116924 - Looks like --display is wrong handled for plugins

Status: VERIFIED FIXED

(Core Graveyard :: Plug-ins, defect)

Description

[Alexei V. Mokeev]

During viewing flash Mozilla crashed, when running with --display option.

Steps to reproduce:

1)set DISPLAY environment variable to something wrong(e.g other machine)
2)Run mozilla with --display <right_display> option
3)Load any page with flash data(required flash plugin)
Note: Mozilla crashed with following messages in console:
About to create new ws_info...
About to create new xtbin of 149 X 20 from 0x8420f08...
About to show xtbin(0x84a7b00)...
completed gtk_widget_show(0x84a7b00)
Nullplugin: plugin received window resize.
Window=(139134180)
W=(149) H=(20)
Gdk-ERROR **: BadWindow (invalid Window parameter)
  serial 8 error_code 3 request_code 1 minor_code 


When using DISPLAY variable,not --display option all works fine.

Comment 1

[Sergi]

Look at gtkxtbin.c method xtbin_new().

When we initialize the Xt toolkit, we create new X connecton with:
    xtdisplay = XtOpenDisplay(app_context, NULL, NULL,
                              "Wrapper", NULL, 0, &mArgc, mArgv);

Second parameter of XtOpenDisplay() is display we should use. In case it's NULL,
XtOpenDisplay() uses DISPLAY environment variable. So, if DISPLAY variable and
--display option parameter differs, we have wrong conditions for plugin window
creating.

I think, every time we'd like to create xtbin, we already have gdk initialized,
with correct display (where --display is more significant than DISPLAY var).
So, we can use gdk_get_display() call. This call returns actual display.

Comment 2

[Sergi]

Attachment: Proposed fix.

Fix for bug, please review.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Sergi, could you attach a unified diff instead of a context diff?  Those are 
much easier to review..

Also, does the xlib or qt code need a similar change?

Comment 4

[av (gone)]

Serge, would you please take a look?

Comment 5

[Sergi]

Neither xlib code nor qt code contains opening of new display.
Only gtk. Do not know why, but initializing of Xt on display was
opened by gdk freezes gdk activity. So in gtk/gdk case we need new
connection to X.

Comment 6

[Sergi]

Attachment: Same patch, unified diff.

Comment 7

[Sergi]

Attachment: Same patch, unified diff.

Comment 8

[av (gone)]

--- Mass reassigning Unix bugs to serge ---

Comment 9

[serge (gone)]

Comment on attachment 62846 [details] [diff] [review]
Same patch, unified diff.

you are right, r=serge

Comment 10

[Peter Lubczynski]

Does this still need super review?

Comment 11

[Sergi]

Yes!

Comment 12

[Patrick C. Beard]

Comment on attachment 62846 [details] [diff] [review]
Same patch, unified diff.

sr=beard

Comment 13

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

a=roc+moz for 0.9.9

Comment 14

[Boris Zbarsky [:bzbarsky]]

checked in on the trunk (pre-branch).

Comment 15

[shrirang khanzode]

v