2013/3/22 Rik Cabanier <cabanier@gmail.com> > > > On Fri, Mar 22, 2013 at 9:51 AM, Benoit Jacob <jacob.benoit.1@gmail.com>wrote: > >> >> >> 2013/3/22 Rik Cabanier <cabanier@gmail.com> >> >>> >>> >>> On Fri, Mar 22, 2013 at 1:48 AM, Robert O'Callahan <robert@ocallahan.org >>> > wrote: >>> >>>> On Fri, Mar 22, 2013 at 7:48 AM, Rik Cabanier <cabanier@gmail.com>wrote: >>>> >>>>> On Thu, Mar 21, 2013 at 11:38 AM, Tab Atkins Jr. < >>>>> jackalmage@gmail.com> wrote: >>>>> >>>>>> Only if you actually throw it. It's not eagerly created every call. >>>>>> >>>>> >>>>> I worked a little with the mozilla code base and it seems that they >>>>> always create the object. >>>>> It's probably not a huge amount of overhead since it's not dynamically >>>>> allocated. >>>>> >>>> >>>> We certainly don't create a JS exception object if an exception is not >>>> actually thrown. >>>> >>> >>> If a function is marked as throwing in the IDL, doesn't it get an >>> exception object as a parameter on the C++ side? >>> >> >> No, in Mozilla's case it gets a so-called ErrorResult object, >> >> >> http://dxr.mozilla.org/mozilla-central/dom/bindings/ErrorResult.h?from=ErrorResult#l34 >> >> C++ can then decide to call its Throw() method to actually produce an >> exception. > > > Yes, that's the one I was thinking of. > > Well, the argument was over whether that would be expensive. It's not (C++ object created on the stack, with its constructor just setting one integer member to a constant -- so a few cycles at most). Benoit
Received on Friday, 22 March 2013 16:59:51 UTC