Bug#788883: perl: drop (most of) debian/prefix_changes.diff

[Dominic Hargreaves]

On Tue, Jun 16, 2015 at 12:21:01AM +0300, Niko Tyni wrote:
> I think it's high time we dropped the EU:MM changes that still allow
>  make install PREFIX=something
> to work with Makefile.PL based packages. Upstream has recommended DESTDIR
> instead since 2002, see #568748 and #545904.
> 
> There's been a somewhat heuristic lintian check for five years visible at
>  https://lintian.debian.org/tags/debian-rules-makemaker-prefix-is-deprecated.html
> and it's currently down to 55 packages. The logic in the check is basically
>   tag if ($uses_makefile_pl && /install.*PREFIX/ && !/DESTDIR/)

Wow, I was completely unaware of this :)
 
> It looks like debian/prefix_changes.diff contains some unrelated
> changes, at least 
> 
> -       \$(NOECHO) \$(POD2MAN) --section=$section --perm_rw=\$(PERM_RW)
> +       \$(NOECHO) \$(POD2MAN) --section=\$(MAN${section}EXT) --perm_rw=\$(PERM_RW)
> 
> that should be broken out in any case.
> 
> I expect affected packages will start to FTBFS, as seen in #545904,
> so a full check would be "just" a simple matter of building the archive
> with and without the patch, comparing the results and filing bugs :)
> This could just as well be tested with the 5.20 packages in sid.
> 
> Alternatively, if a full check seems overkill, we could just check the 55
> packages noticed by lintian and let the rest be fixed in their own time
> when they start to fail.

Well, I should be covering all of the potentially affected packages
with my Depends: perl rebuild test which I'm running at the moment[1]
with perl 5.22. So I'd say add the patch to the next upload of perl
5.22 to experimental and then we can get the test coverage we need
the next time I do a full rebuild as part of the perl 5.22 QA work.

Cheers,
Dominic.

[1] 22 hours runtime, up to 'duck' - but then there's no huge hurry :)