Partial or repeated submission of MR review comments is possible, MR state can become anomalous
Everyone can contribute. Help move this issue forward while earning points, leveling up and collecting rewards.
Sometimes, an MR can get into an anomalous state where review comments get duplicated and/or not fully published.
On one previous occasion, on our self-hosted gitlab CE, I was reviewing an MR. "Submit review" gave an error. On reloading, I found that (most of?) of my comments had been published, but the same coments were still listed as "Pending". It was possible to "Submit review" again, which would duplicate the comments again. There was no way to get the thing to a sane state.[
Right now, I have an MR https://gitlab.torproject.org/tpo/core/arti/-/merge_requests/2258 which has been afflicted by a similar bug. Our instance is having overload problems, which probably made some requests fail. The MR is now in a weird state where some but not all of my review comments have been published, some more than once, but all of them remain Pending.
Needless to say, I would have expected gitlab to use database transactions to ensure that situations like this were impossible. I infer that the db changes invoked during MR review submission are not properly enclosed in transactions.
I tried to search for a previous report of this issue but I wasn't able to come up with any search terms (or earch approach) that didn't produce unmanageable volumes of manifestly irrelevant hits. My apologies if this is a duplicate.