On 2016-05-10 20:48, Ralf Treinen wrote: > Hi, > > On Tue, May 10, 2016 at 11:18:56AM +0200, Aurelien Jarno wrote: > > > We use sbuild wit the cudf resolver on the build daemons, with the > > following optimization criteria: > > > > $aspcud_criteria = > > '-removed,-changed,-new,-count(solution,APT-Release:=/experimental/)'; > > > > Unfortunately this doesn't work when the available packages are newer > > than the one installed in the chroots. > > That is very strange. As a first remark, the optimisation criterion should > not be at stake here. The optimisation criterion cannot make the search for > a solution fail in case there is a solution, the criterion only > serves to decide which solution to select in case there are several > solutions.
Confirmed, I am able to reproduce the problem with only 'apt-get --solver aspcud build-dep qgis'. Note that 'apt-get build-dep qgis' succeed. > > This can also be reproduced using the following command: > > > > | apt-get --solver aspcud -o 'APT::Solver::Strict-Pinning=false' -o > > 'APT::Solver::aspcud::Preferences=-removed,-changed,-new,-count(solution,APT-Release:=/experimental/)' > > build-dep qgis > > A quick test on my home machine does not reproduce the error. Anyway, I guess > you have to specify that you want to satisfy the build-dependencies of > qgis/experimental, I guess ? I have been able to reproduce the problem on mips, but we have seen the problem on various architectures, though for different packages, so probably with different conditions. I am able to reproduce the problem even with only unstable in the sources.list. Unfortunately it seems that qgis build dependencies are currently not available on amd64. I'll try to see if I see a similar issue on amd64 or i386, try to reproduce it and report it here. Thanks, Aurelien -- Aurelien Jarno GPG: 4096R/1DDD8C9B aurel...@aurel32.net http://www.aurel32.net
