess from which repo the original contents came from.
There may be other issues as well...
NOTES:
Jean-Michel (5/14/01 11:03:50 PM)
We must soon agree on this PRs priority.
KM (5/15/01 7:30:08 PM)
If we just connected it to HEAD of the repo it came from we would do what most people would expect.
Optionally we could prompt, but that's a bonus.
JohnA (5/16/2001 7:40:03 PM)
Connecting to HEAD makes sense, but it feels icky to be automagically setting sharing when the
user may not want it. Could prompt...
KM (21/05/01 10:30:20 PM)
If someone wants in a different stream they are likely advanced and will know to change it.
Could make a preference with default=connect to HEAD, but this is PRI 2. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44a54dc1c25ebdb1f23'} | 63 | There are several self hosting scenarios where the user must put projects in particular
places and these places are not the same as where the plugins go. A typical approach
is to set the workspace to be on the plugins dir of some target install. If you are developing
a fragment, configuration or component however, you must put these in a different
location as mandated by the runtime.
NOTES:
JM (5/15/2001 1:28:40 AM)
This is critical to true self-hosting.
JohnA (5/16/2001 2:50:06 PM)
Added in UI. The core now needs to hang onto the location before deleting it,
and set it back on creation. Moving to JML for implementation.
Jean-Michel (16/05/2001 6:17:52 PM)
Added vcm core support. John, can you please fixup the ui to look exactly like
the new project dialog. Also, you must pass-in the fully qualified location when
the project description is first created. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44a54dc1c25ebdb1f24'} | 65 | Have at least two repository connections.
Select a project.
Manage->Merge
Select the project and a repository.
Next.
Back.
Select the project and the OTHER repository.
Next.
Notice the contents of the table have not changed.
NOTES: | normal | VERIFIED | FIXED |
{'$oid': '52e9b44a54dc1c25ebdb1f25'} | 68 | "Compare With" and "Replace With" are right next to each other on the navigator menu.
It is very easy to hit "Replace With"t when you meant "Compare With", losing your work.
NOTES:
EG (5/17/01 11:09:52 AM)
these actions are contributed by VCM - moving PR
JohnA (5/17/2001 9:55:31 AM)
I can see the next PR already... "Compare With and Replace With appear in different
parts of the popup menu. Since these items are related, doesn't it make sense to put them
together? I'm always searching around the popup menu looking for the action I want."
But seriously, we're looking into the fact that replacing doesn't maintain local history.
Once it does, you won't lose work if you hit "Replace With" by accident.
Also, these menus are contributed by the compare plugin, not VCM.
SN (5/18/2001 3:12:48 PM)
Isn't it easier to not lose work? | normal | VERIFIED | FIXED |
{'$oid': '52e9b44a54dc1c25ebdb1f26'} | 66 | There needs to be more options in the context menu for the Catchup/Release
view. Specifically:
1). Show Resource in Navigator
2). Discard outgoing change (from Release view)
My use case is as follows:
- make a bunch of changes
- select a parent resource and choose "Release"
- in the Catchup/Release view notice that there is a change
that you made that you don't want
The only way to get the server version instead of your own
is to navigate manually in the Navigator to that particular
resource and then select Replace with Team Stream Resource.
Also note that if you do this the resource (obviously) remains
in the list of resources with differences and you have to re-do
the Release operation to get an updated view.
NOTES:
BB (5/16/2001 6:41:06 PM)
There are two possible solutions:
1. enable "catch up" even if there is no incoming change.
2. add a "Replace with Team Stream" action.
My preference is 2., and I strongly vote for "Catch Up" to be a safe operation
which only gets unopposed incoming changes. | normal | CLOSED | FIXED |
{'$oid': '52e9b44a54dc1c25ebdb1f27'} | 67 | The api documentation should be updated to mention that if the destination resource is a project
it is expected to exist.
