Re: [Quilt-dev] Tru64 and OSX patches

[Andreas Gruenbacher]

On Monday 30 January 2006 13:46, John Vandenberg wrote:
> On 30/01/06, Andreas Gruenbacher <address@hidden> wrote:
> > On Monday 30 January 2006 09:33, John Vandenberg wrote:
> > > Hi everyone,
> > >
> > > Here are a few more patches that I needed to build and test quilt on
> > > Tru64, OS X, and of lesser importance, MinGW.
> > >
> > > http://zeroj.hda0.net/quilt-20060130/series.php
> >
> > http://zeroj.hda0.net/quilt-20060130/tru64-backup-files.diff
> >   Fixed slightly differently. (Btw, more recent Try64 isn't as crappy
> >   anymore according to Google.)
>
> Was there a problem with mine?

It was a little complicated at a place where it's not important to be 
blazingly fast, and it kept the buffer size at PATH_MAX + 1, which was only 
needed to catch snprintf overflows, that's all.

> > http://zeroj.hda0.net/quilt-20060130/getopt-spaces.diff
> >   Yuck! See below.
>
> It's perl.

Not an excuse ;)

> > http://zeroj.hda0.net/quilt-20060130/test-mail-stdin.diff
> >   Added.
> >
> > http://zeroj.hda0.net/quilt-20060130/hostname_f.diff
> >   This was meant to be worked around by using the --sender option. I'm
> >   updating the test case instead.
>
> $(hostname) does the right thing on a lot of other boxes; I would
> prefer to not need to provide --sender.

Okay, let's try both then. Will the following work for you?

-       opt_sender="${LOGNAME:-$(whoami)address@hidden(hostname -f 2>/dev/null)"
+       hostname=$(hostname -f 2>/dev/null)
+       if [ "$hostname" = "${hostname/.}" ]
+       then
+               hostname=$(hostname)
+       fi
+       opt_sender="${LOGNAME:-$(whoami)address@hidden"

> Is there a .quiltrc setting for this ?

QUILT_MAIL_ARGS="--sender=...", as with any other command

> > > The getopt patch is the most important of these, as it affects all
> > > platforms that don't have a native getopt.  I haven't been able to
> > > complete the new mail test case on either platform due to uniq -D,
> > > date --rfc-822, and other miscellaneous issues.
> >
> > Something is still seriously wrong in compat/getopt.in:
>
> yea; to begin with its written in perl.  It's only intended as a stop
> gap measure, so it catered for the common cases in order that quilt is
> usable.  iirc util-linux is being updated to be portable, so it may
> soon be reasonable to expect that the linux additions are available on
> other platforms.
>
> Another alternative is to change over to using the bash builtin
> getopts.  Is that feasible; after this release?

getopts doesn't support long options.

Andreas