Bugzilla – Bug 1222061
PostgreSQL15-Server fails to load PostGIS-Bitcode
Last modified: 2024-04-03 07:38:16 UTC
When trying to run a statement that utilises PostGIS features the following error is reported: FATAL: failed to load summary "/usr/lib/postgresql15/lib64/bitcode/postgis-3.index.bc": Invalid summary version 9, 1, 2, 3 or 4 expected Searching for this error reports repo-mismatch (distro-provided postgis and postgresql from PostgreSQL, see [1]), which does (probably) not apply here (postgresql-server is from 'SLES 15 Updates for openSUSE Leap 15.5' and postgis is from Application:Geo for Leap 15.5). The following postgres packages are installed: postgresql15, version 15.6-150200.5.22.1 postgresql15-contrib, version 15.6-150200.5.22.1 postgresql15-debugsource, version 15.6-lp155.44.1 postgresql15-devel, version 15.6-150200.5.22.1 postgresql15-llvmjit, 15.6-150200.5.22.1 postgresql15-llvmjit-debuginfo, version 15.6-lp155.44.1 postgresql15-llvmjit-devel, version 15.6-150200.5.22.1 postgresql15-postgis, version 3.4.2-lp155.37.2 postgresql15-postgis-utils, version 3.4.2-lp155.37.2 postgresql15-server, version 15.6-150200.5.22.1 postgresql15-server-devel, version 15.6-150200.5.22.1 An apparmor-profile for postgres is used, but the error also happens if it is not enforced (aa-complain). [1] https://gis.stackexchange.com/questions/377833/postgis-wont-update-geometry-when-importing-csv-file-with-psql
It is a repo mismatch, but in a different way. The distro-provided PostgreSQL packages on Leap 15.5 are actually being built on SLE-15-SP2 (see "150200" in the version string) using LLVM 7, whereas the PostGIS packages for Leap 15.5 from Application:Geo ("lp155" in the version string) are built using the much newer LLVM 15. There are two options to solve this: 1. Deinstallation of postgresql15-llvmjit, so that LLVM is not being used at all 2. Using the PostgreSQL packages from server:database:postgresql, which are build with LLVM 15, so the postgis packages should be compatible with them.
Thanks you for the quick answer. :) 1. resolved the issue and I'll keep 2. in mind. Bug can be closed - I don't know if I can do that myself and which sub-status for RESOLVED to use
Thanks for the feedback, I close it as INVALID, because it was an installation mistake rather than a bug in the code or packaging. But I'll investigate if there is a way to better report such mismatches than the rather cryptic error message you got.