[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [PATCH] Fix for blame -g incorrect revisions

From: Alan Wood <Alan.Wood_at_clear.net.nz>
Date: Tue, 05 May 2009 08:18:45 +1200

Hi Lieven,
 I would like to write a test for this. Only a few things to overcome.

 1) I am not really a python writer, but willing to learn a bit more of this.

 2) I have never managed to the get win-tests.py to run cleanly. About 30% failures
last time if I recall. My attempts to run single tests were not successful either. I wish I
could see how to debug these tests.

 3) I don't have a lot of time. But if that's the fastest way to get all this done then I'll
start doing it.

 I did have a look at the tests and they don't seem to go near this area. They seem
content to only use blame single line files which does not show this bug at all.

 I will try and get the tests working here tonight and see how I get on

Alan Wood

On 4 May 2009 at 12:22, Lieven Govaerts wrote:

> Hi Alan,
>
> are you willing to write a regression test for this issue? There's
> probably a test in blame_tests.py which you can reuse or extend.
> I'm willing to review and commit your patch, but we need the reg. test
> anyhow and I'm short on time to write one myself.
>
> Lieven
>
> On Thu, Apr 23, 2009 at 11:11 AM, Alan Wood <Alan.Wood_at_clear.net.nz> wrote:
> > Hi,
> >
> >  Here is an updated version of the patch in which I have reverted two lines that I
> > changed as I have been unable to convince myself that the change was necessary.
> >
> >  Can't actually come up with a test case that uses the lines so I will leave them alone.
> >
> >  Just trying to understand test system well enough to be able to add tests for this type
> > of issue.
> >
> > Alan
> >
> > On 20 Apr 2009 at 21:33, Alan Wood wrote:
> >
> >> Hi All,
> >>  Here is a patch that hopefully corrects the output from blame -g in
> >> the case where the merged data has an inserted line as the last line
> >> in a chunk and some other cases I saw on the way. I have also
> >> attached a bat file as a simpler example than the original post (
> >> sorry it's windows ).
> >>
> >> I have tried this with 1.5.4 and 1.6.0
> >> Also trunk with with existing wc format.
> >>
> >> Output from script
> >> --------------------
> >> C:\alan\subversion\test>svn --version
> >> svn, version 1.6.0 (Release Candidate 3)
> >>    compiled Mar 15 2009, 12:54:57
> >>
> >> C:\alan\subversion\test>svn blame wc/trunk/a.txt
> >>      1       Alan "line 1"
> >>      1       Alan "line 2"
> >>      6       Alan "line 2.1 branch rev 4"
> >>      6       Alan "line 2.2 branch rev 5"
> >>      1       Alan "line 3"
> >>      3       Alan "line 3.1 trunk rev 3"
> >>      3       Alan "line 3.2 trunk rev 3"
> >>      1       Alan "line 4"
> >>      3       Alan "line 4.1 trunk rev 3"
> >>
> >> C:\alan\subversion\test>svn blame -g wc/trunk/a.txt
> >>        1       Alan "line 1"
> >>        1       Alan "line 2"
> >> G      4       Alan "line 2.1 branch rev 4"
> >>        1       Alan "line 2.2 branch rev 5"
> >> G      1       Alan "line 3"
> >>        1       Alan "line 3.1 trunk rev 3"
> >>        1       Alan "line 3.2 trunk rev 3"
> >> G      1       Alan "line 4"
> >>        1       Alan "line 4.1 trunk rev 3"
> >>
> >> -----------
> >> after patch applied to trunk and running with the same working copy
> >>
> >> C:\alan\subversion\test>svn.exe --version
> >> svn, version 1.7.0 (dev build)
> >>    compiled Apr 20 2009, 17:20:24
> >>
> >> C:\alan\subversion\test>svn.exe blame -g wc/trunk/a.txt
> >>        1       Alan "line 1"
> >>        1       Alan "line 2"
> >> G      4       Alan "line 2.1 branch rev 4"
> >> G      5       Alan "line 2.2 branch rev 5"
> >>        1       Alan "line 3"
> >>        3       Alan "line 3.1 trunk rev 3"
> >>        3       Alan "line 3.2 trunk rev 3"
> >>        1       Alan "line 4"
> >>        3       Alan "line 4.1 trunk rev 3"
> >>
> >> refer to:
> >> http://subversion.tigris.org/ds/viewMessage.do?dsMessageId=1179481&d
> >> sForumId=462
> >>
> >> Possible log message:
> >>
> >> [[[
> >>  Fix issue with blame -g showing incorrect revisions when lines
> >> added from branch
> >>  * subversion/svn/blame-cmd.c
> >>    (blame_receiver): Handle revision comparison when revision or
> >> merged revision is not valid (xml output was already ok)
> >>
> >>  * subversion/libsvn_client/blame.c
> >>  (normalize_blames): Split chunks correctly don't just extend the
> >> later chunk back ( comments were correct but not code )
> >>  (svn_client_blame5): Allow information from merged line to be sent
> >> to callback even when no valid revision pre-merge.
> >> ]]]
> >>
> >>
> >>
> >> Alan Wood
> >>
> >> ------------------------------------------------------
> >> http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1821874
> >
> > ------------------------------------------------------
> > http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1875446
>
> ------------------------------------------------------
> http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2054911

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2059756
Received on 2009-05-04 22:19:09 CEST

This is an archived mail posted to the Subversion Dev mailing list.