Re: svn trunk r17230: FAIL (win32 ra_local fsfs)

On Mon, 7 Nov 2005, Erik Huelsmann wrote:

> On 11/7/05, svn-builder@shadyvale.net <svn-builder@shadyvale.net> wrote:
> Last 100 lines of the build log (full log attached):
>
> Running all tests in commit_tests.py...FAILURE
> Running all tests in update_tests.py...success
> Running all tests in switch_tests.py...success
> Running all tests in prop_tests.py...success
> Running all tests in schedule_tests.py...success
> Running all tests in log_tests.py...success
> Running all tests in copy_tests.py...success
> Running all tests in diff_tests.py...FAILURE

>svn: Can't get attribute information from file 'nul': Incorrect function.
>FAIL: diff_tests.py 2: add a file

>Grr! Seems I'll need to have another look at what's currently done
>with the NUL file. Anybody who has an idea why svn would need
>attributes from an empty file to diff against?

I don't know why *we* need that, but this sends a pretty strong signal to
me. Julian seems to have a good point here. NUL on Windows doesn't seem
that much like a file as it is on Unix. An external diff program could
well want file attributes on its input file.

It seems like we should avoids this smallish optimization and just create
an empty temporary file per diff/merge invocation instead. That's not a
big
deal, and we avoid breaking peoples use of external differs.

Erika, don't forget to leave a note in the code about why we don't use the
null device for this purpose.

Thanks,
//Peter

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Nov 7 13:28:02 2005