[13:17:43] <mszabo>	 so, the good news is that PHP 8.0 randomly segfaults with MediaWiki on production
[13:25:59] <Krinkle>	 mszabo: :|
[13:36:24] <mszabo>	 I'm currently trying to get a coredump, which is sadly a slightly more involved process with kubernetes
[13:40:17] <mszabo>	 this is also 8.0.25 so it should already include Tim's most recent segfault fix
[16:29:30] <mszabo>	 https://gist.github.com/mszabo-wikia/90f304633928eb096d1cf6171f52f0c3
[16:29:37] <mszabo>	 so far that's all I've found
[16:30:03] <mszabo>	 I lack the knowledge to properly interpret it but it seems like heap->free_slot ends up containing invalid pointers somehow
[18:57:28] <mszabo>	 Filed https://github.com/php/php-src/issues/9879, will see where that goes
[19:02:06] * Krinkle subs
[20:05:11] <Reedy>	 lol
[20:05:48] <mszabo>	 Can't have a PHP upgrade without a proper segfault
[20:05:54] <mszabo>	 At least the rate is not atrocious so far
[22:32:17] <mszabo>	 Krinkle: https://grafana.com/blog/2022/11/02/announcing-grafana-phlare-oss-continuous-profiling-database/
[22:45:16] <Krinkle>	 Hmm a Phacility product?
[23:27:38] <Krinkle>	 I'm trying to figure out where it's sense of time comes from in those sample aggregates. It's saying not just % but actual seconds. I'm not seeing anywhere in pprof implementations like for Python where this is derived from other than the "duration", which appears to be how long it samples for. Eg rather than continuously sampling it seems to be modelled around an RPC point that starts it for eg 15 seconds during which every second it 
[23:27:38] <Krinkle>	 takes a sample. But I don't see how you can derive from that how long a function runs
[23:28:02] <Krinkle>	 s/saying/seeing/
[23:29:10] <Krinkle>	 Very interesting though. This is basically Arc Lamp Redis + log + Swift (what we have) as a service with UI rendering client side adhoc from arbitrary ranges of samples. Very nice