[PATCH] test-subrepo-svn.t: partially adapt for Windows

Mads Kiilerich mads at kiilerich.com
Tue Jun 26 09:01:16 CDT 2012


On 26/06/12 13:42, Eduard-Cristian Stefan wrote:
> On 2012-06-26 13:21, Mads Kiilerich wrote:
>> Right - you had a workaround for that in your patch. I left it out 
>> because I didn't know what issue it solved and why simply ignoring 
>> the extra output was the right solution. Now we know what issue it 
>> was working around - but not why there is a need for the workaround 
>> and if there is a real problem hiding here.
>>
>> What temp directory are you using? Could it by any chance be 
>> x:\Administrator\... on a different drive? Or why could subversion 
>> possibly emit this long almost-absolute path?
>>
>> The test passes with 1.6 and 1.7 on linux, and I have seen it pass 
>> with one I-don't-know-which version on windows.
>>
>> Some progress on the other subversion tests will perhaps reveal if 
>> this is a general problem or what is going on.
>
> Just to remember the current setup:
> - hackable folder => d:\hg
> - MSYS root folder => D:/hg/hg-m/MSYS/1.0
> - /tmp folder in MSYS points to C:/Users/ADMINI~1/AppData/Local/Temp
> - MSYS env vars: TEMP=/tmp, TMP=/tmp
> - the working folder is 
> c:\Users\Administrator\AppData\Local\Temp\hgtests.kkslmj\test-subrepo-svn.t
> - svn binaries are located in d:\svn, but even if moved to c:\svn the 
> test doesn't pass

So subversion on your machine shows the absolute path with the top level 
stripped of instead of showing a simple relative path. That makes it 
look even more like a subversion bug. Can you confirm that it works for 
you with 1.7 and perhaps find a bug report or changelog of the fix?

Please also try with the temp folder set to d:/temp and c:/temp and see 
if that makes a difference.

/Mads


More information about the Mercurial-devel mailing list