February 2009 - binary-builders

Re: [Context] pdftex from ConTeXt minimals
by Mojca Miklavec 26 Feb '09

26 Feb '09

On Tue, Feb 24, 2009 at 02:29, Manuel Pégourié-Gonnard wrote: > Hi Mojca, > > I recently installed the *tex binaries from the ConTeXt minimals in > order to test some stuff with a recent luatex, and just ran into the > following problem: > > mpg@roth:~% pdftex --version > pdftex: /lib/libc.so.6: version `GLIBC_2.8' not found (required by pdftex > > I have no problem with the luatex executable from the minimals, nor with > the pdftex executable from TL'08. > > I'd be happy to provide any information you need about the problem if > you tell me what you want to know (no idea about how to investigate this > myself). Hello Manuel, thanks a lot for your feedback. Apparently I have given green light for building the binaries on the wrong machine (not compatible with sligthly older linux). Sorry Luigi, I had to do svn merge -r 370:369 . svn merge -r 368:366 . Karl just gave us access to the same machine that is used to compile TeX Live binaries (debian sarge), but due to lack of rsync and svn I had troubles trying it out and then ran out of time. Arthur - do you have some time to try to make the compilation work there? (We probably need external call of rsync and svn, so the script probably needs to be adapted a tiny bit.) Mojca

2 1

[Context] latest LuaTeX
by Mojca Miklavec 19 Feb '09

19 Feb '09

Hello, The experimental branch of minimals is not quite functional: since rsync doesn't work the way I would want it to work, the initial idea to provide patches in experimental branch only is not really usable. I would like to collect some ideas how to match context version (current, beta, (alpha - not too important), experimental) with the proper luatex version. First question (for Taco): pdftex and xetex are probably OK. We need luatex, do we need some flavour of experimental metapost versions as well? Second question: I have set up rsync to provide current/luatex (latest beta) current/pdftex (svn) current/xetex (svn) current/metapost (latest beta) latest/luatex (svn) latest/pdftex (svn) latest/xetex (svn) latest/metapost (svn) We can probably drop latest/pdftex and latest/xetex, Taco should tell about metapost. The automatic procedure builds the first four. We probably need to compile latest/luatex by default as well. But then the next question ... how to match context version with the right luatex version? My idea is that whenever one requests --context=experimental then luatex binary could be taken from experimental/bin/luatex/platform/ instead of current/bin/luatex/platform/, but what about ConTeXt beta and what to do when Taco releases a new luatex beta and Hans doesn't release new ConTeXt "current"? Hans - I need some of your ideas here. At the moment both ConTeXt current and beta on the garden are freezed (do not match the version on PRAGMA) in order not to fail completely since we have no latest luatex binaries in the repository. Suggestions? Your release plans for the near future? Mojca PS: I will be away for the next two weeks. There might be some internet connection, but I'm not sure about it.

3 2

[Context] How to ignore specific tree when extracting version from SVN?
by Mojca Miklavec 16 Feb '09

16 Feb '09

Hello, When building the binaries for minimals I decided to use the latest SVN version of XeTeX. On the server I use: cd TRUNK && svn info | grep "Last Changed Rev:" | sed 's/^Last Changed Rev: $.*$$/\1/' && cd .. to determine version of XeTeX and whenever the version changes it triggers rebuilding the binaries. But I just figured out that since Jonathan's last commit about two months ago there have been 71 SVN commits into polyglossia (8) and xepersian (all the rest). This means that any fluent binary builder had to compile XeTeX 71 times (and SVN had to store changes) in the last two months for nothing as texmf tree doesn't influence the binaries. On the contrary, ignoring any changes in texmf tree would mean compiling it just twice or so since August. Any volunteer to send me one-liner to fix that? I'm able to write a ruby script to filter that out, but I'm sure that there must be some extremely simple one-liner (or three-liner) in bash to solve that. Thanks a lot for any hints, Mojca

3 3