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

Re: 1.7.15 up for testing/signing

From: Ben Reser <ben_at_reser.org>
Date: Fri, 14 Feb 2014 11:30:13 -0800

On 2/14/14, 7:44 AM, Ivan Zhakov wrote:
> On 14 February 2014 17:51, Philip Martin <philip.martin_at_wandisco.com> wrote:
>> Philip Martin <philip.martin_at_wandisco.com> writes:
>>> Ivan Zhakov <ivan_at_visualsvn.com> writes:
>>>> I am getting diff_tests#10 failure over all protocols on Windows. All
>>>> other tests pass.
>>>> [[[
>>>> FAIL: diff_tests.py 10: diff when property was changed but text was not
>>>> ]]]
>>>> The failure is reproducible. Is it known issue? What I should to check?
>>> The test creates r2 for the file iota that adds svn:eol-style=native.
>>> Then it runs various diffs: -r1:2, -r2:1, etc. The one that fails is
>>> -rPREV which compares r1 in the repository to r2 in the working copy.
>>> The test expects the diff to show a property change only but a text
>>> change is seen.
>>> Perhaps the code that converts the working file to repository form is
>>> failing? The code in question is
>>> libsvn_client/diff.c:diff_repos_wc_file_target:2320
>>> and that code certainly is invoked on my Linux machine.
>> So it looks as if this code is converting the wrong way. It's supposed
>> to convert from working copy form to repository form but is doing the
>> reverse conversion. On Linux the conversion is a noop so it works. At
>> some point we need
>> if (eol_style == svn_subst_eol_style_native)
>> but I don't know if this should go directly in client/diff.c or whether
>> some other translation function should be used.
> I confirm that all tests pass with proposed patch on Windows over ra_local.

The sad thing here is that the buildbot caught this issue.

Here's the build before the broken change was merged:

Here's the one after:

The logs are gone so I can't say for sure what the failure is but I'm fairly
comfortable saying the build bot caught this issue.

I just made a test build against the branch I created to fix the problem and it
passes (ignoring the JavaHL issue):

We have these build bots but they're next to useless if they have persistent
test failures. :(

I should have noticed but I've gotten tired of reviewing the buildbot failures
on branches since it's always been spurious.
Received on 2014-02-14 20:30:44 CET

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

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.