While fixing a CI issue with sci-libs/umfpack-5.6.2, I came upon errors with this packages. It fails to build against sci-libs/amd-2.4.6 and sci-libs/colamd-2.9.6. Additionally it wants to pull in <sci-libs/parmetis-4 which isn't in the tree anymore, so the dependencies for USE=metis can only use <sci-libs/metis-5. Reproducible: Always Steps to Reproduce: 1. emerge =sci-libs/amd-2.4.6 =sci-libs/coldamd-2.9.6 2. emerge =sci-libs/cholmod-2.1.2 3. I'm going to address this with a PR for fixing the umfpack CI issue.
See also bug #716024 and bug #682838
Created attachment 629408 [details] cholmod-2.1.2-20200402-182021.log Build log for building against sci-libs/amd-2.4.6 and sci-libs/colamd-2.9.6
Created attachment 629410 [details] emerge-info.txt
Does anything in the tree need the older versions? I upgraded all of the components of suitesparse to correspond to suitesparse-5.4.0, but I didn't upgrade the metapackage because I didn't realize there was one (bug 711610). So if you don't install the metapackage (or if I version bump it to 5.4.0), you would get all of the new components that work together. In short, it would be simpler to remove the old versions and bump the metapackage if that won't break anything.
I was asked to look after some of the many CI issues currently reported when one opens a PR. This bug came up as a follow-up of such a fix for sci-libs/umfpack. Because it's the stable version, I didn't drop it, as long as 3.0.13 isn't stabilised. It's needed by umfpack-5.6.2, which has the same issue (stable, but current version still unstable). I'm happy with any solution which get's the CI report shortened. I don't run the old version on my desktop and only noticed this due to thorough testing of my changes to umfpack.
I think we have settled on stabilizing the new suitesparse and removing the old one ASAP in bug #716960. (I'm just waiting on the arch teams to add keywords, which I can't do myself.)