NOTES: | normal | RESOLVED | INVALID |
{'$oid': '52e9b44a54dc1c25ebdb1f28'} | 69 | It would be nice if VCM provided the option of organizing projects in a
repository into a
hierachical folder structure. Presently, all projects in a repository are
presented (and
stored) at the same level in the directory tree. Under CVS, there is no reason
why
projects could not be broken down into folders without having to create a
repository
for each.
This would be useful when a given macro-scale story consists of a number of
related Projects with different build options etc... (since these options are
specified on a per-Project basis).
This would also help organize mixed-mode development (Web & C++ & Java) into
related blocks.
NOTES:
My current use for this feature would be to create a little "Jeff" folder in the
development repository to hold all of my private sandbox Projects. The current
solution is to give each Project a name like "Playground_Jeff_<project name
here>".
JB (22/05/2001 12:09:27 PM) | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44a54dc1c25ebdb1f29'} | 70 | We need to test that removing a repo connection from the Repo Explorer does the "right"
thing for projects which are sharing with that connection.
We should search for projects which share it and warn user.
If user accepts to proceed, then the projects which are sharing loose their sharing.
There used to be bugs where the repo connections would show up again. May have been fixed though.
NOTES:
KM (5/19/01 2:36:47 PM)
Item for Olivier?
JEM (9/5/2001 11:30:12 AM)
Hasn't this been done?
JEM (9/6/2001 4:31:13 PM)
Repo removal detects projects using it and just fails gracefully with detailed error message.
It might be good to ask to proceed and then lose the sharing. | normal | VERIFIED | FIXED |
{'$oid': '52e9b44a54dc1c25ebdb1f2a'} | 71 | CVS, SWT
We have several team streams, i.e. we are not just releasing stuff to "head".
Therefore, when we are catching up to - or particularly when we are releasing
- we need to be able to see the name of the team stream.
The reason this is so necessary is that if we happened to make an error setting
the management stream for a project, we would get no notification that we were
releasing into the wrong stream.
Would be nice if the repository name was also included. (We can envision
a scenario where you could easily accidentally release code into the wrong
repository...)
NOTES:
KM (5/18/01 5:36:49 PM)
Has merit but not sure of time/priority.
Jean-Michel (8/23/01 10:21:34 AM)
I also find this very confusing and we should show more context in the sync view (i.e. repo and streams). | normal | VERIFIED | FIXED |
{'$oid': '52e9b44a54dc1c25ebdb1f2b'} | 72 | Would be really nice to have more distinction between catchup & release.
A couple of times now, I have started a "catchup", only to think "was I catching up or releasing?"
I have to stare at the arrows...
I think the UI should help a bit more, to indicate what the user is currently doing.
Please change the title of the Catchup/Release view to "Catchup" if the user said "Catchup",
and "Release" if the user said "Release".
NOTES:
KM (5/18/01 12:15:15 PM)
You can do either from the catchup/release viewer. Making it just one or the other made it more frustrating
for people. Presumably the icons overlays should be sufficiently clear to indicate if its incomming (catching up)
or outgoing (release).
The only thing I can think of is to change the filters to changing the 'mode' of the viewer from either catching up or releasing,
and then change the title and menus. Unlikely we'll change this for June unless we find more people are confused by it.
CM (5/18/2001 2:40:54 PM)
I understand now - I was confused because the "other" menu always seemed to be disabled,
so I thought the operation was actually modal. (i.e. when you say "Catchup", the "Release"
menu is disabled - and vice-versa). But I see that the disabled menu was just because of
the desired direction of the particular item... i.e. it just happened that I haven't yet tried to
release something without catching up first, so I never noticed that it was possible to do both
at the same time.
If we are allowing both, then the initial menu item should not be "Catchup..." or "Release...".
There should only be one menu item, and it should have some name like "Synchonize..." or something.
I think I prefer the "modal" thing, i.e. first ya catchup, then ya release. (less confusing/dangerous)
Anyhow, I will try it for a while, and see if it grows on me <g>.
CM (5/24/2001 1:17:15 PM)
I still find this confusing, and I have watched SN become completely confused also.
