model-writtenrun 8377 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Thomas Munro opened the thread with #1, removing obsolete operating-system configuration instructions and replacing insecure FreeBSD jail advice in doc/src/sgml/runtime.sgml.
- Peter Eisentraut supported the cleanup in #2, while Tom Lane proposed judging old advice by its relevance to fresh PostgreSQL 13 installations and also removing oom_adj instructions in #3; discussion with Magnus Hagander clarified the old-Linux support concerns and the importance of retaining older-version manuals in #4, #5, #6, and #7.
- Thomas confirmed that the cleanup would affect PostgreSQL 13 rather than backbranches and expanded it in #8 to explain modern IPC defaults, qualify advice that only applies to shared_memory_type=sysv, and remove additional obsolete kernel and PostgreSQL guidance.
- Tom supplied three final wording adjustments in #9, including directing Linux users to consult sysctl.conf because it can point to subsidiary files; the revised patch and delta landed unchanged together as c8be915aa9.
- After confirming the cleanup was pushed, Thomas noticed in #10 that the systemd RemoveIPC example still showed a semaphore failure no longer expected on standard Linux builds; Tom explained in #11 that attached System V segments were protected but POSIX shared-memory objects remained vulnerable.
- Thomas posted a separate correction in #12 showing missing shared-memory objects during parallel queries and proposed backpatching to PostgreSQL 10; he declined to document switching dynamic shared memory to sysv because fixing systemd configuration was preferable and unattached System V segments could still be vulnerable.
- Tom caught a spelling error in #13, and the systemd correction landed with only that fix as d094bf9301, with backports through PostgreSQL 10.
- With both documentation changes committed, Peter suggested defaulting both Linux shared-memory settings to sysv in #14, but Tom explained in the final message #15 that this would not fully protect dynamic shared memory and was unnecessary for the primary mapping inherited through fork; no further patch remained pending.