model-writtenrun 4054 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Bharath Rupireddy proposed replacing the misleading snapshot-identifier error with a standard file-open diagnostic containing the path and operating-system error in #1, and Michael Paquier supported that approach in #2.
- Yogesh Sharma suggested explicitly calling it a snapshot file in #3, but Bharath argued that the pg_snapshots path already supplies that context in #4; Daniel Gustafsson supported the existing generic wording because it also avoids extra translation work in #5.
- Michael applied the original patch unchanged as a0d87bcd9b and announced backpatching through PostgreSQL 11 in #6.
- Andres Freund objected that missing snapshots should produce a user-facing diagnostic rather than expose an internal filename in #7, suggested treating ENOENT separately, and requested a revert because changing the missing-snapshot SQLSTATE could break callers in #8.
- Michael initially defended distinguishing malformed identifiers from file-access failures in #9, then reverted the first landing and its backports with be022908cf; in #10 he acknowledged Andres's concern and supplied a development-branch-only replacement with separate ENOENT handling and two regression tests.
- Andres proposed the wording "snapshot does not exist" and ERRCODE_UNDEFINED_OBJECT in #11, while agreeing against backpatching; Michael incorporated both suggestions into #12.
- Yogesh endorsed the revised diagnostic in the thread's last message, #13, and the final attachment subsequently landed unchanged as 78a33bba4c, completing the improvement on the development branch while stable branches retained their original behavior.