I honestly think it is dangerous to allow both, and I still really prefer the "modal" Catchup or Release
that we had in LeapFrog. If I had not been standing there watching SN, he would have released a
whole bunch of old code into our good team stream instead of catching up. He had said "Release..."
to get to the browser because he wanted to see what he had changed. Then he decided to catch up
because the number of changes was significant. He popped up the context menu on the root of
the changes tree, and hovered over the "Release" item, and (since I had just released a ton of
changes that I did not want overwritten) I said "Ahhh! Noooo!" and he realized his error and chose
"catchup" instead. We were nanoseconds away from wiping out a bunch of changes.
I guess it's the "cognitive overload" thing, that particularly hits people at 3am (that's what time
it was, and both of us were tired). But that's when a dangerous error like that is particularly
harmful, because then you have to stay even later to clean up the accidental mess... if you
even realize that you made a mess... <g>).
JohnA (5/27/2001 3:49:01 PM)
I think it's worth noting here that "Catchup" from the catchup/release only applies to incoming changes and conflicts,
"Release" from the catchup/release view only applies to outgoing changes and conflicts.
In SN's case, it's impossible to accidently release when there's an incoming change, thus wiping out
all those incoming changes. The only thing it could have released were conflicts. However, if there were
conflicts in the selection, a warning dialog would having popped up saying "You're about to overwrite stuff
on the server that conflict with your changes, are you sure you want to do that?".
Once in the catchup/release view, you should think of "Catchup" and "Release" as safe operations... the
only thing they do is catchup or release non-conflicting changes in whichever direction. When there's a conflict,
it will warn you before releasing.
For example, in the most common case you have a project with several incoming
changes and several outgoing changes, without conflicts. You can select the project and hit "catchup". This will
load just the incoming changes. You can then compile and test to make sure the incoming stuff did interact badly
with your local stuff. Then, select the project and hit "Release". This will release only the outgoing changes.
If the server or workbench has changed since your last refresh, or if there were conflicts you didn't notice, it will
always prompt you to make sure you're aware of what's about to happen. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44a54dc1c25ebdb1f2c'} | 73 | If I multi-select projects to Add to workspace it asks for the location of each project.
I would expect a dialog in the beggining saying something like:
"Would you like to use the projects default location ? Yes/No"
If yes, it would act like version 0.107 and create every project in the default location.
If no, it should ask the location for *all* projects *before* going to the server. Imagine
how annoying it can be in a 28K modem conection.
NOTES:
RTP (5/18/01 9:18:57 AM)
It would be a nice feature if the vcm plugin could persist this location information and,
for example, if I delete a project and try to add it again, it already points to the last known
location as default. And give the option to change if wanted.
KM (5/18/01 12:11:50 PM)
We should fix the annoying bit. Remembering old locations is out for June.
KM (5/18/01 6:18:23 PM)
We should prompt for single location for ALL projects being loaded.
If you want to load and store in multiple locations, you load each project individually.
See also
1GDVVOM: ITPVCM:WIN2000 - Dialog to specify location for project truncated
KM (21/05/01 10:36:35 PM)
For Olivier?
KM (5/22/01 12:53:41 PM)
Spiffy upgrade to the prompt once story:
Have a "Don't show me this dialog again" checbox on the prompter,
and have a toggle on our preferance page where you can turn it back on.
JW (5/22/2001 9:27:06 AM)
This presentation is troubling. Although there are cases when a user wants to locate the project, this is rare.
Making this choice available at all times makes it hard to use. We need some way to avoid this
query from arising usually.
JohnA (5/22/2001 3:34:34 PM)
There is now a user preference, off by default, for prompting for location on load.
I still like the idea of prompting for each project, especially now that it's an optional thing.
However, as Rodrigo says, it should do all prompting before starting the load.
JohnA (5/23/2001 10:58:45 AM)
Moving the prompting to before the load is not as important... moving to LowPriority. | normal | VERIFIED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f2d'} | 74 | While releasing, I type text into the left pane of the compare window
expecting that the new text will get released, but it does not.
