Le blog de Olivier Crête

Today, there was a meeting of the Gentoo/AMD64 team to discuss the blubb/dang proposal to support support multilib, but the meeting didn’t get much resolved since eradicator (our multilib man) could not there. I find funny that this proposal has quite a few things in common with my June 2003 proposal: allowing every lib to be built for both API’s, maintaining dep trees for both. Its also funny that killing multilib and chroots are also proposed/rejected. But we’ve had progress since 2003, we have get_libdir() and all of eradicator’s great work. Its nice that we are now discussing minor details like the -config files (like gtk-config, etc)

I find blubb proposal too complicated. Instead of modifying every -config (that’s crazy) or putting both 32bit and64bit -config files in special directories and replacing them in bin/ with special wrappers. Or having a special portage wrapper for each of them.. which would anyway break when they are called inside configure script, etc.. Or would be hard to implement for non-portage builds. We could leave the 64bit ones where they belong (in /usr/bin) and put the 32bit ones somewhere else (lets say /usr/bin/32bitconfigfiles/), with the same name. And then when we want to build a 32bit application, we can just put that directory in the PATH before the /usr/bin.. and there you go, problem solved. We should probably even provide some kind of non-portage script (lets call it build32) that would set PKG_CONFIG_PATH, PATH, a gcc=i686-gcc alias, etc to make building 32bit easy.

After getting my new Athlon64 box, the first thing that I did was to install Gentoo/AMD64. And it has really improved since I left the project.. I was surprised to see that I did not have too many packages to keyword myself, although I did keyword at least a dozen in a last week..

I also started a new job at Maximum Throughput, its my first “real” job (not an internship). I will be working mostly on the next generation of InfinARRAY, their high performance distributed filesystem.