Bug #30552 for Math-Pari: Cannot be built in a way suitable for including in binary distributions

Wed Nov 07 12:21:05 2007 GWOLF [...] cpan.org - Ticket created

Subject:

Cannot be built in a way suitable for including in binary distributions

Hi, I'm writing this as one of the members of the Debian pkg-perl group. There is interest in having this module packaged for Debian GNU/Linux - But as the build process is structured right now, it is impossible to do so. It boils down that, as our software is auto-built and must be completely buildable from our archive's sources, we cannot download GP/PARI's sources at build time. I tried specifying the directory where our libpari-dev include files are (/usr/include/pari), but your build script insists on having PARI's test suite - Please refer to our bug report (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=440527) for fuller details. Our request is simple: Would it be possible to build this module with only the include files, specifying it _not_ to require the test suite, having only the includes? Thank you very much!

Wed Nov 07 20:30:30 2007 nospam-abuse [...] ilyaz.org - Correspondence added

CC:	undisclosed-recipients: ;
Subject:	Re: [rt.cpan.org #30552] Cannot be built in a way suitable for including in binary distributions
Date:	Wed, 7 Nov 2007 17:30:11 -0800
To:	Gunnar Wolf via RT <bug-Math-Pari [...] rt.cpan.org>
From:	Ilya Zakharevich <nospam-abuse [...] ilyaz.org>

On Wed, Nov 07, 2007 at 12:21:07PM -0500, Gunnar Wolf via RT wrote: Show quoted text

> I'm writing this as one of the members of the Debian pkg-perl group. > There is interest in having this module packaged for Debian GNU/Linux - > But as the build process is structured right now, it is impossible to do > so. It boils down that, as our software is auto-built and must be > completely buildable from our archive's sources, we cannot download > GP/PARI's sources at build time. I tried specifying the directory where > our libpari-dev include files are (/usr/include/pari), but your build > script insists on having PARI's test suite - Please refer to our bug > report (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=440527) for > fuller details. > Our request is simple: Would it be possible to build this module with > only the include files, specifying it _not_ to require the test suite, > having only the includes? > Thank you very much!

If you feel fine shipping a broken module, it makes sense to disable the test suite. Just do not run 'make test'. Does this answer your question? Yours, Ilya

Wed Nov 07 20:30:33 2007 The RT System itself - Status changed from 'new' to 'open'

Thu Nov 08 10:38:44 2007 GWOLF [...] cpan.org - Correspondence added

On Wed Nov 07 20:30:30 2007, nospam-abuse@ilyaz.org wrote: Show quoted text

> If you feel fine shipping a broken module, it makes sense to disable > the test suite. Just do not run 'make test'. > > Does this answer your question?

No, sorry for not being explicit enough - I'm copying from my Debian report - libpari-dev is the package which contains the PARI development headers (i.e. /usr/include/pari): Curiously, the package does not build even with libpari-dev installed - It _does_ require the full sources (i.e. the test suite, at least): ------------------------------------------------------------ $ perl Makefile.PL paridir=/usr/include/pari/ Setting up Math::Pari with Perl 5.008008 on linux version 2.6.17-2-vserver-amd64; cc=cc, gccversion=4.2.3 20071014 (prerelease) (Debian 4.2.2-3), ccflags='-D_REENTRANT -D_GNU_SOURCE -DTHREADS_HAVE_PIDS -DDEBIAN -fno-strict-aliasing -pipe -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64', ld=cc, ldflags=' -L/usr/local/lib', optimize='-O2'. Could not extract version from '/usr/include/pari/config/version'; Do you remember that paridir should be location of PARI/GP source tree, not the location of compiled+installed files? Trying extract from the directory name... Directory `/usr/include/pari' has unknown syntax... Found GP/PARI build directory in /usr/include/pari (integer-formatted version /usr/include/pari). Cannot find tests in /usr/include/pari/src/test/in: No such file or directory at utils/Math/PariBuild.pm line 513. ------------------------------------------------------------

Thu Nov 08 16:49:45 2007 nospam-abuse [...] ilyaz.org - Correspondence added

Subject:	Re: [rt.cpan.org #30552] Cannot be built in a way suitable for including in binary distributions
Date:	Thu, 8 Nov 2007 13:49:06 -0800
To:	Gunnar Wolf via RT <bug-Math-Pari [...] rt.cpan.org>
From:	Ilya Zakharevich <nospam-abuse [...] ilyaz.org>

On Thu, Nov 08, 2007 at 10:38:51AM -0500, Gunnar Wolf via RT wrote: Show quoted text

> > Queue: Math-Pari > Ticket <URL: http://rt.cpan.org/Ticket/Display.html?id=30552 > > > On Wed Nov 07 20:30:30 2007, nospam-abuse@ilyaz.org wrote:

> > If you feel fine shipping a broken module, it makes sense to disable > > the test suite. Just do not run 'make test'. > > > > Does this answer your question?