This happens when you see that you are about to release bad code
or a bad comment and you attempt to fix it. Also, instead of accepting
and merging changes, you sometimes want to write a 3rd piece of code
that is based on the other two. It's tempting to type that code into this
window.
It might be better to make the text widgets read only?
NOTES:
AW (18.06.01 17:05:25)
I agree, when releasing the left hand side of the synch view should be made readonly.
Otherwise one can accidentally release untested code to the team.
Moving to ITPVCM for comment.
KM (6/18/01 12:56:09 PM)
Intended usage is you can indeed modify the text on the left and then save it.
This modifies your local resource. You can then choose to release it.
I assume this PR originated before we improved this. Steve's first part
of his scenario should now work as described.
As for AW's comments, if you want to release untested, that's up to you.
We don't support making a new resource from the sync view and
doesn't feel like it fits the intended usage pattern.
JEM (8/14/2001 10:29:52 AM)
When the user releases with unsaved changes we should warn them and ask if they
want to save prior to releasing. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44b54dc1c25ebdb1f2e'} | 75 | A catch up reported conflicts although I have never touched the corresponding files in my workspace. Comparing them
revealed that the are the same. I then catched up with the conflicts. Doing another catchup displayed the same files
still as conflicts.
I have zipped my workspace if helpful.
NOTES:
KM (21/05/01 6:01:57 PM)
Since it sounds repeatable, the zip would be very helpful! Where can we get it?
Jean-Michel (29/05/2001 11:23:48 AM)
I have not had time to try and reproduce yet.
TPE (2001-05-30 13:27:41)
The workspace is now in ftp://hubba/INCOMING/forVCMTeam/aGE2HV8.zip
TPE (2001-06-01 10:44:05)
Have not reproduced the problem yet -- looking into it. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44b54dc1c25ebdb1f2f'} | 76 | If you pick a project its "YourProject in HEAD in repo..."
but if you pick a resource its "YourProject/SomeResource in repo..."
NOTES:
JohnA (5/22/2001 4:34:36 PM)
Hmm, this is because files in the repository view are now file version handles, not
team stream handles. This was done to be able to get the version name without a
round-trip. File version handles don't know what their team stream is, do they? | normal | RESOLVED | INVALID |
{'$oid': '52e9b44b54dc1c25ebdb1f30'} | 77 | When you compare a Java project with the team stream version,
the resulting editor should have an "Expand All..." menu option
(like the Catchup/Release view) so you can quickly drill down
through folders to get to your changes.
NOTES:
KM (5/22/01 3:51:02 PM)
Is this us or compare framework?
AW (01.06.01 10:54:34)
It's you. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f31'} | 78 | The repositories view is useless when working with more than one team stream.
After I added the three SWT team streams I had no idea which project actually is in what
team stream. This is very error prone and could be avoided by a better UI.
If I select the Managing page of the project properties it knows which team stream the project
is in. It should be possible to filter out projects that are not actually in the team stream
instead of providing the "raw" CVS view.
Also, I would expect to be able to right click on a team stream and choose Catchup and
Release.
NOTES:
KM (5/22/01 3:51:56 PM)
This is how CVS works. Future.
Jean-Michel (8/1/01 4:30:27 PM)
What we should have is a CVS specific preference or an extra menu item
in the repositories view.
No matter how the user configures a deep refresh when querying for
projects in a stream, the code below will do the right thing:
/**
* Returns the project names in the given stream.
*
* There is no easy way to determine if a CVS project is contained
* in a stream. To do so, you have to look at each file in the project
* and check if it tagged with the stream name. By default we simply
* return all projects in HEAD, but if the user would like, we can perform
* a refresh infinite on each project to correctly determine if a
* project has a file in that stream.
*
* [Note: Running <code>fetchRemoteTree</code> with depth zero will
* not work since each file in the project has to be examined to
* determine if the project is contained in a stream.]
