617525 - MozAfterPaint events are delivered differently for in-process and out-of-process frame scripts

Status: RESOLVED FIXED

(Core :: IPC, defect)

Description

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

I saw tests fail in bug 615386 because the in-process message-manager was delivering MozAfterPaint events differently than out-of-process (out-of-process was actually passing the test, in-process failing it).

Tentatively marking as a blocker of bug 615386, but not a hard blocker.  I can work around this, but I'd rather not (obviously).

Test coming up.

(smaug, please let me know if this is the preferred component for mm bugs.)

Comment 1

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Attachment: Test

STR
 (1) Apply patch, rebuild
 (2) Run
$ firefox -no-remote -chrome chrome://global/content/test-ipcbrowser.xul 2>&1 | grep '\[content\]'
 (3) Edit layout/ipc/test-ipcbrowser.xul so that the <browser> has remote="false"
 (4) Rebuild
 (5) Repeat step (2)

The results I see are

remote="true"
[content] Saw MozAfterPaint for subdocument (data:text/html,<font%20color=red>old%20innermost)
[content] Saw MozAfterPaint for subdocument (http://people.mozilla.com/~cjones/subdoc-inner.html)
[content] Saw MozAfterPaint for document (http://people.mozilla.com/~cjones/subdoc.html)
[content] Saw MozAfterPaint for subdocument (data:text/html,<body%20onload='function%20w(md,%20t)%20{%20%20%20%20t.p(0);%20%20%20%20md.write(t.replacementForMiddleFrame);%20%20%20%20t.p(3);%20%20%20%20t.p(md.documentElement.innerHTML);%20%20%20%20md.close();%20%20%20%20t.p(4);}w(parent.document,%20top);'><font%20color=blue>new%20innermost)
[content] Saw MozAfterPaint for document (http://people.mozilla.com/~cjones/subdoc.html)

remote="false"
[content] Saw MozAfterPaint for document (http://people.mozilla.com/~cjones/subdoc.html)
[content] Saw MozAfterPaint for subdocument (data:text/html,<body%20onload='function%20w(md,%20t)%20{%20%20%20%20t.p(0);%20%20%20%20md.write(t.replacementForMiddleFrame);%20%20%20%20t.p(3);%20%20%20%20t.p(md.documentElement.innerHTML);%20%20%20%20md.close();%20%20%20%20t.p(4);}w(parent.document,%20top);'><font%20color=blue>new%20innermost)

I expect the events delivered to be the same for remote=true and =false.

I can work around this in reftests by attaching my MozAfterPaint listener to the top-level window.

Comment 2

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Also FTR, the test page linked from test-ipcbrowser.xul, http://people.mozilla.com/~cjones/subdoc.html, is just a slightly modified version of layout/reftests/bugs/212563-2.html.

Comment 3

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Another data point:

With remote="false" and this change to test-ipcbrowser-content.js,

//addEventListener("MozAfterPaint", logMozAfterPaint, false);
const CI = Components.interfaces;
var gTopWindow = docShell.QueryInterface(CI.nsIDocShellTreeItem)
                 .parent.QueryInterface(CI.nsIInterfaceRequestor)
                 .getInterface(CI.nsIDOMWindow);
gTopWindow.addEventListener("MozAfterPaint", logMozAfterPaint, false);


I see the event log

[content] Saw MozAfterPaint for subdocument (undefined)
[content] Saw MozAfterPaint for document (http://people.mozilla.com/~cjones/subdoc.html)
[content] Saw MozAfterPaint for subdocument (data:text/html,<body%20onload='function%20w(md,%20t)%20{%20%20%20%20t.p(0);%20%20%20%20md.write(t.replacementForMiddleFrame);%20%20%20%20t.p(3);%20%20%20%20t.p(md.documentElement.innerHTML);%20%20%20%20md.close();%20%20%20%20t.p(4);}w(parent.document,%20top);'><font%20color=blue>new%20innermost)
[content] Saw MozAfterPaint for document (http://people.mozilla.com/~cjones/subdoc.html)
[content] Saw MozAfterPaint for subdocument (undefined)

which is different than either of the other two.  *scratches head*

Comment 4

[Olli Pettay [:smaug][[e-mail]]]

Attachment: wip

This is still untested.

Comment 5

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

For remote=false, this patch gives a fourth, different result.

[content] Saw MozAfterPaint for document (http://people.mozilla.com/~cjones/subdoc.html)
[content] Saw MozAfterPaint for subdocument (data:text/html,<body%20onload='function%20w(md,%20t)%20{%20%20%20%20t.p(0);%20%20%20%20md.write(t.replacementForMiddleFrame);%20%20%20%20t.p(3);%20%20%20%20t.p(md.documentElement.innerHTML);%20%20%20%20md.close();%20%20%20%20t.p(4);}w(parent.document,%20top);'><font%20color=blue>new%20innermost)
[content] Saw MozAfterPaint for document (http://people.mozilla.com/~cjones/subdoc.html)

This includes the MozAfterPaints important to this reftest (the ones to "document"), so ought to work.  Will give a full rundown when other work settles.

Comment 6

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

I'm really not the right guy to review this change, but I can feedback+ that it seems to fix the reftest issue I saw.

smaug, would roc perhaps be better?

Comment 7

[Robert O'Callahan (:roc) (email my personal email if necessary)]

http://hg.mozilla.org/mozilla-central/rev/b4a5b419d513