Fix build against PostgreSQL 19

[devrimgunduz]

plprofiler.h:

• PG19's header include-what-you-use (IWYU) cleanup means <storage/lwlock.h>, <storage/shmem.h> and <utils/tuplestore.h> are no longer pulled in transitively by the headers plprofiler.h already included. Without them, LWLockAcquire/LWLockRelease, RequestAddinShmemSpace/RequestNamedLWLockTranche, ShmemInitStruct/ ShmemInitHash/GetNamedLWLockTranche and tuplestore_begin_heap/ tuplestore_putvalues were all implicitly declared, producing the "call to undeclared function" / "incompatible integer to pointer conversion" errors seen on both clang (LLVM bitcode build) and gcc. Add the three missing includes to plprofiler.h.
• profilerSharedState.lock was declared as "LWLockId", a typedef that was removed from PostgreSQL back in 9.4 when LWLocks moved to a pointer-based API. The plprofiler code itself already treats plpss->lock as a pointer (e.g. "plpss->lock = &(GetNamedLWLockTranche(...))->lock;"), so the struct member had been silently relying on an implicit int-to- pointer conversion that recent compilers (and C99+) no longer allow. Change the member to the correct "LWLock *" type.

plprofiler.c:

• PG19 changed the signature of ShmemInitHash() from HTAB *ShmemInitHash(const char *name, long init_size, long max_size, HASHCTL *infoP, int hash_flags); to HTAB *ShmemInitHash(const char *name, int64 nelems, HASHCTL *infoP, int hash_flags); collapsing the separate init_size/max_size arguments into a single nelems argument. Both call sites in profiler_shmem_startup() (for the "plprofiler functions" and "plprofiler callgraph" shared hash tables) were passing the old 5-argument form, which now fails to compile ("too many arguments to function call" / argument-type mismatches). Drop the redundant init_size argument at both call sites so only one size argument (nelems) is passed.

Hacked by Claude, tested by me.

Per pgdg-packaging/pgdg-rpms#213

[luss]

kewl

…

On Fri, Jul 10, 2026 at 5:02 AM Devrim Gündüz ***@***.***> wrote: plprofiler.h: - PG19's header include-what-you-use (IWYU) cleanup means <storage/lwlock.h>, <storage/shmem.h> and <utils/tuplestore.h> are no longer pulled in transitively by the headers plprofiler.h already included. Without them, LWLockAcquire/LWLockRelease, RequestAddinShmemSpace/RequestNamedLWLockTranche, ShmemInitStruct/ ShmemInitHash/GetNamedLWLockTranche and tuplestore_begin_heap/ tuplestore_putvalues were all implicitly declared, producing the "call to undeclared function" / "incompatible integer to pointer conversion" errors seen on both clang (LLVM bitcode build) and gcc. Add the three missing includes to plprofiler.h. - profilerSharedState.lock was declared as "LWLockId", a typedef that was removed from PostgreSQL back in 9.4 when LWLocks moved to a pointer-based API. The plprofiler code itself already treats plpss->lock as a pointer (e.g. "plpss->lock = &(GetNamedLWLockTranche(...))->lock;"), so the struct member had been silently relying on an implicit int-to- pointer conversion that recent compilers (and C99+) no longer allow. Change the member to the correct "LWLock *" type. plprofiler.c: - PG19 changed the signature of ShmemInitHash() from HTAB *ShmemInitHash(const char *name, long init_size, long max_size, HASHCTL *infoP, int hash_flags); to HTAB *ShmemInitHash(const char *name, int64 nelems, HASHCTL *infoP, int hash_flags); collapsing the separate init_size/max_size arguments into a single nelems argument. Both call sites in profiler_shmem_startup() (for the "plprofiler functions" and "plprofiler callgraph" shared hash tables) were passing the old 5-argument form, which now fails to compile ("too many arguments to function call" / argument-type mismatches). Drop the redundant init_size argument at both call sites so only one size argument (nelems) is passed. Hacked by Claude, tested by me. Per pgdg-packaging/pgdg-rpms#213 <pgdg-packaging/pgdg-rpms#213> ------------------------------ You can view, comment on, or merge this pull request online at: #27 Commit Summary - 8e9173e <8e9173e> Fix build against PostgreSQL 19 File Changes (2 files <https://github.com/bigsql/plprofiler/pull/27/files>) - *M* plprofiler.c <https://github.com/bigsql/plprofiler/pull/27/files#diff-60116c4d2aebd136948b7c18a16861854e322b9b0449757118aabd99a81513c7> (2) - *M* plprofiler.h <https://github.com/bigsql/plprofiler/pull/27/files#diff-0978dd799f513cf12a996eaadfb8a216766610437dbab99a0eecc85a8f17e7b5> (5) Patch Links: - https://github.com/bigsql/plprofiler/pull/27.patch - https://github.com/bigsql/plprofiler/pull/27.diff — Reply to this email directly, view it on GitHub <#27?email_source=notifications&email_token=AAMWOHRVY3YG34HLM76NSDL5ECWL7A5CNFSNUABEM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UF42DAMRYGEZTAOBQGGTHEZLBONXW5KTTOVRHGY3SNFRGKZFFMV3GK3TUVRTG633UMVZF6Y3MNFRWW>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAMWOHWID4GGTKOYKWNTXUD5ECWL7AVCNFSNUABFKJSXA33TNF2G64TZHMZDCOBTG4YDMNZQHNEXG43VMU5TIOBVGM3DANBVHA4KC5QC> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[adeshthack]

is there any update on pg19 compatibility?