*
* [Issue: Why do we remove CVSROOT? This will force CVS users
* to use the command line to modify administration files. We should
* instead allow these to be edited within Eclipse.]
*/
public String[] fetchTeamStreamProjectNames(
String teamStreamName,
IProgressMonitor monitor) throws VCMException {
String[] topLevelModules =
internalFetchMemberNames(Path.EMPTY, teamStreamName, monitor)[0];
List result = new ArrayList();
result.addAll(Arrays.asList(topLevelModules));
result.remove("CVSROOT");
// For each project see if we can get the resource tree
if( /* option enabled via CVS preferences dialog */ true ) {
if (!teamStreamName.equals("HEAD")) {
for (Iterator it = result.iterator(); it.hasNext();) {
String projectName = (String) it.next();
IServerWorkspace workspace = getTeamStreamProject(teamStreamName, projectName);
try {
IServerResourceTree tree =
fetchRemoteTree(
workspace,
Path.EMPTY,
IResource.DEPTH_INFINITE,
monitor);
if (tree.getResource() == null) {
it.remove();
}
} catch( CoreException e ) {
// An internal error occured
throw new VCMException(VCMStatus.ERROR, null, "fetchTeamStreamProjectNames", e);
}
}
}
}
return (String[]) result.toArray(new String[result.size()]);
}
Jean-Michel (9/14/01 1:36:31 PM)
Fixed, released into HEAD. We still need a little work on the filter
wizard so that it can be cancellable and have the correct sizing. | normal | VERIFIED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f32'} | 79 | AK (5/23/01 10:45:38 AM)
i had ca. 1700 outgoing changes to release.
after releasing i had to wait for like 5 minutes (sic!) to get the tree updated
and the ide is frozen during that time (+ 100% CPU usage)
NOTES:
KM (5/23/01 3:10:20 PM)
It sounds like what you need here is a "forced release" that overwrites the server without
providing any compare? (You could never manually compare 1700 resources anyway).
Would you expect it to inform you of conflicts or not?
KM (5/23/01 6:24:57 PM)
From Adam:
i guess i should explain better what happened
_after_ i'd released everything it took forever to just update the tree viewer (maybe it runs the compare engine again(?))
KM (5/23/01 6:26:19 PM)
Moving to future. Insufficient time left for looking at performance of this nature. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f33'} | 81 | Platform: JDK 108 W2K
If you import a large file into a project, then release it to a repository, attempting to cancel
the operation does not interrupt the transfer of the large file.
TPE (2001-05-29 13:57:21)
Need to check if it is safe to interrupt large file transfer (to ensure the server does
not interpret partial results incorrectly). | normal | RESOLVED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f34'} | 82 | Platform: JDK 108 W2K
If you create a folder and release it to the repository, then cause a conflict by,
say, adding a member _and_ modifying it in the repository, the catch-up/release
browser shows that the folder is in conflict (the red diamond) but givs no indication
about why or how. The only action is to release local state or accept the repository
renditioning. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f35'} | 80 | Delete key in repo view should:
when a repository connection is selected, remove the connection
when a stream other than HEAD is selected, remove the stream
otherwise do nothing
NOTES: | normal | VERIFIED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f36'} | 85 | 1) Get a java class from the CVS repository
2) Modify the class.
3) Compare with the Stream resource
4) Click on "Copy current change frm left to right" for a few of the changes.
5) Close the "Compare C <a> and <b>" tab.
6) You are prompted to save the changes - say yes.
7) Go back and look at class in packages view - the changes you said to save have not been saved.
NOTES:
AW (05.07.2001 10:15:23)
Duplicate of 1GF0UT6: ITPVCM:WIN2000 - Compare with each other: saves files?
