On Wed, Jan 09, 2019 at 02:09:47AM +0100, Tim Düsterhus wrote: > Pieter, > > Am 08.01.19 um 23:37 schrieb PiBa-NL: > > Got a coredump of 1.9.0-8223050 today, see below. Would this be 'likely' > > the same one with the 'PRIORITY' that 1.9.1 fixes? > > Without knowing much about the mux code: This is highly unlikely to be > related. In my tests the bug lead to an immediate crash when receiving > the bogusrequest. In your case it appears to me that the crash happened > while sending the response.
For me it looks like this one, first one fixed in 1.9 : af97229 ("BUG/MEDIUM: h2: Don't forget to quit the sending_list if SUB_CALL_UNSUBSCRIBE.") > Generally I believe it's best to send those kind of reports to Willy in > private. If this is a actually something that can be triggered by an > unauthenticated client it does not need to be exposed for the whole > world to see :-) Well, as long as there's no reproducing data, public report often increase the likeliness of the "me too" effect which often helps narrow down an issue. > > Anyhow i updated my system to 2.0-dev0-251a6b7 for the moment, lets see > > if something strange happens again. Might take a few days though, IF it > > still occurs.. I'm confident you will not see this one anymore. Interestingly, 1.9 was much more difficult to develop than 1.8 but looks much less chaotic after the release. I think that the cleaner architecture and the regtests have significantly reduced the opportunities for introducing stupid bugs, this is great. Thanks, Willy