Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.
Bug 376988 - Cleanup auto tagging reports in 4.2 primary builds
Summary: Cleanup auto tagging reports in 4.2 primary builds
Status: RESOLVED FIXED
Alias: None
Product: Platform
Classification: Eclipse Project
Component: Releng (show other bugs)
Version: 4.2   Edit
Hardware: PC Linux
: P3 normal (vote)
Target Milestone: 4.2 M7   Edit
Assignee: Paul Webster CLA
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2012-04-17 11:25 EDT by David Williams CLA
Modified: 2012-04-25 17:09 EDT (History)
3 users (show)

See Also:


Attachments
report from I20120417-1000 (4.46 KB, text/plain)
2012-04-17 11:25 EDT, David Williams CLA
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description David Williams CLA 2012-04-17 11:25:00 EDT
Created attachment 214115 [details]
report from I20120417-1000

As first reported in bug 375807 comment 58, 
the current "reports" from auto tagging have a few stray 
"change directories" listed, along with the "changed repositories". 

Attached is similar example from 
I20120417-1000

See bug 375807 comment 60 for reason why this happens (assuming there's a hint for a fix in there somewhere :)
Comment 1 Markus Keller CLA 2012-04-24 08:51:54 EDT
Remaining problem I see:

1. Bad folders in "The following projects have changed:" section:
The problem is that there are two common git repository layouts, e.g.:
- eclipse.platform.ui/bundles/org.eclipse.ui.ide/<project-contents>
- eclipse.jdt.ui/org.eclipse.jdt.ui/<project-contents>

=> To distinguish the two, you either need a hardcoded list, or you need to inspect the repos and see on which level there is a .project file. The fully correct strategy to find the project from a full path to a file is to walk up from the file and find the directory that contains a .project file.

=> As an intermediate solution, better err on the side of showing too many prefix folders. Including the repo also wouldn't hurt, e.g.
- eclipse.platform.ui/bundles/org.eclipse.ui.ide
- eclipse.jdt.ui/org.eclipse.jdt.ui/ui
- eclipse.jdt.ui/org.eclipse.jdt.ui/ui refactoring
- ...


2. Bug summary should not escape HTML entities (&apos;, &quot;, &gt;, etc.)

3. First section header says "The build contains the following changes:". This is too bold a statement, especially if the list contains open bugs (which e.g. originate from a commit message "workaround for bug 123").

=> I would change this to "Commit messages mentioned these bugs:".
Comment 2 Markus Keller CLA 2012-04-24 11:52:46 EDT
Paul mentioned on the mailing list that the bug list is a bit too much for a summary mail. I agree with that.

If others want to keep the bug list, then please move the bug list to the end of the mail.
Comment 3 Paul Webster CLA 2012-04-24 13:42:41 EDT
(In reply to comment #2)
> Paul mentioned on the mailing list that the bug list is a bit too much for a
> summary mail. I agree with that.

Sorry, I meant the commit comments would be too much for a summary email (I don't need to see a git log integration..<last_build>), as there are often multiple commits for one bug.

I do expect to see the list of bugs mentioned in the commit messages.  I'd be fine with putting them at the end, and changing the sentence at the beginning of the section.

PW
Comment 4 David Williams CLA 2012-04-24 13:44:32 EDT
I'm going to assign this to Paul, for now. I (in our SDK builds) just use the tools he provides from elsewhere ("e4 repo"?) ... so, if he wants to fix those, fine ... if it is left to me, it would be a while before I'd look at it. 

If you'd prefer, I could just leave out the report and just send a message that "build has started" .... but ... seems to me its useful to have a report? 

While its not currently, the idea is to also link to it from the builds download page. I suspect too, we should well emphasize "since previous build", and maybe even reference it "Since previous build: ${oldBuildId}" since sometimes it might have just been from a few hours previously, sometimes a week or two. 

If the bug list is moved to the end, the first part would just be "repositories changed", right? (Just wanted to be sure I wasn't missing the point).
Comment 5 David Williams CLA 2012-04-24 13:45:10 EDT
Paul, feel free to move back to inbox (or me) if you definitely won't be working on this.
Comment 6 Paul Webster CLA 2012-04-24 14:56:53 EDT
http://git.eclipse.org/c/e4/org.eclipse.e4.releng.git/commit/?id=020535f83b81375fa99eb5eb379803f71a125fcb

I've swapped the sections around and updated the project detection script.  That should address #1 and #3 from comment #1.