AW (30.07.2001 15:06:00)
fixed in 0.127
TC (9/10/2001 4:22:52 PM) - when you copy contents over and then close the compare viewer you are atill
not prompted for change confirmation
STEPS
1) Create a .java file with the following contents
public class TestClass {
public static void main(String[] args) {
System.out.println("Line1");
System.out.println("Line2");
}
}
2) Release it to a Stream
3) Delete the second line and save it
4) Synchronize with stream
5) Copy over the change with the button
6) Close the browser - no warning and changes are lost
AW (11.09.2001 09:54:01)
Compare with Stream is VCM functionality.
I've fixed "Compare With Each Other"
KH (9/13/2001 9:57:39 AM)
Moving to VCM based on comments from AW.
JEM (9/13/2001 2:56:01 PM)
Related to
1GJW3AI: ITPVCM:WIN2000 - Compare with Stream Contents does not prompt for save changes. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f37'} | 64 | The comment field can get very long (e.g. see the Crystal project in sourceforge) and verbose. Users actually
use this field often. The history view has the following problems when showing the comment:
1. The comment is truncated (high)
2. We don't honor the cariage returns so the user's formatting is lost (lower priority). We could possibly
show the formated comment in a tip or add a context menu option to Browse Comment.
We should also consider the other places that show comments.
NOTES:
Jean-Michel (15/05/2001 9:52:49 AM)
See also: 1GD1WU0: ITPVCM:ALL - CVS: do we need validation for release comments?
JohnA (5/15/2001 7:21:04 PM)
1. Are you sure we're truncating? I tried it, and if I expand the column, I get more of a comment.
2. Can't help -- table cells only get one line.
Kevinm (5/15/01 7:27:08 PM)
The multi-line is interesting since we only prompt for single line. Do people often make them multi-line?
Jean-Michel (8/23/01 10:20:40 AM)
Yes, this is really bad for OS projects because most of their comments are multi-line. You basically
cannot see these comments from within Eclipse. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f38'} | 83 | Platform: JDK 108 W2K
Canceling an add to workspace operation leaves the work-accomplished-to-date in the workspace.
I expected it to cancel the entire menu operation (i.e. not to load anything)
NOTE:
KM (5/23/01 2:59:18 PM)
I think this is the best we can do. As long as that which was loaded has intergrity we are ok.
(e.g. project is setup right with .vcm_meta, parent folders in place, sync info correct for all).
Does this match what you've observed?
TPE (2001-05-25 15:03:12)
The partial results are correctly represented, but the load does not make sense in Java-terms. | normal | RESOLVED | WONTFIX |
{'$oid': '52e9b44b54dc1c25ebdb1f39'} | 84 | If I create another stream, it automatically contains all projects and folders. Can't we
hide projects that don't have a tag for that stream? That way it would only show
resources that have been explicitly released to that stream.
NOTES:
KM (5/23/01 7:49:35 PM)
Actually as it stands I find it totally confusing. As do the SWT folk who are actually trying to use streams.
You have no idea what's *really* in the stream.
Jean-Michel (5/25/01 2:10:09 PM)
There is no foolproof way of doing this with cvs. We can try and be
smart (e.g. it will work if the project directory has at least one file) but
that is it. Maybe this should become an option?
I suggest that we investigate after release?
JohnA (5/27/2001 12:31:30 PM)
How about: if it does contain a file (covers 99% of cases), we look for stream tag and
filter appropriately. If it doesn't contain any files, we include it in all streams.
If this isn't too much work, I think it's worthwhile to do.
BB (5/29/2001 4:15:48 AM)
I'm not sure about the 99%. If you don't put class files in a "bin" directory, you can easily
have a project that has its first file in the third level down (org/eclipse/vcm/...)
Jean-Michel (29/05/2001 9:44:02 AM)
Moving to time-permitting...
Jean-Michel (9/14/01 1:37:10 PM)
Dup of: 1GE558R: ITPVCM:WINNT - Confusing repositories view | normal | VERIFIED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f3a'} | 86 | I feel that catchup/release mode is a wishy-washy name for that mode, and it doesn't
make it very distinct from the other modes. I think we should change it to
be called "Power User Mode". This better expresses the reason for the mode, encourages
novices to stay away from it, and is a much catchier name. This is also a term
that already has some weight to it since Windows also has a "Power User" setting.