> > No, sorry for not being explicit enough - I'm copying from my Debian > report - libpari-dev is the package which contains the PARI development > headers (i.e. /usr/include/pari): > > Curiously, the package does not build even with libpari-dev installed

I do not see why this is "curious". Show quoted text

> - It _does_ require the full sources (i.e. the test suite, at least):

No wonder... Show quoted text

> ------------------------------------------------------------ > $ perl Makefile.PL paridir=/usr/include/pari/ > Setting up Math::Pari with Perl 5.008008 on linux version > 2.6.17-2-vserver-amd64; > cc=cc, gccversion=4.2.3 20071014 (prerelease) (Debian 4.2.2-3), > ccflags='-D_REENTRANT -D_GNU_SOURCE -DTHREADS_HAVE_PIDS -DDEBIAN > -fno-strict-aliasing -pipe -I/usr/local/include -D_LARGEFILE_SOURCE > -D_FILE_OFFSET_BITS=64', > ld=cc, ldflags=' -L/usr/local/lib', optimize='-O2'. > Could not extract version from '/usr/include/pari/config/version'; > > Do you remember that paridir should be location of PARI/GP source tree, > not the location of compiled+installed files?

Does not this answer your questions? Show quoted text

> Trying extract from the directory name... > Directory `/usr/include/pari' has unknown syntax... > Found GP/PARI build directory in /usr/include/pari > (integer-formatted version /usr/include/pari). > Cannot find tests in /usr/include/pari/src/test/in: No such file or > directory at utils/Math/PariBuild.pm line 513. > ------------------------------------------------------------

This is just the tip of the iceberg... Sorry to sound pedantic, but I still have no idea what it is you wanted to communicate about. [There is a (quite small) chance that Math::Pari build can be performed when only libpari-dev is available. Do you want to check for this chance, or what? In which way I may be useful to you?] Thanks, Ilya

Thu Nov 08 16:57:38 2007 nospam-abuse [...] ilyaz.org - Correspondence added

Subject:	Re: [rt.cpan.org #30552] Cannot be built in a way suitable for including in binary distributions
Date:	Thu, 8 Nov 2007 13:57:15 -0800
To:	Ilya Zakharevich via RT <bug-Math-Pari [...] rt.cpan.org>
From:	Ilya Zakharevich <nospam-abuse [...] ilyaz.org>

On Thu, Nov 08, 2007 at 04:49:47PM -0500, Ilya Zakharevich via RT wrote: Show quoted text

> Sorry to sound pedantic, but I still have no idea what it is you > wanted to communicate about. [There is a (quite small) chance that > Math::Pari build can be performed when only libpari-dev is available.

Oups, I see that I was not clear enough. Let us rewrite it as There is a (quite small) chance that Math::Pari build could be performed (given appropriate modifications of Math::PariBuild module) when only libpari-dev is available. Yours, Ilya

Wed Dec 28 10:46:05 2016 AKHUETTEL [...] cpan.org - Correspondence added

On Thu Nov 08 16:57:38 2007, nospam-abuse@ilyaz.org wrote: Show quoted text

> > Oups, I see that I was not clear enough. Let us rewrite it as > > There is a (quite small) chance that Math::Pari build could be > performed (given appropriate modifications of Math::PariBuild > module) when only libpari-dev is available. >

Well, the internal bundling against libpari goes against Gentoo QA guidelines as well. Please consider supporting the build against an installed copy (with dynamic linking). See e.g. for a years-old blog post, https://blog.flameeyes.eu/2009/01/bundling-libraries-for-despair-and-insecurity/

Sat Jan 14 16:44:15 2017 nospam-abuse [...] ilyaz.org - Correspondence added

Subject:	Re: [rt.cpan.org #30552] Cannot be built in a way suitable for including in binary distributions
Date:	Sat, 14 Jan 2017 13:43:37 -0800
To:	Andreas K. Hüttel via RT <bug-Math-Pari [...] rt.cpan.org>
From:	Ilya Zakharevich <nospam-abuse [...] ilyaz.org>

On Wed, Dec 28, 2016 at 10:46:05AM -0500, Andreas K. Hüttel via RT wrote: Show quoted text

> Queue: Math-Pari > Ticket <URL: https://rt.cpan.org/Ticket/Display.html?id=30552 > > > On Thu Nov 08 16:57:38 2007, nospam-abuse@ilyaz.org wrote:

> > > > Oups, I see that I was not clear enough. Let us rewrite it as > > > > There is a (quite small) chance that Math::Pari build could be > > performed (given appropriate modifications of Math::PariBuild > > module) when only libpari-dev is available. > >

> > Well, the internal bundling against libpari goes against Gentoo QA guidelines as well. Please consider supporting the build against an installed copy (with dynamic linking).

This is very close to being out of question. First, one needs to be able to build with a supported distribution IN ANY WAY — which is with very high priority on my TODO list — but it is there for almost a decade! Show quoted text

> See e.g. for a years-old blog post, > https://blog.flameeyes.eu/2009/01/bundling-libraries-for-despair-and-insecurity/

This is extremely long and unclear; practically unreadable. My conjecture is that it boils down to “Fix all the problems you see upstream, so there is a hope of resolving all the newly discovered problems in a timely manner”. Is there anything else there? Yours, Ilya P.S. In addition to supporting a contemporary version, I would need to link agains a (non-supplied) private header file. I presume that supplying such a header file (or a replacement?) with Math::PARI would not contradict the distribution guidelines?