#2 is what curl -k https://bugs.eclipse.org/bugs/show_bug.cgi?id=${BUG}\&ctype=xml returns.  In theory after we parse out the fields we care about, we would need to run it through some html to text translator.  If anybody has any ideas, let me know :-)

A report on e.jdt.ui and e.platform.ui from I20120416-1508 to master now generates the following:


The following projects have changed:
bundles/org.eclipse.core.commands
bundles/org.eclipse.e4.core.commands
bundles/org.eclipse.e4.ui.workbench
bundles/org.eclipse.e4.ui.workbench.addons.swt
bundles/org.eclipse.e4.ui.workbench.renderers.swt
bundles/org.eclipse.ui
bundles/org.eclipse.ui.ide
bundles/org.eclipse.ui.navigator
bundles/org.eclipse.ui.workbench
org.eclipse.jdt.text.tests
org.eclipse.jdt.ui
org.eclipse.jdt.ui.tests
org.eclipse.jdt.ui.tests.refactoring
tests/org.eclipse.e4.ui.tests
tests/org.eclipse.ui.tests

The build contains commits that mentioned these bugs:
+ Bug 86168 - [type wizards] Allow the creation of a compilation unit called package-info.java (FIXED)
+ Bug 130922 - [quick assist] &apos;Assign parameter to new field&apos; must not make field final&apos; (FIXED)
+ Bug 184056 - [hovering] Tool tip on closing curly bracket (FIXED)
+ Bug 200847 - [templates] Variable resolver resolves to array of specified type (FIXED)
+ Bug 243912 - IProject#delete(..) should suppress deltas from CharsetDeltaJob (FIXED)
+ Bug 262593 - [Commands] Show in package explorer shortcut not working in linux (FIXED)
+ Bug 335132 - [actions] Override / Implement dialog size / position are not saved (FIXED)
+ Bug 335249 - [preferences] Improve grouping of &quot;Potential programming problems&quot; on Errors/Warnings page (FIXED)
+ Bug 336062 - [Dialogs] Open Resource dialog should have &quot;Show In &gt; ...&quot; context menu (FIXED)
+ Bug 348720 - [DND] Splitting self in shared area causes an SWTException (FIXED)
+ Bug 352380 - [Compatibility] Empty &apos;Team&apos; menu shows up in context menus (FIXED)
+ Bug 352785 - [extract method] Refactor &gt; Extract Method with break statement in switch causes NPE (FIXED)
+ Bug 353126 - improve StubCreator (FIXED)
+ Bug 353472 - [override method] generating a method override should copy annotations (FIXED)
+ Bug 357997 - [type hierarchy] Type Hierarchy View has button bar painting issues on workspace startup (FIXED)
+ Bug 360639 - [KeyBindings] shortcut for redo does not follow standard linux shortcuts (FIXED)
+ Bug 369159 - [Compatibility] ICommandService/IExecutionListener not fired (FIXED)
+ Bug 372992 - EditorPart#isSaveOnCloseNeeded() not called when closing Editor Part (FIXED)
+ Bug 373501 - Huge Performance regression in NonInitialTypingTest#testTypeAMethod() (FIXED)
+ Bug 373529 - [Compatibility] IPageListener&apos;s pageOpened(*) is sent out too early (FIXED)
+ Bug 373885 - [jar exporter] save runnable jar export as ant script with relative name saves in eclipse program folder not workspace (FIXED)
+ Bug 374553 - Part ToolBar is not re-rendered correctly when a ToolItem is hidden (FIXED)
+ Bug 374605 - Unreasonable warning for enum-based switch statements (FIXED)
+ Bug 375576 - Controls are not fully realized when partOpened(...) , partBroughtToTop(...) and partActivated(...) are called (FIXED)
+ Bug 376269 - Editor Mapping Lookup in EditorRegistry assume file extension is not empty (FIXED)
+ Bug 376381 - [preferences][clean up] ICleanUpConfigurationUI parent Composite has excess margins (FIXED)
+ Bug 376688 - [type wizards] New class wizard doesn&apos;t remember setting for which method stubs to create (FIXED)
+ Bug 376779 - [navigation] JavaElementHyperlinkDetector#detectHyperlinks(..) is wasting time (FIXED)
+ Bug 377031 - Empty/broken workbench.xmi file makes workspace unusable (FIXED)
+ Bug 377184 - [content assist] Override method proposal does not format in subclass with type parameter (FIXED)
+ Bug 377328 - Ctrl+Y does not work for Paste in Emacs scheme (FIXED)
+ Bug 377393 - [content assist] Remove final modifier from several JavaCompletionProposal subclasses (FIXED)
+ Bug 377445 - [content assist] Code complete is vague when multiple Options exist for Enum member type on annotation (FIXED)
+ Bug 377559 - menuContribution not working in Juno, but works fine in Eclipse 3.7 (FIXED)
Comment 7 David Williams CLA 2012-04-24 18:33:20 EDT
I think the new report look great and are clearer. 

