657550 - jslinux broken in Nightly

Status: RESOLVED FIXED

(Tech Evangelism Graveyard :: English Other, defect)

Description

[Jeff Muizelaar [:jrmuizel]]

http://bellard.org/jslinux/ works in Aurora but not in the nightly from 2011-05-16.

It gives a "invalid arguments" error in the console.

Comment 1

[Jan de Mooij [:jandem]]

works: 2011-05-11 TM - http://hg.mozilla.org/tracemonkey/rev/b83c8374358a
fails: 2011-05-12 TM - http://hg.mozilla.org/tracemonkey/rev/5b479a987cda

Regression range:

http://hg.mozilla.org/tracemonkey/pushloghtml?fromchange=b83c8374358a&tochange=5b479a987cda

Comment 2

[Jan de Mooij [:jandem]]

Bug 649133 removed mozResponseArrayBuffer. jslinux uses response/responseType like this:
--
Le = new XMLHttpRequest();
Le.open('GET', Ke, false);
if (Le.hasOwnProperty('responseType')) {
    Le.responseType = 'arraybuffer';
} else {
    Le.overrideMimeType('text/plain; charset=x-user-defined');
}
Le.send(null);
if (Le.status != 200 && Le.status != 0) {
    throw "Error while loading " + Ke;
}
if (Le.mozResponseArrayBuffer) {
    Me = Le.mozResponseArrayBuffer;
} else if (Le.hasOwnProperty('responseType')) {
    Me = Le.response;
} else {
    Me = Le.responseText;
}
fd = Me.byteLength;
Ne = new Uint8Array(Me, 0, fd);
for (i = 0; i < fd; i++) {
    this.st8_phys(ha + i, Ne[i]);
}
return fd;
--

Comment 3

[Jan de Mooij [:jandem]]

javascript:alert(new XMLHttpRequest().hasOwnProperty("responseType"));

This alerts |true| in Chrome, |false| in Firefox 6.

Comment 4

[Masatoshi Kimura [:emk]]

> new XMLHttpRequest().hasOwnProperty("responseType")
This test doesn't look correct to me. It should be
> XMLHttpRequest.prototype.hasOwnProperty("responseType")
per WebIDL <http://dev.w3.org/2006/webapi/WebIDL/>.
Unfortunately the latter test doesn't work on WebKit because WebKit doesn't comply the WebIDL spec...

Comment 5

[Boris Zbarsky [:bzbarsky]]

There are actually a number of sites now broken that were testing for mozResponseArrayBuffer....  I hope unprefixing response/responseType will fix most of them, but for jslinux we should just get the author to fix the code per comment 4.

Comment 6

[Boris Zbarsky [:bzbarsky]]

But maybe we should keep mozResponseArrayBuffer for now...

Comment 7

[Jeff Muizelaar [:jrmuizel]]

(In reply to comment #5)
> for jslinux we should just get the author to fix the
> code per comment 4.

I've emailed Fabrice about this.

Comment 8

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

i guess I would be fine with bringing mozResponseArrayBuffer back for a release if we also put a warning in the error console when it's used.

However I wonder how much trouble this will be in practice since it's a very new property and it sites which are using it should still be maintained. And there's always the risk that keeping it longer will just mean that more sites come to depend on it.

Comment 9

[Boris Zbarsky [:bzbarsky]]

Yeah, fair.  Let's not worry about it for now, then; we can bring it back in aurora or beta if it becomes an issue.

The page has been changed to use 'in' instead of hasOwnProperty.  Now we just need to remove the prefixing on reponseType/response.  ;)

Comment 11

[Jeff Muizelaar [:jrmuizel]]

jslinux now works in the nightly

Comment 12

[Masatoshi Kimura [:emk]]

"'mozResponseType' in" test is no longer required because responseType/response properties have been unprefixed. No versions of Firefox will be shipped with the prefixed mozResponseType/mozResponse.