83355 - [xlib] Xlib port doesn't use MIT-SHM extension

Status: RESOLVED WONTFIX

(Core :: XUL, defect)

Description

[pocemit]

XlibRGB does not use SHM.
gfx/src/xlib doesn't use SHM either.

Making some patches later today.

Comment 1

[pocemit]

Attachment: patch adds MIT-SHM support to xlibrgb with fallback to standard X routines

Comment 2

[pocemit]

This makes xlibrgb use MIT-SHM extension if available.
This is a "add only" patch, no code has been changed or removed.
The entire thing is under #ifdef MOZ_USE_SHM.
It has been tested on little endian hardware, though if I followed the logic
correctly it should work on either endian.

Reviews are welcome.

Comment 3

[Roland Mainz]

The code still relies on XFlush() to be sure that the image has been rendered...
the correct way should be to listen to XShmCompletionEvent events...

Misc:
- Setting dependicy to bug 57820 which plays around a little bit with xlibrgb.c
- Added this bug to the Xlib-toolkit tracking bug (bug 79119)...
- CC:'ing Xlib-toolkit folks

Comment 4

[pocemit]

hi gisburn :)
I am going on vacation
and you love MIT-SHM :)

Comment 5

[Roland Mainz]

Yes, I do... :-)

Accepting...

Comment 6

[Roland Mainz]

Setting target milestone to 0.9.4(=soon :-)

Comment 7

[Roland Mainz]

imoT:
AFAIK you have a nice patch which kills XCreateGC() in xlibrgb.c ... wanna
attach it here, please ? I'd like to include that in this patch, please...

Comment 8

[Tomi Leppikangas]

Attachment: patch that gets ride of "own_gc" in xlibrgb.c

Comment 9

[Roland Mainz]

Does that patch really work with 1bit deep StaticGray visuals (e.g. WeiredX JAVA
server or WiredX (http://www.wiredx.com/) JAVA Xserver) ?

Comment 10

[Roland Mainz]

Sorry... that is http://www.wiredx.net/ ... :-)

Comment 11

[John Morrison]

May God have mercy on us all. The 212 bug spam-o-rama is Now!

Comment 12

[Asa Dotzler [:asa]]

0.9.4 is out the door.

Comment 13

[Asa Dotzler [:asa]]

0.9.5 is out the door. bumping TM up by one.

Comment 14

[Jonadab the Unsightly One (Nathan Eady)]

Too late for 0.9.6, this needs retargeting.

Comment 15

[pocemit]

cmon, roland, this has been sitting here for months.
can't be that hard to forward-port this patch into current tree

Comment 16

[Roland Mainz]

I am on it... right now...

Comment 17

[Roland Mainz]

Adding testcase...

Comment 18

[Roland Mainz]

pocemit:
The final patch is in bug 112635, incl. support for MIT-SHM. Wanna r= it, please
?

Comment 19

[Roland Mainz]

Main patch "in" as part of patch in bug 112635. MIT-SHM support is currently
|#ifdef'ed| out as it causes color issues on LSB machines...

I'll take a look at that issue in 0.9.9 ...

Comment 20

[Mats Palmgren (inactive)]

Xlib port has been removed from trunk (bug 326152)

-> WONTFIX