| *** clarkb is now known as Guest298 | 01:19 | |
| *** Guest298 is now known as clarkb | 01:20 | |
| timburke | ok, here's my minimal repro so far: https://paste.opendev.org/show/bzcT5rtgKIa7SqfltyGV/ | 01:40 |
|---|---|---|
| timburke | there's still a whole bunch of swift stuff that gets imported, but i tried to minimize a bunch of the magic and get it down to just one line that makes the difference between a segfault and no segfault | 01:40 |
| timburke | gdb backtrace looks like https://paste.opendev.org/show/bdTlNAalM7CXARfga89x/ | 01:41 |
| timburke | but if you comment out the `self.x` assignment, everything's fine. if you change the assignment to *just* 'str', it's fine (presumably because 'str' is a short, commonly used string, so we just increment a refcount somewhere instead of allocating a new string) | 01:43 |
| timburke | i still haven't gotten to a non-swift repro yet, though | 01:44 |
| timburke | zigo, fyi ^^^ | 01:44 |
| timburke | in playing around with `-X dev`, i started seeing backtraces like https://paste.opendev.org/show/bBFhJ9arhGzz5Ogv7u09/ -- which make me want to take a closer look at https://github.com/python/cpython/pull/28802 tomorrow... | 06:59 |
| opendevreview | Matthew Oliver proposed openstack/swift master: WIP sharder: update own_sr stats explicitly https://review.opendev.org/c/openstack/swift/+/852283 | 07:40 |
| opendevreview | Matthew Oliver proposed openstack/swift master: Sharding: No stat updates before CLEAVED state https://review.opendev.org/c/openstack/swift/+/837811 | 07:40 |
| opendevreview | Matthew Oliver proposed openstack/swift master: WIP: sharding: Block moving to CLEAVED if cleaved rows < expected https://review.opendev.org/c/openstack/swift/+/843973 | 07:40 |
| zigo | timburke: Are you saying it's an issue in the 3.11 interpreter itself then?!? | 09:08 |
| opendevreview | Tim Burke proposed openstack/swift master: Malformed CompleteMultipartUpload request should 400 https://review.opendev.org/c/openstack/swift/+/865348 | 17:12 |
| timburke | zigo, that's my concern, yeah. there's a chance it's got something to do with greenlet or something, but our db code doesn't touch c-level mallocs directly, only by way of stdlib | 17:16 |
| zigo | timburke: Should I try to rebuild greenlet and see if unit tests are failing? | 17:16 |
| timburke | looks like i might have a workaround -- declaring slots on the cursor seems like it might help... lemme try this out... | 17:17 |
| timburke | unfortunately, trying to bisect cpython from 3.10.0 to 3.11.0 got hairy because of dependencies wanting to take advantage of new hotness from 3.11 that isn't always there while i'm bisecting :-/ | 17:18 |
| opendevreview | Tim Burke proposed openstack/swift master: db: Declare slots for GreenDBConnection/GreenDBCursor https://review.opendev.org/c/openstack/swift/+/866051 | 17:44 |
| timburke | *now* i'm getting somewhere! pure stdlib repro: https://paste.opendev.org/show/bC71TAbuxeWRwq8LIEAN/ | 19:01 |
| opendevreview | Tim Burke proposed openstack/swift master: Fix DB tests on py311 https://review.opendev.org/c/openstack/swift/+/866051 | 19:02 |
| timburke | so the good news is, cpython's main branch has this fixed (by https://github.com/python/cpython/commit/de388c0a7b71c094d36ce40fecef87bdbb8a87d3) -- the 3.11 branch doesn't (yet?) have a fix, though | 20:25 |
| timburke | zigo, https://github.com/python/cpython/issues/99886 | 21:09 |
| opendevreview | Tim Burke proposed openstack/swift master: Add swift-ring-packer utility https://review.opendev.org/c/openstack/swift/+/866082 | 22:47 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!