Less gross code, less error paths to worry about.

Comment on attachment 651897 [details] [diff] [review] convert it Does my definition of JSHolder in xpcprivate.h look reasonable, Justin? I'm making a hash table from void* to nsScriptObjectTracer*. I'm reasonably confident about the rest, but these key types weird me out a bit.

Not an nsDataHashtable?

(In reply to Andrew McCreight [:mccr8] from comment #2) > Comment on attachment 651897 [details] [diff] [review] > convert it > > Does my definition of JSHolder in xpcprivate.h look reasonable, Justin? I think Ms2ger is right, nsDataHashtable would make more sense. Although this is not bad at all.

Ah, nice, I didn't notice that! Thanks, I'll switch it over.

Comment on attachment 652898 [details] [diff] [review] JSDHash to nsDataHashtable The documentation seems to imply that we'd want an nsClassHashtable here, since nsScriptObjectTracer isn't a primitive type. Is that what we want here? I have no idea. This looks reasonable, but I'm no expert in our hash tables, nor have I seen this cycle collection code before. But r=bholley if you're comfortable with that.

nsClassHashtable just looks like a type that does nsRefPtr on its values, which isn't what we want here. I'll check that this works in a try run.

Pushed with a slightly inspecific commit message. https://hg.mozilla.org/integration/mozilla-inbound/rev/c0d1f69cc901 try run: https://tbpl.mozilla.org/?tree=Try&rev=6763c07cc2be (the permaorange there is something that has since been disabled, see bug 681138)