Roland Levillain <roland@lrde.epita.fr> writes:

Olena's BuildBot is up and running since April 8, 2010.

Yesterday we had our first successful automated builds in months since the migration to Git! Two of the build slaves successfully went through configuration, compilation, tests and distchecking for branch `next':

https://buildfarm.lrde.org/buildfarm/oln/waterfall?branch=next&reload=none

* i486-linux-gnu-gcc-4.1 * i486-linux-gnu-icc-11.0

The other builds have failed for various reasons:

1. powerpc-apple-darwin8-gcc-4.0.1: Host is offline. 2. i486-linux-gnu-gcc-3.3: Compilation failure (to be addressed). 3. i486-linux-gnu-gcc-4.2: Known weird behavior/bug of g++ 4.2. 4. i486-linux-gnu-icc-10.1: Compiler suddenly no longer working (weird). 5. x86-64-linux-gnu-gcc-4.1: Build host in bad condition.

Cases #1 and #5 are hardware issues. Problem #2 needs some examination. I know how to work around issue #3. Case #4 is weird: previous builds of `next' never exhibited such a compiler issue.

I'll try to fix #2 then #3. And I'll try to diagnose #4.

Done.

We need to work on #1 and #5 with Geoffroy.

Build #1 on antalya is probably outdated now. We want a fast Mac OS X 10.5 or 10.6 host instead. I'll probably disable the powerpc-apple-darwin8-gcc-4.0.1 build. As for #5, I have just installed a build slave for x86-64-linux-gnu-gcc-4.1 on seattle. Wait and see. [...]

I think we also need to set up more build hosts. For instance, I'd like to have builds

* using g++ 4.4 on GNU/Linux * using g++ 4.3 on GNU/Linux * on Cygwin (either using g++ 3.4 or g++ 4.x) * possibly with MinGW * possibly using gcc-snapshot (the future g++ 4.5) * possibly on Mac OS X 10.5 or 10.6

And builds using llvm-gcc and llvm-clang, too.