On Mon, Oct 27, 2025 at 09:17:03PM +0100, Jim Jones wrote: > On 27/10/2025 17:23, Nathan Bossart wrote: >> This was briefly mentioned upthread, but I'm a little concerned that this >> doesn't respond to commands like SET TRANSACTION READ ONLY. I wonder if we >> should mark transaction_read_only as GUC_REPORT and use that instead. FWIW >> I see that we marked search_path as GUC_REPORT somewhat recently (see >> commit 28a1121). > > You're right, it doesn't. I like the idea, but I'm not sure how to > integrate a transaction-scoped variable into this feature. Would that > mean we also need to change the reset mechanism for GUC_REPORT variables > when the transaction ends?
Hm. You're right, that seems to have problems (I'm curious about the use of stmt->is_local in SetPGVariable() for SET TRANSACTION statements). I also see some past discussions in this area [0] [1] [2]. [0] https://postgr.es/m/flat/3a40f835-116d-0f95-aede-d5236337bbf0%402ndquadrant.com [1] https://postgr.es/m/flat/CA%2BTgmoZsHrHeqh5dYpoH%2BWW5EmT-egMGuyrLTsjKz80WajT4tg%40mail.gmail.com [2] https://postgr.es/m/flat/CAFj8pRBFU-WzzQhNrwRHn67N0Ug8a9-0-9BOo69PPtcHiBDQMA%40mail.gmail.com -- nathan
