Commits on Aug 21, 2026
-
ecpg: Fix auto_mem cleanup on thread exit.
The pthread key destructor registered to free memory allocated with ecpg_auto_alloc() on thread exit had two problems: 1. It didn't do anything, because the OS clears the thread-specific slot (ie the value that get_auto_allocs() will return) before calling the destructor. We need to use the value that is passed to the destructor as an argument. 2. It was using ECPGfree_auto_mem(), but that would free the actual allocated values, and not just the list structure. The values are supposed to be freed by user code (for example the free(r) call in src/interfaces/ecpg/test/thread/alloc.pgc). We need to use ecpg_clear_auto_mem() instead, or valid .pgc code will double-free the most recently auto-allocated objects on thread exit. The destructor still doesn't run at all on Windows, but a proposed patch will address that problem. Discussion: https://postgr.es/m/CA%2BhUKGLv202ndx%3DkP1fZJapNcbfnpq-rezHNyt%2BcQ4pdxr1akg%40mail.gmail.com Backpatch-through: 14 Reviewed-by: Thom Brown <thom@linux.com>
authored and Commitfest Bot committed
Aug 21, 2026 -
port: Provide pg_threads.h API.
This is based on C11 <threads.h>, which is not available on all systems we support and a little light on features, but still provides a standards-based way to nail down a lot of gnarly details and avoid having to invent our own strange names for things. Our API has pg_ prefixes on all identifiers, since client code might be using <threads.h> directly. The thread_local macro is an exception and doesn't get a prefix, since there is no doubt about its definition. Two implementations are provided to redirect the API to underlying threading APIs: * <pthread.h>, a thin wrapper for POSIX and MinGW systems * <threads.h>, a thin wrapper for Visual Studio's C11 threads A native Windows version using traditional Windows APIs is also possible. Some local extensions are also provided in pg_threads_ext.h: * pg_rwlock_t * pg_thrd_barrier_t * pg_thrd_error_string() * pg_thrd_error_string_with_detail() * PG_CND_INIT * PG_MTX_PLAIN_INIT Later commits will make use of this API as a centralized replacement for several module-local thread API abstractions. Discussion: https://postgr.es/m/CA%2BhUKG%2BK%2BLmYtz54%2BSv1j5Zu1Qo6szzAMbMyuKSeQmb7De4Ewg%40mail.gmail.com Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi> Reviewed-by: Bryan Green <dbryan.green@gmail.com>
macdice authored and Commitfest Bot committed
Aug 21, 2026 -
* adopt pg_threads.h for mutexes * remove a local pthreads abstraction Discussion: Reviewed-by:
macdice authored and Commitfest Bot committed
Aug 21, 2026 -
pg_basebackup: Use pg_threads.h.
Previously, Windows systems used a thread to stream WAL, and POSIX systems used a subprocess. Now all systems use a thread. Discussion: Reviewed-by:
macdice authored and Commitfest Bot committed
Aug 21, 2026 -
pg_signal_processor: Infrastructure for cleanup.
Provide a way to register signal handlers than run serially in a special thread. This is intended to allow frontend programs to use more complicated shutdown code when you hit ^C, without having to worry about async-signal-safety. XXX experimental Discussion: Reviewed-by:
macdice authored and Commitfest Bot committed
Aug 21, 2026 -
fe_utils: Provide cancel_set for fast quit paths.
Provide a simpler way for frontend programs managing many connections to arrange cleanup. Now that we can run more complicated code on eg SIGINT via pg_signal_processor, this API provides an easy way to manage a set of connections that need to be canceled. We couldn't have done this before, because it requires locks. A follow-up patch will use this for pg_dump/pg_restore. XXX Experimental! Discussion: Reviewed-by:
macdice authored and Commitfest Bot committed
Aug 21, 2026 -
XXX pg_dump: Refactor to use threads on all platforms.
XXX See Bryan Green's patches. This was just an experiment that grew out of switching to pg_thread.h and then accidentally continuing to hack. Clearly we also need to get rid of the pipes that were only needed because of the make -j architecture, which his patches address. Discussion: Reviewed-by:
macdice authored and Commitfest Bot committed
Aug 21, 2026
