590010 - Bug 583857 causes 426 lines of spam at the beginning of non-Windows reftest logs

Status: RESOLVED WORKSFORME

(Core :: Graphics, defect)

Description

[Daniel Holbert [:dholbert]]

Now when I start reftests, I'm greeted with 142 consecutive iterations of this bundle of three lines (426 lines of output in total):
*********
JavaScript strict warning: chrome://reftest/content/reftest.js, line 341: reference to undefined property CC[NS_GFXINFO_CONTRACTID]
REFTEST INFO | Dumping JSON representation of sandbox
REFTEST INFO | {"isDebugBuild":true,"xulRuntime":{"widgetToolkit":"gtk2","OS":"Linux","XPCOMABI":"x86-gcc3"},"d2d":false,"cocoaWidget":false,"gtk2Widget":true,"qtWidget":false,"winWidget":false,"http":{"userAgent":"Mozilla/5.0 (X11; Linux i686; rv:2.0b5pre) Gecko/20100823 Minefield/4.0b5pre","appName":"Mozilla","appVersion":"5.0","vendor":null,"vendorSub":null,"vendorComment":null,"product":"Gecko","productSub":"20100823","productComment":null,"platform":"X11","oscpu":"Linux i686","language":"en-US","misc":"rv:2.0b5pre"},"haveTestPlugin":true,"windowsDefaultTheme":false,"nativeThemePref":true,"prefs":{"__exposedProps__":{"getBoolPref":"r","getIntPref":"r"},"_prefs":{"root":"","PREF_INVALID":0,"PREF_STRING":32,"PREF_INT":64,"PREF_BOOL":128}}}
*********

(note that the last line -- the JSON dump -- is also quite long)

I noticed this on my own linux reftest run today, but I've confirmed that it affects linux tinderboxen (see log linked in URL field), and I'm guessing it affects all non-windows platforms.

This is from a block added to reftest.js in bug 583857, here:
http://hg.mozilla.org/mozilla-central/diff/d3293a155e32/layout/tools/reftest/reftest.js

Comment 1

[Serge Gautherie (:sgautherie)]

(In reply to comment #0)
> JavaScript strict warning: chrome://reftest/content/reftest.js, line 341:
> reference to undefined property CC[NS_GFXINFO_CONTRACTID]

I filed bug 629617.

Comment 3

[Serge Gautherie (:sgautherie)]

Iiuc, BuildConditionSandbox() is called once per manifest.

I wonder whether the sandbox could be built once only then cloned?

Assuming it couldn't, can properties like CC[NS_GFXINFO_CONTRACTID] actually change from one sandbox to another?
If not (hopefully), these property values should be retrieved once at the start only then reused. (At least to speed things up.)

{
   495     gDumpLog("REFTEST INFO | Dumping JSON representation of sandbox \n");
   496     gDumpLog("REFTEST INFO | " + JSON.stringify(sandbox) + " \n");
   497 
   498     return sandbox;
}
Yet, unless all sandboxes are "identical", there's nothing more that can be done about it. Are they all "identical"?

Comment 4

[u279076]

Dan, is this bug still relevant?

Comment 5

[Daniel Holbert [:dholbert]]

Nope, I haven't hit anything like this in quite a while. --> WORKSFORME