Sep 2021

3 commits
process startup: Split single user code out of PostgresMain().
18 Sep 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #81 of 1 reviewer saw the final version

The single-user extraction matches #8, including PostgresSingleUserMain and removal of manufactured argv from BackendRun. Differences are comments in src/backend/tcop/postgres.c, including the switch-parser description and startup-time explanation.

  • Kyotaro Horiguchisaw the final versiontrailerHe discussed the single-user entry-point design in #4 and explicitly approved the revised patch after examining processing-mode initialization in #9.
process startup: Do InitProcess() at the same time regardless of EXEC_BACKEND.
16 Sep 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surelanded as postedfrom patchset #81 of 1 reviewer saw the final version

The added and removed lines match the InitProcess patch in #8: BackendStartup initializes PGPROC, while PostgresMain retains that responsibility only for standalone startup.

  • Kyotaro Horiguchisaw the final versiontrailerIn #9, he explicitly said the InitProcess patch looked fine in conjunction with the single-user extraction.
process startup: Initialize PgStartTime earlier in single user mode.
15 Sep 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #81 of 1 reviewer saw the final version

The PgStartTime relocation in src/backend/tcop/postgres.c matches #8. The commit only expands the comment explaining that its timing roughly matches postmaster startup.

  • Kyotaro Horiguchisaw the final versionLLM analysis, 100% sureHe explicitly approved the PgStartTime patch in #9, and 2c7615f77b credits him through a misspelled Reviewd-By trailer.

Aug 2021

9 commits
pgstat: Schedule per-backend pgstat shutdown via before_shmem_exit().
7 Aug 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1Also in: shared-memory based stats collector (part of a posted patchset)Also in: Dividing line between before_shmem_exit and on_shmem_exit?
pgstat: Bring up pgstat in BaseInit() to fix uninitialized use of pgstat by AV.
7 Aug 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1Also in: shared-memory based stats collector (part of a posted patchset)
process startup: Always call Init[Auxiliary]Process() before BaseInit().
5 Aug 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surelanded as postedfrom patchset #11 of 1 reviewer saw the final version

The initialization-order patch from #1 matches across all eight files, modulo context and file ordering. It places InitProcess or InitAuxiliaryProcess before BaseInit, makes standalone shared-memory creation explicit, and removes InitBufferPoolBackend.

  • Kyotaro Horiguchisaw the final versiontrailerHe examined startup sequencing and standalone initialization in #4, and the credited patch landed without substantive changes.
process startup: Centralize pgwin32_signal_initialize() calls.
5 Aug 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 99% surelanded as postedfrom patchset #12 of 2 reviewers saw the final version

The pgwin32_signal_initialize relocation from #1 is preserved in src/backend/utils/init/miscinit.c, with corresponding removals from main.c and postmaster.c. The main.c hunk is adjusted for the checker-mode patch having landed earlier in the sequence.

  • Kyotaro Horiguchisaw the final versiontrailerHe reviewed startup design in #4, and Andres acknowledged his review when announcing the initial cleanups in #6.
  • Robert Haassaw the final versiontrailerHe characterized the initial cleanup group as mechanical and uncontroversial in #5, and Andres acknowledged that review in #6.
process startup: Remove bootstrap / checker modes from AuxProcType.
5 Aug 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #12 of 2 reviewers saw the final version

The AuxProcType cleanup and --check interface match #1. The commit renames check_mode to check_only, documents its purpose, and additionally removes the obsolete AmBootstrapProcess macro from src/include/miscadmin.h.

  • Kyotaro Horiguchisaw the final versiontrailerHe examined checker mode's relationship to bootstrap and auxiliary processes in #4, after which only light cleanup changes followed.
  • Robert Haassaw the final versiontrailerHe described the initial cleanup group as mechanical and uncontroversial in #5, with his review acknowledged in #6.
process startup: Move AuxiliaryProcessMain into its own file.
5 Aug 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #12 of 2 reviewers saw the final version

The move into src/backend/postmaster/auxprocess.c and its new header matches #1. The landed version polishes file-header descriptions and identification blocks, with inherited comment reflow but no functional redesign.

  • Kyotaro Horiguchisaw the final versiontrailerHe examined the bootstrap and auxiliary-process separation in #4, and Andres acknowledged his review in #6.
  • Robert Haassaw the final versiontrailerHe described the initial mechanical cleanups favorably in #5, and Andres thanked him when announcing their landing in #6.
process startup: auxprocess: reindent block
5 Aug 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 99% suremodified before commitlight changesfrom patchset #12 of 2 reviewers saw the final version

The block unindent in src/backend/bootstrap/bootstrap.c matches #1. The committed version additionally reflows the PGPROC and ProcSignalSlot comments through pgindent.

  • Kyotaro Horiguchisaw the final versiontrailerHe reviewed the surrounding startup separation in #4, with his contribution acknowledged in #6 before only formatting changes followed.
  • Robert Haassaw the final versiontrailerHe regarded the initial cleanup group as mechanical and uncontroversial in #5, which Andres acknowledged in #6.
process startup: Separate out BootstrapModeMain from AuxiliaryProcessMain.
5 Aug 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #12 of 2 reviewers saw the final version

Applying the corresponding patch from #1 to the commit parent produces the same code apart from indentation and an expanded checker-mode comment in src/backend/bootstrap/bootstrap.c. The comment explains why a fully separate checker implementation would currently duplicate too much code.

  • Kyotaro Horiguchisaw the final versiontrailerHe discussed whether checker mode should remain with bootstrap and supported a separate auxiliary-type argv entry in #4.
  • Robert Haassaw the final versiontrailerHe characterized the initial separation and cleanup work as mechanical and uncontroversial in #5, with his review acknowledged in #6.
process startup: Rename postmaster's --forkboot to --forkaux.
5 Aug 2021Andres Freundmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surelanded as postedfrom patchset #12 of 2 reviewers saw the final version

All three --forkboot to --forkaux replacements in src/backend/postmaster/postmaster.c match the first patch in #1.

  • Kyotaro Horiguchisaw the final versiontrailerHe examined the auxiliary-process startup interface in #4, and Andres acknowledged his review of the initial cleanups in #6.
  • Robert Haassaw the final versiontrailerHe described the initial cleanup group as mechanical and uncontroversial in #5, and the rename landed unchanged.

Credits across 12 commits

Author
Andres Freund12
Kyotaro Horiguchi1
Reviewer
Kyotaro Horiguchi9
Robert Haas6
Committer
Andres Freund12