Mochabot log - CommonJS IRC channel: #commonjs on irc.freenode.net

2010-01-09:

[6:02] <jmerlin> the problem is using their locker class, :(
[6:14] <jmerlin> i doubt they make much use of it themselves in chrome
[6:46] <jmerlin> weeew maybe that's because the documented usage of Locker is a bit weird, and no example case is shown, reading bug trackers with similar issues, it seems like v8 just "crashes later" if you don't use a locker object correctly (rather than throwing an exception in your face or having an assertion pop up that says: "you can't do that!"), manifesting itself in either crashes or in program-exit error messages
[6:48] <jmerlin> i think if i wrap EVERY little bit of v8 code in locker objects it should fix the issue but the documentation in v8.h does say "any code that accesses or creates new v8 handles" is where a locker object should be used but it doesn't stress that this is necessary even when threading hasn't become an issue, it's not sufficient to protect only code that may execute concurrently (for instance your startup code won't)
[6:48] <jmerlin> that just seems stupid to me, a bad design choice at best
[6:50] <jmerlin> i may just end up switching to another engine that doesn't stand in the way of threading so readily, threading is extremely important when you need asynchronous callbacks into your JS :\
[6:59] <jmerlin> yep, wrapping the initializing code in lockers fixes it
[6:59] <jmerlin> wow that's really bad :|
[7:54] <jmerlin> bad isp is bad
[7:55] <jmerlin> http://code.google.com/p/v8/issues/detail?id=471#c20
[7:55] <jmerlin> looks like that bug is now a "feature", it's not documented woooo
[8:17] <Dantman> Hmm... the svg spec has something interesting... An onactivate event, noted as a preferred event to clicking for best support to other input devices. ^_^ I think I just found a good name for a faux event that'll replicate the behavior of how <a> handles hrefs.
[10:28] <jmerlin> ondras: looks like v8 only does gc when its resource pool gets near its maximum size
[10:28] <jmerlin> or it determines there's too much memory being used by the process
[10:28] <jmerlin> im creating my objects which are 4KB and I have to make nearly 2GB worth before it'll run a GC cycle
[10:29] <jmerlin> i may write a little function that lets me force a full GC sweep
[12:36] <ondras> hmhm
[14:11] <jmerlin> have you tried that?
[16:00] <Dantman> Wow... MongoDB 1.3 has a very interesting new feature, findandmodify
[16:03] <ashb> whats it do?
[16:07] <Dantman> http://www.mongodb.org/display/DOCS/findandmodify+Command
[16:09] <ashb> ah useful

 

 

Logs by date :