cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs: https://quodlibet.duckdns.org/irc/pypy/latest.log.html#irc-end ) | use cffi for calling C | if a pep adds a mere 25-30 [C-API] functions or so, it's a drop in the ocean (cough) - Armin
<cfbolz>
mattip: #3351 is probably a cpython bug too, if you get creative enough with objects with __del__s that you pass to the Connection, creating a cycle somehow. but that's not our immediate problem
<mattip>
the reproducer is pretty simple, I don't think we should fail there
<cfbolz>
yeah
<cfbolz>
I understand the bug now, I think
<cfbolz>
not quite clear how it should be fixed
<cfbolz>
I can probably make a smaller reproducer
<mattip>
isn't it a failure of the lock?
<mattip>
logic
forgottenone has joined #pypy
<cfbolz>
mattip: yeah, the logic is broken
<cfbolz>
mattip: ok, got a simpler case at alest
<cfbolz>
least
lritter has quit [Ping timeout: 260 seconds]
<cfbolz>
in other news: tornado looks quite good now
<cfbolz>
the sqlite3 stuff is a pile of hacks :-(
<cfbolz>
mattip: I have a plan for a fix though. Probably solves both problems
<mattip>
cool on both counts
dmalcolm__ has quit [Remote host closed the connection]
<cfbolz>
antocuni: autogenerated code by a templating engine, in this case
<antocuni>
nice
<cfbolz>
antocuni: but i wonder whether eg translating has also stuff like that
<antocuni>
in my experience, I had similar problems in the past. Often they were caused because the unoptimized trace was very long, but after optimizations it would become very short
<antocuni>
but if you hit the limit, you are screwed
<antocuni>
e.g., capnpy is designed in such a way that most operations for accessing fields are optimized away
<cfbolz>
yip
<cfbolz>
ah, there are probalby more problems anyway
<mattip>
I just watched a presentation of papyri,
<mattip>
which takes your sphinx doc source (like numpy, scipy, pandas)
<mattip>
and renders it into an intermediate representation, that you ship
<mattip>
then the end-user can render it
<mattip>
so, bottom-line, maybe the sphinx benchmark is not needed ;)