# [FFmpeg-devel] some renames in the regression testsuite

Reimar Döffinger Reimar.Doeffinger
Mon Mar 30 19:30:32 CEST 2009

On Mon, Mar 30, 2009 at 06:59:27PM +0200, Diego Biurrun wrote:
> On Mon, Mar 30, 2009 at 06:35:00PM +0200, Reimar D?ffinger wrote:
> > On Mon, Mar 30, 2009 at 05:32:18PM +0200, Diego Biurrun wrote:
> > > On Mon, Mar 30, 2009 at 03:54:32PM +0200, Reimar D?ffinger wrote:
> > > > On Mon, Mar 30, 2009 at 02:22:46PM +0200, Diego Biurrun wrote:
> > > > > Those changes are done.  Next I wanted to change the filenames of the
> > > > > generated regression files to use lavf instead of libav, but when I
> > > > > apply the attached patch, the lavftest target fails with a difference
> > > > > during the gxf test.  I'm completely puzzled, all this does is rename
> > > > > a few files, how can this affect the result?
> > > >
> > > > Upload the two gxf files (assuming they differ).
> > >
> > > I have uploaded a dir named 'regtest_fail_file_rename' with 'before' and
> > > 'after' subdirectories to incoming.
> >
> > gxf_write_material_data_section writes the file name into the output.
>
> Then it's OK to just update the regression test checksum?

I'd prefer to have it fixed first, to be sure there is not more that
changes.

> > As you can probably guess I don't like that kind of stuff anyway, but
> > there are at least some bugs:
> > 1) it assumes that / is the directory separator - while e.g. Windows
> > allows that \ can be used, too.
> > 2) it probably should not be done with bitexact enabled
>
> Yes, it does not sound like a good idea at all.  Wouldn't the filename
> without the full path suffice?

Huh? It does not write the full path, only the last name - as long as
you are so considerate to use UNIX-style paths with '/'.
So writing to "/tmp/file.gxf" makes it write EXT:/PDR/default/file.gxf
in the file.
If you used something like "C:\tmp\file.gxf" on Windows it would write
EXT:/PDR/default/C:\tmp\file.gxf into the file though...