These are chat archives for dropbox/pyston

25th
Mar 2015
Travis Hance
@tjhance
Mar 25 2015 00:41
wait what does <slice>=<slice> mean?
Chris Toshok
@toshok
Mar 25 2015 04:07
oh, just something like: a = [1,2,3,4,5]; b = [1,2,3,4] a[1:2] = b[:-1]
Travis Hance
@tjhance
Mar 25 2015 04:07
ooh
Chris Toshok
@toshok
Mar 25 2015 04:08
right now we heap allocate the rhs before processing the lhs
Marius Wachtler
@undingen
Mar 25 2015 09:47
I'm trying to get exception_subclasscheck.py to pass
the second 'strange' subclasscheck <class '__main__.E'> output comes from a PyErr_NormalizeException call when encoutnering the except statement. Is it important to get the behaviour exactly the same as cpython?
Marius Wachtler
@undingen
Mar 25 2015 10:07
We would probably have to call PyErr_NormalizeException in SET_EXC_INFO to get the output exactly the same (atleast for this case..)
Kevin Modzelewski
@kmod
Mar 25 2015 10:54
hmm I would guess it's not too important
I don't think the subclasscheck stuff is used that much
Marius Wachtler
@undingen
Mar 25 2015 13:26
I would like to get the withstatement to work on old style classes
Marius Wachtler
@undingen
Mar 25 2015 13:34
Looks like the ___enter___ wont get found on our old style classes because it is uses a ClsAttibute lookup. How can I fix that?
Kevin Modzelewski
@kmod
Mar 25 2015 21:05
oh man
so it looks like that with statements have special lookup rules
that if it's an instance, they just call getattr() on it
Chris Toshok
@toshok
Mar 25 2015 21:06
you mean like JS’s with statement?
like: with (o) { print(a); } // <- prints o.a
@undingen maybe put that check in objmodel.cpp::getclsattr / getclsattrInternal?
Chris Toshok
@toshok
Mar 25 2015 21:09
heh from that pep:
    Yes, the verb tense is messed up in a few places. We've been
    working on this PEP for over a year now, so things that were
    originally in the future are now in the past
Marius Wachtler
@undingen
Mar 25 2015 21:10
thanks will try to fix it
thanks for the hint
btw: tried the llvm timing output
Marius Wachtler
@undingen
Mar 25 2015 21:21
from >2.5secs inside llvm, spending ~0.7secs on inst selection and dwarf emission and reg alloc does not sound to bad. I'm wodnering where most of the time gets spend...
Kevin Modzelewski
@kmod
Mar 25 2015 21:23
I think perf is able to give the llvm function names
I wonder if MCJIT adds any more time? (I don't think that would show up here)
Marius Wachtler
@undingen
Mar 25 2015 22:48
You are right the MCJIT time does not show up in the report
:-(