Re: [Quilt-dev] CVS test suite fixes

[John Vandenberg]

On 1/11/06, Jean Delvare <address@hidden> wrote:
> Hi Markus,
>
> > Some small patches to make the test suite work again.
>
> Good catch, I've committed the fixes related to the changes to "quilt
> diff". This, together with another minor fix from me, makes the test
> suite work again on my systems.
>
> > I also had to add "--sort" to "diff --snapshot" in snapshot.test, probably
> > because of filesystem related issues.
>
> I did not apply these changes. The test suite is meant to detect the
> problems, and I guess you just hit one. The "--sort" shouldn't be
> necessary as far as I can see (but I am not familiar with quilt
> snapshot), so instead of changing the test suite, I think we should
> investigate the issue and maybe fix quilt itself if needs be.
>
> Can you describe the error you get? Then can you try to understand why
> you get it (when others, including me, don't)?
>
> > Finally it would be nice if test/{Makefile,run} could be placed under
> > Autoconf control, so that actually the newly installed version will get
> > tested and not the first quilt encountered on path.
>
> Even better would be John Vandenberg's way: test the to-be-installed
> version of quilt. There are patches floating around (search the mailing
> list archive) if you want to take your chance.

Over the last week I have updated my series of patches, incorporating
Makefile cleanup, patch argument improvements and OS X support.

http://zeroj.hda0.net/quilt-0.43-patches/

I have yet to test on Solaris, and have a few improvements yet to make
before its ready.

Also, I would appreciate some assistance with one problem I found with
the previous local-quilt.diff submitted to this list.  After
compat/sendmail is created as a symlink to /usr/sbin/sendmail, if
Makefile is modified, the symbolic link is always regenerated because
make uses the date of the linked file(/usr/sbin/sendmail) rather than
the symlink(compat/sendmail).  This can be avoided by using the
--check-symlink-times option.

I was not able to find any way that this option can be enabled from
within the distributed Makefile, so the need to use this additional
option would probably come up on the list a bit.  To avoid that, the
updated patch generates a virtual symlink as a shell script, with a
restriction that only 9 parameters will be passed through.  I think I
may be able to remove that restriction by using: set -o noglob;
/bin/whatever $@, but I have yet to test that thoroughly.

Can this be improved on?

--
John