These are chat archives for dropbox/pyston

4th
Mar 2015
Chris Toshok
@toshok
Mar 04 2015 00:00
#if defined(PYOS_OS2)
Kevin Modzelewski
@kmod
Mar 04 2015 00:16
lol
Chris Toshok
@toshok
Mar 04 2015 00:16
it’s been a while since I saw an OS2 ifdef, heh
Kevin Modzelewski
@kmod
Mar 04 2015 23:01
@tjhance @dagar I think I found the bug that gcc exposed and it's pretty interesting
we currently do scopes[e] = getScopeInfoForNode(e) to analyze a scope in a certain case
and getScopeInfoForNode first checks to see if the node exists in the scopes cache otherwise in computes it
in clang this works as expected, but in gcc it seems to first evaluate scopes[e], which actually inserts a NULL value into the map and then returns a reference
and then getScopeInfoForNode thinks we already calculated the scope info :P
Daniel Agar
@dagar
Mar 04 2015 23:08
weird
I think we're still going to hit the assert in parser
Kevin Modzelewski
@kmod
Mar 04 2015 23:10
yeah
at least hopefully it won't spew those messages about buffer overflow or whatever
(that test is supposed to crash, at least for now)
Daniel Agar
@dagar
Mar 04 2015 23:11
I rebased
the old build is still going though
Kevin Modzelewski
@kmod
Mar 04 2015 23:17
ok, I restarted it
Marius Wachtler
@undingen
Mar 04 2015 23:21
mmh the scopes[e] = getScopeInfoForNode(e) behavior really surprises me
but it is probably a form of a[i] = i++; which is undefined
Kevin Modzelewski
@kmod
Mar 04 2015 23:26
yeah, I guess the order of execution of the lhs and rhs are undefined?