Source: llvm-toolchain-6.0 Version: 1:6.0-3 Severity: normal PostgreSQL 11 is optionally using llvm to jit query execution at runtime. This works fine on amd64 and i386, but fails on all other platforms.
https://buildd.debian.org/status/logs.php?pkg=postgresql-11&ver=11~beta1-1&suite=experimental 2018-05-22 22:23:21.218 UTC [23734] pg_regress/strings STATEMENT: SELECT chr(0); terminate called after throwing an instance of 'std::bad_function_call' what(): bad_function_call 2018-05-22 22:23:21.325 UTC [23614] LOG: server process (PID 23738) was terminated by signal 6: Aborted 2018-05-22 22:23:21.325 UTC [23614] DETAIL: Failed process was running: INSERT INTO TEMP_GROUP SELECT 1, (- i.f1), (- f.f1) FROM INT4_TBL i, FLOAT8_TBL f; 2018-05-22 22:23:21.325 UTC [23614] LOG: terminating any other active server processes The fix for this is r328687 in llvm upstream: http://llvm.org/viewvc/llvm-project?view=revision&revision=328687 (Andres is also the author of the llvm support in PostgreSQL) https://www.postgresql.org/message-id/[email protected] It would be nice if the llvm package in Debian unstable would be updated with this fix. I'm not totally sure about the status of the 6.0.1 release, but I think it doesn't have the patch yet. Thanks, Christoph
