mozilla :: #ionmonkey

17 Jul 2017
10:49jandemevilpie: found the bug causing the test failure, unrelated to your patch. I'll file a bug
14:26bbouvierdjvj-ic, ping
14:53evilpiejandem: ARM build is failing
14:54jandemevilpie: i had to add the argument to the definition in the .cpp file, I fixed it locally
14:54jandembefore i pushed
15:10bbouvierwhy is the InlineFrameIterator stored in the FrameIter and not the JitFrameIterator?
15:10bbouvierjandem, nbp?
15:10bbouvieralso, since JitFrameIterator is in the jit:: namespace, could we get rid of the Jit prefix in the name?
15:11jandembbouvier: because not every JitFrameIterator consumer cares about inline frames
15:11jandemmaybe, but it would be more confusing when code wants to use both FrameIterator and JitFrameIterator
15:13bbouvierjandem, good call for the renaming question
15:14bbouvierjandem, not sure for the first one, though: it seems nextJitFrame() always tries to create an inlineFrameIterator for jit frames
15:14bbouvierwell, only if isIonScripted()
15:14jandembbouvier: but that's FrameIter. There are other places where we use JitFrameIterator and ignore inline frames
15:16bbouvierjandem, oh that's why, ok
15:19nbpbbouvier: you mean in terms of file names?
15:20nbpbbouvier: oh Stack.h impl, yes Jan answer.
15:20nbpbbouvier: the GC does not care about iterating the inline Frames, it only cares about marking the Jit frames.
15:20nbpbbouvier: so we do not want to iterate over the inline frames when we are traversing the stack for the GC.
15:21bbouviernbp, makes sense too
15:21nbpbbouvier: on the other hand arguments.caller, or error reporting wants to report any stack frame, including the inline frames.
15:26bbouvierso would it make sense to hoist the ionInlineFrameNo (unsigned) from Data to the FrameIter itself?
15:35bbouvierapparently not (one of the FrameIter ctors uses a Data's ionInlineFrameNo)
15:35jandembbouvier: Data is used for the debugger, it sometimes want to store a heap allocated frame iterator
15:36jandemstoring an inline frame iterator would be unsafe
16:20nbpanybody for a quick review: ?
16:20nbpbbouvier: maybe ^
16:22nbpbbouvier: these are reported by OOMTest function from ion/bug1284485.js.
16:23bbouvierjust adding memory reporting, right?
16:23nbpbbouvier: yes
16:23bbouviernbp, ok r=me then
16:23nbpbbouvier: thanks
18 Jul 2017
No messages
Last message: 9 days and 6 hours ago