vakaobr commented on issue #2588: URL: https://github.com/apache/age/issues/2588#issuecomment-6078761564
@jrgemignani @crdv7 apologies for the direct ping, and please redirect us if there is a better owner. Tagging you two specifically because of what you have each touched rather than at random, and because we think we may have found a candidate commit. **Is #2517 a candidate for this, and is it heading to PG17?** `Bound per-row memory in writable clauses` (#2517, merged to `master` 2026-09-14) sits in exactly the path where this fault fires: the writable clause of an edge `CREATE`. Our symptom is a label name resolved from memory that appears to be adjacent to the agtype payload of the row being written, which reads like a per-row allocation lifetime problem. As far as we can tell it is not in the line we run: ``` we run AGE 1.7.0, PG17/v1.7.0-rc0 PG17 branch head 2026-08-26 (#2517 merged 2026-09-14, base master) PG17 cypher_create.c } last changed 2026-01-20 (#2309) PG17 cypher_utils.c } ``` There is also no 1.8.0 on the PG17 line, so we cannot get this by upgrading within PG17. Two questions, and we are happy to do the work either way: 1. Does #2517 look like it could explain this, or is it unrelated to label-name resolution? 2. If it is a plausible fit, is a PG17 backport planned? If not, **we will cherry-pick it onto `PG17` locally, build, and run it against the live workload that reproduces this at about one occurrence per 20 documents.** That would give you a clean signal on whether it is the same bug, at no cost to you. Tell us if you would rather we test something else instead. @jrgemignani, the label-cache angle is why you are tagged: `ag_cache.c` on PG17 last changed in your `Fix issue 2245 - Creating more than 41 vlabels causes crash`, and our label catalog stays correct throughout while the *name* comes back wrong, which is what made us suspect the cache rather than the catalog. Also, if one of you has triage rights: could the `bug` label be added? We cannot label on this repo and the issue is currently unlabelled. Full diagnostics, including the `jit = off` result and the `pg_stat_activity` and memory-context collection we have armed, are in the **Update 2026-10-09** section of the issue body. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
