[tlbuild] patches for TeX Live

[Peter Breitenlohner]

On Thu, 25 Aug 2011, Mojca Miklavec wrote:

> (redirecting from tex-live mailing list from "synctex patches in TeX Live")
>
> On Thu, Aug 25, 2011 at 00:14, Karl Berry wrote:
>>    My question is: assuming that a linux distribution is releasing TeX
>>    Live 2011 now - does it make sense to include the following patch?
>>    http://tug.org/svn/texlive?view=revision&revision=23644
>>
>> As far as I know, yes.  Perhaps Peter will read this in due course and
>> confirm.
>>
>> I haven't tried to build the sources with just that
>> patch and only that patch applied, but looking at it, I find it hard to
>> imagine that it interacts with anything else.
>>
>> (Of course, it would make sense for any kind of distribution, of course,
>> not just [GNU/]Linux.  The patch is not platform-specific.  For the
>> record ...)
>
> In the meantime we tested. Compilation runs fine and the resulting PDF
> works much better.
>
> ........

Hi Mojca,

first of all thanks for your comments and suggestions, although many of them
may be hard to implement mainly due to lack of resources and/or manpower.

The patch you mention (r23644) is certainly a bug fix and should be included
by anyone building for a distro.  Unfortunately classifying all patches into
such categories may be difficult and applying just the bug fixes may not
work because they might not apply cleanly without the other ones (e.g.,
improvements or the recent introduction of (e-)upTeX & Co).

================

I am working exclusively with the trunk and porting just the bug fixes
(critical or not) to a different branch would require first such a branch to
exist and then extra work.

During the freeze for TL2011 a lot of patches had accumulated in my working
copy and I wanted to apply them in small pieces according to subject.  By a
mistake they were then all commited as one large patch set; I mention this just
to demonstrate the kind of problems one might encounter.

================

Before committing patches I always (try to) make sure that a build on y
system (native for i686-linux and cross linux => i386-pc-mingw32) succeeds. 
Of course, this does not mean that everything builds for all different
architectures, nor that everything would be useful for distros, but at least
everything might be a candidate.

But I certainly egree with you, that taking a random SVN snapshot for a
distro build might not be the best possible choice.

================

BTW: Feedback from more frequent attempts to build for all kinds of systems
could spare us the pain of last minute fixes for architecture specific build
problems and would even make it feasible to forward such problems to the
upstream maintainers of various packages and have them fixed there.

Regards
Peter Breitenlohner <peb at mppmu.mpg.de>