Hi Julien, On Sat, Oct 08, 2016 at 11:47:29AM +0200, Julien Cristau wrote: > Is there a reason cmdtest can't be m-a: foreign? Wouldn't that fix the > issue as well?
I agree that marking cmdtest m-a:foreign would fully resolves this bug. I'm less sure that such a marking would be correct. cmdtest includes a python module yarnlib, which uses architecture dependent Python extensions. It seems to me that putting it in a m-a:foreign package runs afoul the multi-arch interpreter problem. So the correct solution could be splitting cmdtest into python-yarnlib, switching users of yarnlib over to the separate package and then marking cmdtest m-a:foreign. Given the necessary package split here, I considered annotating Build-Depends of 5 packages more economical. If you disagree with that trade-off, I suggest: tags -1 - moreinfo patch reassign -1 cmdtest retitle -1 make cmdtest m-a:foreign Helmut