Ken Moffat wrote:
On Tue, Dec 09, 2014 at 10:00:15AM -0600, Bruce Dubbs wrote:
Ken Moffat wrote:
On Mon, Dec 08, 2014 at 10:11:52PM -0600, Bruce Dubbs wrote:
Ken Moffat wrote:
On Fri, Dec 05, 2014 at 08:54:04PM +0000, Ken Moffat wrote:
This is just a heads up that there is something rotten in the state
of texlive when built from source, at least on i686.
[...]
2. My test script for xindy now fails, because the index (which is
the whole point of xindy) is not created. That was how things used
to be when my clisp build was inadequate.
And since then, it has turned out to be broken on i686 in LFS-7.6
(and until last Friday, nobody had noticed).
I've got a 686 with gcc-4.7.0. Is there something explicit you would like
me to test?
[...]
OK, I'll hold off.
-- Bruce
For the record, I have now managed to compile gcc-4.8.2 on recent
svn, using instructions modelled on gcc pass 2 from LFS-7.5 (i.e.
use the matching versions of gmp, mpfr, mpc in the build), and put it
in /opt/gcc48.
For gcc, which is all that is needed for clisp and (newer, separate)
xindy-2.5.1 that is good enough. For g++, the conventions have
changed, trying to test-compile a g++ package failed because a
library from poppler (compiled with gcc-4.9.2) did not link.
But it did not help, my xindy test still fails to create the index
on i686.
Along the way, I have added 'ulimit -s unlimited' to clisp on i686
(tests still failed somewhere, but apparently not with the segfault
I had seen on x86_64), and then added ---with-debug' to the i686
configure. On gcc-4.9.2 (and 4.8.2) 'make check' now passes. I
have NOT got around to testing that on x86_64, but I think both
changes might be heading towards the book.
For xindy, I am totally out of ideas about why it does not work on
i?86. I'm coming towards the view that we need a note "This does not
work on i?86". (Normally, I would say i686 because I doubt anyone still
has a usable i{4,5}86 and ISTR glibc no longer supports i386 itself,
but the binary texlive insists it is for i386. Maybe "32-bit x86" ?
I'll now see if I can find upstream's latest version of asymptote,
and see if I can confirm that the failure while installing the docs
is caused by gcc-4.9.2. Even if that does indicate what to blame,
it does not provide a solution (other than to indicate that on both
architectures, gcc-4.9.1 worked fine), but perhaps it will actually
point out that the problem is somewhere else :-(
Although I started the "build all texlive from source" process by
assuming that it woud be good enough to just use the versions in the
texlive source, and upgrade once a year, I've had to upgrade
biber/biblatex to get that to work with current perl, and I'm now
thinking it _might_ be better to use separate xindy and asymptote.
For the moment, that will not solve any problems, but it's something
I'd like to keep in mind if the two differnt issues (xindy - busted
on i686, asymptote - docs now fail to install) ever get resolved.
BTW - did I ever mention how much I hate this stuff ?
The only thing I can say is that we appreciate it.
BTW, did you ever install a commercial distro, say debian or fedora and
test their versions of xindy and asymptote.
-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
