The progress properties should have a consistent behavior and allow to distinguish the case where the size of the file being downloaded is zero bytes from the case where it is unknown. Add tests so that this behavior is not regressed in the jsdownloads API.

This adds regression tests for the zero-byte cases and more general progress tests, but also refactors the existing tests which use interruptible requests, that previously had issues with the handler being unregistered and an error page being returned.

Comment on attachment 717527 [details] [diff] [review] The patch Review of attachment 717527 [details] [diff] [review]: ----------------------------------------------------------------- ::: toolkit/components/jsdownloads/src/DownloadCore.jsm @@ +604,5 @@ > + > + // Ensure we report the value of "Content-Length", if available, even > + // if the download doesn't generate any progress events later. > + if (aRequest instanceof Ci.nsIChannel && > + aRequest.contentLength >= 0) { may not aRequest need an explicit QueryInterface to nsIChannel before accessing contentLength? ::: toolkit/components/jsdownloads/test/unit/head.js @@ +258,5 @@ > //////////////////////////////////////////////////////////////////////////////// > //// Initialization functions common to all tests > > let gHttpServer; > +let gEmptyRequestReceived; this thing looks hackish, could you make it more like deferNextResponse?

(In reply to Marco Bonardo [:mak] (Away Mar 1) from comment #2) > > + // Ensure we report the value of "Content-Length", if available, even > > + // if the download doesn't generate any progress events later. > > + if (aRequest instanceof Ci.nsIChannel && > > + aRequest.contentLength >= 0) { > > may not aRequest need an explicit QueryInterface to nsIChannel before > accessing contentLength? "instanceof" takes care of that. > > +let gEmptyRequestReceived; > > this thing looks hackish, could you make it more like deferNextResponse? I'll try!

(In reply to Paolo Amadini [:paolo] from comment #3) > "instanceof" takes care of that. oops, you're right