On many systems, GCC14 should be fine: Recent linux distributions have now moved to GCC15, which is a problem. But then changing the build options to use gcc14 works fine (see https://groups.google.com/g/sage-devel/c/tJCbTnelvSg/m/IfzQF0HGBQAJ ). On those same distributions, sagemath is running fine on Python 3.13. It look to me the readme might have not been updated for a while (or for some reason you have a really old one ... in which case that could be a lead to follow why you're getting inconsistent build results.
On Wednesday, 27 August 2025 at 11:09:22 UTC-7 Richard_L wrote: > Nils - It's always possible, but I don't see any obvious culprits in my > path. OTOH, I do see is that readme calls for "GCC versions from 8.4.0 to > 13.x" whereas I'm running gcc 14.3.0. (Has anyone else tested gcc 14.x?) I > also see that readme calls for python3 versions "3.9.x to 3.12.x" whereas > I'm running python3 3.6.15, which induces a build of python by the build > script. Since the symbol in question is in a cpython .so file, the bug may > have to do with sage's building of python3-3.12.5 in > "~/sage-10.7/local/var/lib/sage/venv-python3.12.5". (Has anyone else tested > with the script-built python3-3.12.5?) > > On Tuesday, August 26, 2025 at 5:32:23 PM UTC-7 Nils Bruin wrote: > >> A loader symbol lookup error probably means that the dynamic library >> loaded during runtime differs from the one that was assumed during compile >> time. "agg" is a component of matplotlib, so it could be that the >> inconsistency lies within matplotlib itself. Could it be the case that an >> old version of a library is hanging around in ~/.local or something like >> that, which ends up shadowing the actual (hopefully freshly installed) >> version of the library? There are some big changes in the >> build/installation system in sage-10.7, so it could be that an artifact >> that was previously harmless now gets in the way. While you're waiting for >> an expert, you could try and see if building/running in a cleaner >> environment makes the problem go away. >> >> On Tuesday, 26 August 2025 at 16:14:39 UTC-7 Richard_L wrote: >> >>> 1) Sorry, I forgot to mention, I am building sage-10.7 from source on >>> openSUSE-15.6. >>> 2) It turns out not to matter whether sage builds matplotlib or if it is >>> installed from the OS repository. >>> >>> Suggestions welcome. >>> >>> On Tuesday, August 26, 2025 at 8:35:04 AM UTC-7 Richard_L wrote: >>> >>>> All(?) tests invoking matplotlib show at least one of two error >>>> messages. See attachment "plot_fail". However, the build finished and >>>> reported success (install.snip), while 'make ptestlong' shows 200+ errors >>>> (test.log). Neither matplotllb nor matplotlib-inline report failure. [ >>>> Note: install.log too big to attach; therefore the tail of install.log >>>> attached here as install.snip. ] >>>> Please inform if some other log file is needed to debug this. >>>> Thanks >>>> >>> -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/sage-devel/dda2e7c5-1b4a-4307-9a95-0c59a0f861b2n%40googlegroups.com.
