Apr 2026

1 commit
jit: Change the default to off.
2 Apr 2026Tomas Vondramaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #111 of 11 reviewers saw the final version

The commit preserves the three default-setting changes from #1 in src/backend/jit/jit.c, src/backend/utils/misc/guc_parameters.dat and src/backend/utils/misc/postgresql.conf.sample. It removes the historical-default parenthesis from doc/src/sgml/config.sgml and rewrites the commit message, exactly as Tomas Vondra reports in #19.

  • Jim Jonessaw the final versionLLM analysis, 88% sureJim endorsed changing the default and reported having repeatedly encountered JIT problems in #2.
  • Laurenz Albesaw the final versionLLM analysis, 85% sureLaurenz endorsed the change and concurred with the reported JIT problems in #3.
  • Bernd Helmlesaw the final versionLLM analysis, 91% sureBernd approved disabling JIT by default based on affected users and the ability to enable it explicitly in #4.
  • wenhui qiusaw the final versionLLM analysis, 80% surewenhui explicitly agreed with the proposed default and enabling JIT when needed in #5.
  • Euler Taveirasaw the final versionLLM analysis, 94% sureEuler reviewed the compatibility consequences for OLAP users and argued for documentation or threshold changes instead in #6.
  • Pierre Ducroquetsaw the final versionLLM analysis, 95% surePierre challenged disabling all JIT and proposed targeting tuple deforming instead in #7, renewing that objection in #15.
  • Christoph Bergsaw the final versionLLM analysis, 97% sureChristoph supported the change after examining installation defaults, workload predictability and the limitations of documentation or threshold adjustments in #8.
  • Greg Sabino Mullanesaw the final versionLLM analysis, 92% sureGreg endorsed disabling JIT by default unless a reliable costing adjustment could prevent inappropriate activation in #10.
  • Andres Freundsaw the final versionLLM analysis, 99% sureAndres endorsed the change based on partitioning and LLVM costs in #11 and explained why disabling tuple deforming would not address the dominant compilation costs in #17.
  • Tom Lanesaw the final versionLLM analysis, 88% sureTom endorsed proceeding with the change before beta so that adverse reactions would surface earlier in #14.
  • Tomas Vondrasaw the final versioncommittercommitted the patch

Credits across 1 commit

Author
Jelte Fennema-Nio1
Reviewer
Tomas Vondracommitter1
Committer
Tomas Vondra1