model-writtenrun 7938 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- David Turon reported that psql's table description cut off a stored generated-column expression while still displaying its closing syntax, making the result misleading #1.
- Tom Lane identified the longstanding 128-character limit, and Bruce Momjian initially favored keeping it; Tom then explained why wrapping a truncated expression in generated-column syntax was particularly confusing and raised a possible memory leak #2 #3 #4.
- David TuroĊ suggested an ellipsis or configurable truncation, Peter Eisentraut advocated removing the limit so psql could display the complete expression, and Oleksandr Shulgin argued that any retained truncation needed to be visible #5 #6 #7.
- Tom Lane supported removing the limit after tracing it to an obsolete fixed-size output buffer, undermining the rationale for retaining it as a display policy #8.
- Tom Lane posted the patch removing truncation and explicitly tracking whether the output string must be freed; he withdrew the memory-leak claim but explained that the existing ownership logic relied on fragile assumptions, and proposed backpatching to PostgreSQL 12 #9.
- Bruce Momjian gave brief assent without reporting code review or testing, and Tom Lane committed the patch unchanged as 314fb9baea, with backpatches through PostgreSQL 12; the thread ended with Tom confirming completion #10 #11.