model-writtenrun 4669 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Peter Eisentraut proposed accepting B as a byte-unit alias to match configuration parameters, choosing this over adding 'bytes' support to GUC, although the initial proposal incorrectly named pg_size_pretty() #1.
- Justin Pryzby clarified that the change affects pg_size_bytes() input and requested an explanation of duplicate unit entries #2; Peter corrected the patch subject and added an alias comment in #3.
- David Rowley demonstrated that aliases in the shared formatting table could produce inconsistent output for larger units and proposed a separate alias table #4, prompting Peter to rework the implementation accordingly in #5.
- David recommended referencing the original unit table by index to avoid nested name searches #6, while Dean Rasheed proposed a fully separate input-unit table and identified documentation attached to the wrong function #7.
- David favored the smaller alias table because additional nonbinary units were not planned #8, and Dean acknowledged that existing behavior effectively fixed MB at its binary meaning #9; Peter retained the alias design, adopted direct indexing, and revised the documentation in #10.
- David reviewed the final submission, spotted a comment typo, and requested explicitly documenting PB instead of using 'etc.' in the pg_size_pretty() unit list #11.
- Peter split the pg_size_pretty() documentation hunk from #10 into af4d5715fc, explicitly listing PB and backpatching the correction, then committed the remaining B-alias feature with light editorial fixes as ce1215d9b0.
- Peter confirmed both changes had landed in #12, and David acknowledged the fix in #13, leaving no outstanding patch in the thread.