Hi, On 17/02/23 at 20:17 +0100, Timo Röhling wrote: > Control: tags -1 + moreinfo > > Hi Lucas, > > this seems to be bug #1030493 again. > > On Fri, 17 Feb 2023 07:50:37 +0100 Lucas Nussbaum <[email protected]> wrote: > > > E - '')] > > > E ? ^ > > > E > E + ''), > > > E ? ^ > > > E > E + (_INOTIFY_EVENT(wd=1, mask=1073742336, > > cookie=0, len=16), > > > E + ['IN_DELETE', 'IN_ISDIR'], > > > E + '/tmp/tmpmmbhfdhl', > > > E + 'new_folder')] > > Apparently, the returned order of INOTIFY_EVENTS is different with > your builds than it is on the buildds, my machine, and Bastian's > machine. Neither Bastian nor I could reproduce the bug. Initially, I > assumed this is some sort of flakiness and added test retries, but > it looks like your build system always returns events in a > different order (or at least often enough so two retries are not > sufficient). > > I suspect that you are using a different filesystem, which processes > changes differently and thus makes the inotify API return events in > a different order, but I don't know for sure.
I'm using ext4, so this is unlikely. > Can you help us track down the underlying cause, so we can reproduce > the failure and then think of an appropriate way to work around it? Sure. I took a look, and it seems that the different behaviour is caused by me running an old stable kernel, rather than the latest one. With: Linux 5.10.0-9-cloud-amd64 #1 SMP Debian 5.10.70-1 (2021-09-30) amd64 (x86_64) it fails consistently. With: Linux 5.10.0-21-cloud-amd64 #1 SMP Debian 5.10.162-1 (2023-01-21) amd64 (x86_64) It works consistently. Just upgrading the kernel makes the problem disappear for me. So you should be able to reproduce the problem by downgrading the kernel (using snapshot.d.o). I had not updated the image (AWS AMI) I am using for the rebuilds for a long time. Sorry about that. Lucas
