Archive rebuild proposal for Lucid

Michael Bienia michael at bienia.de
Thu Oct 29 10:04:20 GMT 2009


On 2009-10-29 10:25:39 +0100, Morten Kjeldgaard wrote:
> I think this is a very sensible thing to do. It's been a problem in  
> karmic that the rebuild happened so late in the cycle;  many of the  
> packages on the FTBFS list had -*ubuntu1 tags, so it would be relevant  
> to take care of these problems very early on in the cycle, especially  
> because many of the problems are toolchain related.

We should do an archive rebuild twice: once at the beginning of a
release cycle to see what doesn't build anymore with the new toolchain
and to have enough time to fix it and once near the end to catch those
packages which broke during the development of the next release.

E.g. while karmic changed very early to gcc-4.4, it only changed from
glibc 2.9 to eglibc 2.10 in the middle of karmic schedule. The
combination of g++ 4.4 and eglibc 2.10 enforces a more strict ISO C++
compliance which caused several FTBFS found by the late archive rebuild.
You wouldn't find them in an early archive rebuild.

> In addition, I suggest that the rebuilding script files a bug on LP  
> when a package FTBFS. This will help us get the problems fixed in the  
> normal work-flow, and the bug-fix history of the package will be  
> recorded under +source which will help Debian maintainers and ourselves.

Please don't or only if you can filter out false positives. Some FTBFS
are caused by one arch being faster than an other and thus a
build-dependency being temporarily not installable.

Michael



More information about the ubuntu-devel mailing list