Unless you say otherwise, I will add, on my end, in masterBuild.sh, the "comparison point". I think that makes since, since sometimes you'll be passed empty string or "NONE" or otherwise invalid "starting point" (e.g. the first time a build runs") so the calling program is the one best to know that to do in those cases .. though, for now, our builder will just say "NONE" :) 

    echo "$eclipseStream Build: $buildId started"
    echo " "
    echo "   Report of changes based on comparaison to" 
    echo "   previous build or tag of $oldBbuildId"
    echo " "
    echo "$reporttext" 

Let me know if any one prefers otherwise.
Comment 8 David Williams CLA 2012-04-24 19:20:03 EDT
Oh, and one of the complaints on mailing list had to do to with "not escaping special characters" such as 

field&apos; must
not make field final&apos;

or 

dialog should have &quot;Show In &gt;

I think the note used to send with text/html mime type but I changed to text/plain, thinking that HTML would take more more work, might "run the text together" without some sort of <pre></pre> tag? But ... not sure ... was the intent to "display" this with HTML? (And, if we do that, ... you know someone is going to want real links in the bugzillas :)
Comment 9 David Williams CLA 2012-04-25 17:09:16 EDT
At first I thought today's 3.8 I build must have had a "bad report", but looking at it closely, I bet its working "as designed" and that one bug just touched a lot of executables? 

If we are done here, I suggest we close it as "fixed" and deal with other improvements as enhancement requests? 



3.8.0 Build: I20120425-1600 started
 
   Report of changes based on comparaison to
   previous build or tag of I20120425-0800
 
The following projects have changed:
bundles/org.eclipse.equinox.executable
bundles/org.eclipse.equinox.launcher.carbon.macosx
bundles/org.eclipse.equinox.launcher.cocoa.macosx
bundles/org.eclipse.equinox.launcher.cocoa.macosx.x86_64
bundles/org.eclipse.equinox.launcher.gtk.aix.ppc
bundles/org.eclipse.equinox.launcher.gtk.aix.ppc64
bundles/org.eclipse.equinox.launcher.gtk.hpux.ia64_32
bundles/org.eclipse.equinox.launcher.gtk.linux.ppc
bundles/org.eclipse.equinox.launcher.gtk.linux.ppc64
bundles/org.eclipse.equinox.launcher.gtk.linux.s390
bundles/org.eclipse.equinox.launcher.gtk.linux.s390x
bundles/org.eclipse.equinox.launcher.gtk.linux.x86
bundles/org.eclipse.equinox.launcher.gtk.linux.x86_64
bundles/org.eclipse.equinox.launcher.gtk.solaris.sparc
bundles/org.eclipse.equinox.launcher.gtk.solaris.x86
bundles/org.eclipse.equinox.launcher.motif.aix.ppc
bundles/org.eclipse.equinox.launcher.motif.hpux.ia64_32
bundles/org.eclipse.equinox.launcher.motif.linux.x86
bundles/org.eclipse.equinox.launcher.motif.solaris.sparc
bundles/org.eclipse.equinox.launcher.win32.win32.ia64
bundles/org.eclipse.equinox.launcher.win32.win32.x86
bundles/org.eclipse.equinox.launcher.win32.win32.x86_64
bundles/org.eclipse.equinox.launcher.wpf.win32.x86
bundles/org.eclipse.osgi
bundles/org.eclipse.osgi.tests

The build contains commits that mentioned these bugs:
+ Bug 377510 - Resolver gives false positives on conflicting constraints from a fragment (FIXED)