On Tue, May 1, 2012 at 12:39 PM, Paulo César Pereira de Andrade
<paulo.cesar.pereira.de.andrade(a)gmail.com> wrote:
> Ok. Worth mentioning is that in Mandriva I did also package several
> optional "databases", see:
>
>
http://svn.mandriva.com/cgi-bin/viewvc.cgi/packages/cooker/pari/current/S...
>
> and it was already packaging gp2c that comes from an extra upstream
> tarball.
Okay, once we get the new pari in the distribution, it would be good
to look at these databases and see what we should have in Fedora.
> Also, I did build pari with --disable-tls, as I recall having issues with
> it when building sagemath. Not sure if sage-on-gentoo also had issues
> with it
https://github.com/cschwan/sage-on-gentoo but I recall some
> problems they had that looked familiar to me, and I commented in
> #sagemath about it some months ago.
Hmmm, interesting information. I'll CC the pari maintainer on this
email to make him aware of a potential issue there. Do you have any
idea if the problem was a bug in pari itself, or in Sage's use of
pari?
I think I did it "defensively" when debugging problems in the
sagemath package, as newer pari uses a lot of tls variables, e.g.
$ grep -r __thread .
./src/headers/parisys.h:# define THREAD __thread
$ grep -r THREAD . | wc -l
84
and it has things like __thread static variables in functions, several
global variables, etc.
But at least in pari 2.5.1 it is disabled by default.