model-writtenrun 7129 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Tristan Partin proposed preventing NULL arguments from reaching memcpy, suppressing expected executable leaks, and experimentally increasing max_stack_depth, explicitly warning that the stack-limit change did not work and must not be committed #1; he supplied the three patches in #2.
- Heikki Linnakangas explained that Neon should pass an empty string instead of NULL, recommended checking message size if empty registrations were skipped, and questioned why additional sanitizer support was needed when CI already used it #3.
- Andres Freund explained that ASan’s shadow stack confused PostgreSQL’s stack-depth checks and supplied detect_stack_use_after_return=0 as a workaround while favoring a compatible stack check #4; he also recommended compiler-provided sanitizer detection and a shared static library for executable sanitizer options #5.
- Alexander Lakhin reported additional XML and pg_bsd_indent sanitizer issues and found that clang-18 still failed the subtransaction test with the workaround, although clang-16 passed #6; Tristan agreed to investigate #7.
- Tristan’s replacement series dropped the logical-message fix in favor of changing Neon, disabled ASan instrumentation on stack-checking functions, centralized executable leak suppression, and added PL/Python instrumentation and suppression rules #8; he reported passing ordinary tests but left sanitizer detection, naming, a reproduced XML leak, and Alexander’s additional issues unresolved.
- Months later, Junwang Zhao reported that the original executable patch still allowed an initdb failure from save_ps_display_args leak reports and that the replacement series no longer applied cleanly #9; the thread ended with Tristan promising a rebase, without another submission or a reported landing #10.
- The later Debian Trixie CI migration cited Andres’s workaround #4, but its entire committed patch came from Nazir Bilal Yavuz’s later submission #602988 in thread #52442, so be9efd4929 belongs to that thread and does not establish a landing of either sanitizer series here.