NOTES:
JohnA (6/7/2001 3:47:00 PM)
Not likely to be approved for June, but possibly a future item. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f3b'} | 89 | Connect to a windows cvs server
Create a new team stream.
The error appears as "team stream already exists"
1) It should be capitalized: "Team stream already exists"
2) It should be more verbose; in fact, on windows servers you
can only have one team stream. We should either differentiate
these two cases or be more helpful in the error message.
NOTES:
KM (5/28/01 9:08:35 PM)
Fixed typo.
Don't see what we can do better. Don't want to write CVSNT specific code.
The error occurs if
adapter.fetchTeamStreamNames(monitor)
contains the name of the stream to be created. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44b54dc1c25ebdb1f3c'} | 90 | If you select a project in another stream (one that has actually been released into that stream)
in the repo explorer, all the properties in the property sheet save "name" are blank
(eg. author, comment, etc.).
NOTES: | normal | RESOLVED | INVALID |
{'$oid': '52e9b44b54dc1c25ebdb1f3d'} | 87 | Check menu separators and ordering for pop in catchup/release.
NOTES: | normal | RESOLVED | INVALID |
{'$oid': '52e9b44b54dc1c25ebdb1f3e'} | 88 | If the user selects to run a release/catchup on files that are modified, but not saved yet, should we prompt the user that
the files should be saved first or help him save them?
NOTES:
KM (5/23/01 10:23:24 PM)
Also there is an existing PR suggesting we should refresh from local too.
KM (05/27/01 3:19:55 PM)
Future. This is equivalent to prompting that they save before run/debug (which we don't). | normal | RESOLVED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f3f'} | 91 | My workspace was built by KH.
I caught up, loaded everything.
Changed a few files.
Changed the project property to use my name and password instead of KH.
Selected the project->mananging->catchup in the navigator.
The catchup view shows that everything is different but I changed only 3 or 4 files and
no one released other version, ie, the files are the same.
NOTES:
KM (5/24/01 11:05:24 AM)
I assume: we are throwing away the connection, so throwing away all the sharing info.
We could fix by allowing editing of the login name. | normal | VERIFIED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f40'} | 93 | 1. Enter an invalid password when logging into a repo.
2. Click cancel.
3. Discard the repo location.
4. Create another connection to the same repo, the old password is used and I have no chance of changing
my password. The only way out is to manually delete the key ring file.
We need a way of forcing a login to a server, even after either cancelling the login or entering a wrong password.
NOTES:
KM (5/24/01 2:52:12 PM)
This seems to work now.
Also, UI should support a preference page where all the keys are shown and they can be removed.
Moving to Future as reminder to UI to provide this support. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44b54dc1c25ebdb1f41'} | 92 | Several parts of the CVS UI directly reference internal CVS core stuff.
The UI should only be interacting with a clearly defined API, not referencing
anything else in core. Our convention is that the ".base" packages are API
between VCM core and VCM UI.
This should be easy and safe to fix, it may just involve moving a couple of
classes around.
NOTES:
KM (5/31/01 2:14:13 PM)
Future. We need an entire walkthrough of our APIs this summer.
For example, I suspect there is code in the UI that should be in the core. | normal | RESOLVED | INVALID |
{'$oid': '52e9b44c54dc1c25ebdb1f42'} | 94 | When browsing files in the Repositories View, if you try to open a file it will
always open in the text editor even if it is not associated to it. The view does
not behave like the navigator - it should open the file using its associated
editor, if it is not associated, it should open using its system default viewer.
Steps to repeat:
1) Open the managing view (Perspective->Open->Other...->Managing, OK)
2) Connect to any CVS repository location
3) Browse for a file that has an external viewer, (not associated to an editor,
i.e: *.gif, *.xml)
4) Open the file, notice even a .gif file will open in the editor
To see how the Repositories View should behave, try opening a .gif file from the
Navigator - it will launch your gif viewer (IExplorer by default).
