This is the mail archive of the
mailing list for the Cygwin project.
Re: [64bit] Biber packaging questions
- From: "Yaakov (Cygwin/X)" <yselkowitz at users dot sourceforge dot net>
- To: cygwin-apps at cygwin dot com
- Date: Thu, 13 Jun 2013 13:57:45 -0500
- Subject: Re: [64bit] Biber packaging questions
- References: <51B8813F dot 6060207 at cornell dot edu> <51B99A6C dot 10608 at users dot sourceforge dot net> <51BA1164 dot 4070909 at cornell dot edu>
On 2013-06-13 13:37, Ken Brown wrote:
On 6/13/2013 6:09 AM, Yaakov (Cygwin/X) wrote:
On 2013-06-12 09:10, Ken Brown wrote:
1. Should these build prerequisites be added to the 64bit distro?
Otherwise it will be difficult for others to rebuild biber from source.
These should be added to both, although I suspect many are noarch, so
you should only need to build some of those once.
I'll go ahead with this for the 64bit distro now, and handle the 32bit
distro later. The differences in the way Perl modules are handled makes
it difficult to treat the two distros in the same way. I hope they will
eventually be in sync.
If you're talking about perl_vendor, I think that's going to go away
By the way, I'm also going to have to add perl-Unicode-Collate, which
will become obsolete once Perl is updated, because the version of
Unicode::Collate included in perl-5.14 is too old.
That's fine; we'll just need to remember that whenever we move to Perl 5.16.
3. There is a completely different approach I could take. Namely, I
could simply package Biber as a perl module and forget about packing it
into a Perl Archive. If I do this, then users will need perl 5.16 or
later, as well as most or all of the perl modules listed above, so the
RFU will have to wait for a perl update; but that's probably not
serious. Would this be preferable? I'm not aware of any Linux distros
that do this, though someone did do it unofficially for Fedora:
I strongly recommend this route. For one, it is probably faster (not
having to decompress so much on the fly), but more importantly, it does
not involve bundling code (which is to be avoided for the same reasons
as static library linkage).
That's my preference too. In fact, I had already built and tested this
using essentially the same biber.cygport as the one you suggested.
The bottom line is that using PAR makes sense only for the standalone
TeX Live, where they want to avoid system dependencies as much as
possible. For a distribution managed TeX, however, there is really no
need for it.
Let me know if you need any help packaging the Perl modules.