01:19 <mattip> hexology: can you use homebrew? I think the new release should be available there soon

01:19 <mattip> https://github.com/Homebrew/homebrew-core/pull/87517

01:20 <hexology> nice, i do plan to use brew when the release is up

01:20 <hexology> it's also missing from pyenv, not sure if anyone is working on that

01:45 <mattip> pyenv PR was merged 11 hours ago https://github.com/pyenv/pyenv/pull/2111

01:46 <mattip> but I am not sure that installs via homebrew on macOS

07:17 <hexology> mattip: it's the opposite, now the homebrew package for pyenv needs to be updated to the latest version

07:20 <antocuni> ctismer: there is sys.pypy_translation_info['transltion.lldebug'], but depending on how you compile pypy, it might contain the wrong value

07:21 <antocuni> i.e., I think that that value is true only if you translated it using "rpython --lldebug", but not if e.g. you translate normally and then run "make lldebug" manually

07:49 <ctismer> antocuni: OK, I just need the first option to skip code that breaks when lldebug0? is selected.

07:50 <ctismer> antocuni: Thanks a lot, you made my morning :)

07:50 <antocuni> you're welcome :)

09:42 <cfbolz> olliemath: very nice!

10:00 <Olliemath> cfbolz thanks - makes my editor much easier to use so hopefully handy for others :)

10:00 <cfbolz> olliemath: heh, ye

10:00 <cfbolz> s

14:04 <cfbolz> mattip: I moved it to a branch

14:05 <cfbolz> seems more complicated

14:08 <cfbolz> and I maybe found a possible reason for why coverage is so massively much slower on pypy

14:12 <cfbolz> ah no, seems cpython has the same slow code

14:15 <cfbolz> but I probably found the bug

15:25 <mattip> fwiw, I set up a simple module with two functions and a test function, and used pytest-cov to test it

15:25 <mattip> the time to run it under pypy was the same as the time to run it under cpython

15:26 <mattip> so the jury is still out on whether ci with a pure-python library and coverage is really 3x or 5x slower on pypy

16:02 <cfbolz> mattip: ok

16:07 <LarstiQ> cfbolz: now I'm curious what the bug is :)

16:07 <cfbolz> LarstiQ: the logic when to emit a line event was wrong under some circumstances:

16:08 <cfbolz> when executing a backward jump into the middle of a line (like with a for loop)

16:08 <cfbolz> then the target line was not traced

16:08 <cfbolz> there was some extra logic in for loops to compensate

16:08 <cfbolz> but the combination of those two pieces of logic produced really weird corner cases

16:08 <cfbolz> now we just follow cpython much more precisely

16:10 <LarstiQ> ah