On 2024/01/06 11:17, Daniel Dickman wrote: > > > > On Jan 5, 2024, at 11:40 AM, Stuart Henderson <s...@spacehopper.org> wrote: > > > > On 2024/01/02 12:08, Stuart Henderson wrote: > >>> On 2024/01/01 15:23, Daniel Dickman wrote: > >>> The next version of matplotlib requires numpy 1.25 or newer. > >>> > >>> The below diff updates numpy to 1.25.2 so matplotlib can be updated. > >> ... > >>> p.s. Testing on !amd64 is always useful as there has been > >>> platform-specific breakage in the past. Bulk tests would also be helpful > >>> given how much of the tree depends on numpy. > >> > >> Reading over it LGTM, I will run tests on i386. > > > > Lots of faffing to get numpy to build at all on i386 thanks to the > > "exists" check for ld.lld from llvm/13 (used by lang/gcc), but got there > > in the end, tests looking pretty reasonable: > > > Thanks a lot for testing! I wasn’t sure if you also did a bulk and ran into > problems in any consumers?
I did, and no prblems related to numpy. > I actually think that given how much breakage this update could cause I don’t > think I’m going to be brave enough to commit it. It should probably go > through bulks on all platforms that folks care about to identify any > platform-specific issues ahead of being committed. Bulks on that mamy platforms aren't not going to happen before commit and I don't think need to be done amyway. I'm basically ok with this thpugh given what you've said It would be nice to have build and "make test" of numpy itself on another arch or two (sparc64 would be a good one if anyone has time to do that). > If someone does go ahead and commit it, I’ll be happy to update matplotlib > afterwards.
