lalpulsar tests failing with frame reading segfaults on clang:13-15 and gcc:7
All recent pipelines, on master or not, fail with two sets of failures: one is a runner-specific "out of space" issue for wheels jobs (-> https://git.ligo.org/computing/helpdesk/-/issues/5407 ) while the others are compiler tests failing with Segmentation fault (core dumped)
in either lalpulsar_Makefakedata_v5
or
lalpulsar_MakeSFTs
. This happens for clang:13-15 and gcc:7, on runners different from the out-of-space one, at both CIT and UWM. Oddly enough, the older clang:12 and all newer gcc seem to be fine.
Example pipeline from this morning: https://git.ligo.org/lscsoft/lalsuite/-/pipelines/596569
My only guess is that some dependency version must have changed in a way that the combination of those specific compilers and lalpulsar is allergic to... ideas @karl-wette @duncanmmacleod @adam-mercer ?