Re: libseed-list importer GC bug (was Re: js extensions for native modules)

[Jonatan Liljedahl <lijon kymatica com>]

Ok, thanks!

/Jonatan

Alan Knowles wrote:
> Commited in, I changed it to  JSValueProtect, to get rid of the compiler warnings.
>
> Regards
> Alan
>
>  --- On 29/Jun/2010, Jonatan Liljedahl wrote: 
> > This change exposed another bug: sometimes the object returned from 
> > imports.sandbox was screwed up.
> >
> > The reason this happens is that the GC won't find the module in our 
> > cached file_imports hash table, thus allowing the GC to collect it and 
> > we'll get a wild pointer in the file_imports cache.
> >
> > And the reason this never happened before is that it didn't reuse the 
> > cached module object but always made a new one for each reference of 
> > imports.sandbox.
> >
> > Solution: protect the module object after adding it to the file_imports 
> > cache. Updated patch attached.
> >
> > I don't know if this should be done for js-modules or gir-modules too?
> >
> > PS. With this patch the "workaround WebKit GC bug" (which actually was a 
> > seed bug) in seed-cairo.c shouldn't be needed anymore.
> >
> > /Jonatan