These are chat archives for dropbox/pyston

8th
Oct 2015
An Long
@aisk
Oct 08 2015 02:45
@undingen Cool I'll try it latter, thx ! I tried call freeze on every ast type, but some times got crashed.
Marius Wachtler
@undingen
Oct 08 2015 13:03
@kmod do you have an idea whats the cause of the deopt problem is seeing in https://travis-ci.org/dropbox/pyston/jobs/84262536?
The asserts says that asgn->value != after_expr
I printed out the CFG it is: https://gist.github.com/undingen/79ca5da48c97562c0db2
Marius Wachtler
@undingen
Oct 08 2015 13:08
oh it looks like that it maybe caused by the object cache :-(
Marius Wachtler
@undingen
Oct 08 2015 13:55
the issue is really weird... it has something todo with the object cache but even if I add a line which disables loading any object from the object cache it fails. But if I instead of adding the current_stmt as a symbolic entry emit it directly as a constant number the issue disappears aka the unwinder finds the correct current_stmt...
Marius Wachtler
@undingen
Oct 08 2015 14:05
but now that I know that the issue is probably not directly deopt related I think I will find the issue on my own
An Long
@aisk
Oct 08 2015 14:54
Hi @undingen , I got the same problem when add DEFAULT_CLASS_SIMPLE, there is my code:
static BoxedClass* AST_cls;

class BoxedAST : public Box {
public:
    AST* ast;
    HCAttrs attrs;
    BoxedAST(){};

    DEFAULT_CLASS_SIMPLE(AST_cls);
};
Kevin Modzelewski
@kmod
Oct 08 2015 20:07
Hi @aisk do you have your code up in a branch?
@undingen interesting, haven't seen something like that
hmm I wonder if those values aren't real pointer values 0x7f3920555620 0x7f39205556e0
Sun
@Daetalus
Oct 08 2015 21:58
Does Pyston has plan to use or will generate *.pyc file in future?
Kevin Modzelewski
@kmod
Oct 08 2015 23:23
we do use pyc files
they're not the same as CPython ones though
Sun
@Daetalus
Oct 08 2015 23:24
k
Kevin Modzelewski
@kmod
Oct 08 2015 23:26
parsing large codebases would be quite painful without them :)