NOTES:
KM (5/24/01 4:38:51 PM)
In general you cannot open a specialized editor on a remote resource because they can sometimes
be dependent on other resources and expect these to be local. For example, an HTML page which
references a graphic, the HTML editor is going to try to look for the graphic to display it.
You can't be sure that the editor was written to accomodate remote or 'missing' resources.
We discussed this some time ago and felt it was safer to just use vanilla editors for now.
See also 1GE6I1Q: ITPJUI:ALL - Opening repository resources doesn't honor type | normal | RESOLVED | DUPLICATE |
{'$oid': '52e9b44c54dc1c25ebdb1f43'} | 95 | JDK 110 W2K
Was working in cvslnx.ott.oti.com:/home/cvs.vcmrepo on my laptop.
Copied the entire workspace to a desktop machine and did some work with it there.
Then went back to the laptop (note: didn't copy the workspace back) and went to catchup.
Compare works ok, but continually getting CVS errors when trying to catchup with a change.
Retrying the operation does not help.
The workspace (without JRE) is in:
ftp://walleye/va2000/INCOMING/forVCMTeam/1GEA876_wks.zip
A screen shot of the problem message is in:
ftp://walleye/va2000/INCOMING/forVCMTeam/1GEA876_scr.jpg
Steps to recreate:
Take the workspace and try to catch up with either
org.eclipse.vcm.core
or org.eclipse.vcm.core.cvs
Jean-Michel (29/05/2001 11:22:19 AM)
I have not had a chance to look into this problem. Have you debugged
it at all Tim? | normal | RESOLVED | INVALID |
{'$oid': '52e9b44c54dc1c25ebdb1f44'} | 96 | Subscriber.getAdapter(IResource) should throw VCMException because it does
return getRepository(resource).getAdapter();
and getRepository(resource) can legitimately return null.
Throwing an exception here would be more helpful than an NPE
and is a common code path.
See also
1GEGR0U: ITPVCM:WINNT - NPE refreshing SyncView with deleted project
NOTES: | normal | RESOLVED | INVALID |
{'$oid': '52e9b44c54dc1c25ebdb1f45'} | 97 | There is an option for turning on prompting for project location on add.
Default is "don't prompt". Its not evident that this option exists.
NOTES: | normal | RESOLVED | INVALID |
{'$oid': '52e9b44c54dc1c25ebdb1f46'} | 98 | We need a "New" button in the repo connection dialog.
Otherwise its not clear how you get a new one when first time you release a
project and the list comes up emtpy.
NOTES:
KM (5/31/01 8:59:26 PM)
We'ld require a wizard that had you choose the repo as its first page,
or
if there was only one type did not display the first page. | normal | RESOLVED | DUPLICATE |
{'$oid': '52e9b44c54dc1c25ebdb1f47'} | 101 | merge wizard
new cvs connection
new team stream.
NOTES:
JEM (6/1/01 5:41:48 PM)
Also ignore preference page.
JohnA (6/2/2001 3:04:42 PM)
Added the code hook for the preference page.
JohnA (6/13/2001 1:06:19 PM)
Moving to future. Need to reconsider help for wizards when we have time. | normal | RESOLVED | FIXED |
{'$oid': '52e9b44c54dc1c25ebdb1f48'} | 100 | It would be nice to be able to lock the teamstream or the project being
released into it for the duration of the release process.
This would be intended to prevent another user from catching up with some
inconsistent state.
NOTES:
KM (5/29/01 3:04:51 PM)
Yes, we'ld like as well!
Not supported in June (and not supported in all repos).
BB (6/4/2001 8:19:02 AM)
Note that CVS does not support locking of a repository at all. Fallback
is social protocols... | normal | RESOLVED | INVALID |
End of preview. Expand
in Dataset Viewer.
No dataset card yet
New: Create and edit this dataset card directly on the website!
Contribute a Dataset Card- Downloads last month
- 22