Server: Ubuntu 14.04 amd64, xpra version 0.15.0 13-Feb-2015 08:00 Client: MS Windows 7 32bit, xpra version 0.15.0 r8661
I forgot this: Maximize Firefox first, then minimize...
Once again, I can't reproduce this one. Can you please post "xpra info" when this happens? Does it help if you select "raise windows" from the tray? Does it help if you toggle opengl on or off from the tray?
Once again, I can't reproduce this one.
I will reproduce again, maybe this helps:
Can you please post "xpra info" when this happens?
xpra info: attachment/ticket/809/lost_image_log_150224.txt
Does it help if you select "raise windows" from the tray?
No.
Does it help if you toggle opengl on or off from the tray?
Both server and client don't support opengl.
John1221: nothing there looks suspicious :( I'll have to test with the exact same versions and OS..
Until then, does it make any difference if you use a different encoding? (ie: rgb)
antoine: don't have any difference if I use others encoding ( png, rgb, jpeg,...) After lost image, If I make the Firefox from maximize to normal state ( press restore button ), then the window images return to normal.
OK, thanks.
The fact that this also affecting firefox, and also related to maximized state makes me think that this could be the same bug as #790: maybe the application thinks it isn't mapped when it is, so the contents are empty.
Update: don't need to minimize and restore INSTANTLY, just need to minimize and then, restore a window.
Is this fixed as per #790?
@antoine: Tested with Firefox and xterm with server ( trusty, xpra 0.15 r8802), client( MS Windows 7, xpra 0.15 r8826) Firefox is fixed, but xterm still occurs ( lost all text in xterm)
Great for Firefox. In a way, that's good, because xterm if far is easier to debug and test.
Sorry to ask again but can you give me exact steps to reproduce with just an xterm? (as simple as you can make them)
And if the bug is to do with maximize, please add to #790 instead of this one.
@antoine: Reproduce:
Hope this helps :)
Sorry for taking so long to test this very obvious bug. The reproduction steps were very easy.
This is fixed in r9009. I think this was caused by the work in #775. We now synchronize the "iconified" flag too when we detect a change, and not just when the window gets unmapped (hopefully this won't cause regressions).
Feel free to test the latest server builds (the client is unchanged) and report back.
I couldn't find the beta build r9009. So I tested the latest server(TRUSTY) builds on 10 April, 2015(http://xpra.org/beta/trusty/main/binary-amd64/xpra_0.15.0-1_amd64.deb), the latest client(WINDOWS) builds (r8987) at the same time. Xterm still occurs this bug.
@John1221: the fix is in r9009, so r8987 is too old. I have just posted some newer builds (trusty only).
@antoine: It works in the server builds r9016. Thank you very much.
This bug is reappeared.
Server: trusty amd64 with xpra 0.15.6-r10632 Client: MS Windows 7 64bit with xpra 0.15.6-r10632
As per ticket:790#comment:31, this is apparently fixed in trunk.
If we can identify the fix in trunk, we can then backport it.. The steps in this ticket seem simple enough, so I will give it a go.
Seems to be a client issue, seems broken with 0.15.7 client but ok with latest 0.16.0 beta clients, no matter which server version I connect to. Bisecting the client:
So the problem is r10790, but it looks right and works fine in 0.16.0
And since I've spent far too much time on this already... So I am keeping this ticket open, but will close when we stop supporting 0.15 (hopefully soon).
Sorry, 0.15.x is going to be retired soon and I don't have time for this: the solution will be to upgrade clients to 0.16. Feel free to re-open if you can reproduce with a version 0.16 onwards.
I've upgraded xpra server to 0.15.8-3, xpra client to 0.16.0-r11206. This bug can reproduce like comment:17 I can also reproduce with xpra client version 0.16.0 r10853 and 0.16.0 r11176. Haven't test with some versions between r10853 and r11176 yet.
Interesting: I tested with 0.16 latest and could not reproduce this particular problem at all (even though I did see it last time in comment:20). I have tested with many servers:
with Firefox 42, XP client.
I did see something fishy going on with the window style: the print dialog does not have a minimize or maximize option when it first shows up but as soon as the fixup_window_style() hook runs, it does. This is fixed in r11284. You can still use the taskbar to minimize things and try to trigger this bug.
@John1221: can you reproduce this bug with the latest win32 beta?
I also tested with server(Ubuntu 14.04) and Firefox 42, but Windows 7 64bit client
It still appear with the latest r11304 win32 beta
I have now tested again with:
Still cannot reproduce. I cannot follow the instructions from comment:17 exactly, as the minimize button is now missing from the print dialog. I have tried using the taskbar and using the main window to minimize all firefox windows.
@John1221: can you give me more specific steps? Can you try a few more things to narrow it down? (opengl on/off, other applications? etc..)
I'm sorry. Has a mistake at here "Press minimize button from the Print dialog", it should be "Press Firefox windows button from the taskbar to minimize all Firefox windows". Maybe the Print dialog has a minimize button, but from previous version of Firefox.
I tried narrowing opengl off, and images were still good. I think my problem is opengl. From "Session Info", I can see the server doesn't support (or not yet) opengl(n/a) but the client does.
Maybe the Print dialog has a minimize button, but from previous version of Firefox.
As per comment:23, this was a bug in xpra - now fixed.
I think my problem is opengl. From "Session Info", I can see the server doesn't support (or not yet) opengl(n/a) but the client does.
No, this is unrelated: the server never shows its opengl information.
I can reproduce this bug with 0.15.x clients, but not with any recent 0.16.x clients. So unless you can give me accurate and reliable steps to reproduce this bug, I will have to close this as "worksforme".
Prepare:
xpra start :100 --start-child=firefox --bind-tcp=0.0.0.0:10000
<path-to-xpra-client>\Xpra-0.16.0-r11304\Xpra_cmd.exe attach tcp:IP:10000 --opengl=yes
Reproduce:
You can also watch this video.
Watched the video and this is exactly what I am doing when testing. The only difference being that I don't have opengl enabled in the win32 client. (because it is running in a virtual machine) I see that you are forcing it on, why? Have you tried disabling opengl in the client? Either from the command line or after connecting from the systray?
I tried disabling opengl in the client, as I said in comment:27, from the command line and this bug don't appear again. Btw, I think if I force opengl on, the client will be using GPU and improve Xpra's performance.
I tried disabling opengl in the client, as I said in comment:27, from the command line and this bug don't appear again.
Ah, I didn't understand this:
I tried narrowing opengl off, and images were still good
Please post your wiki/ClientRendering/OpenGL debugging information (the gl_check tool output from the client). Is opengl enabled or disabled by default on your system? Is this why you force it on?
OpenGL debugging information from the client:
OpenGL_accelerate module loaded OpenGL properties: * GLU extensions : GL_EXT_bgra * GLU version : 1.2.2.0 Microsoft Corporation * accelerate : 3.1.0 * display_mode : DOUBLE * gdkgl.version : 6.1 * gdkglext.version : 1.2.0 * gtkglext.version : 1.2.0 * has_alpha : True * max-viewport-dims : (16384, 16384) * opengl : 4.2 * pygdkglext.version : 1.0.0 * pyopengl : 3.1.0 * renderer : Intel(R) HD Graphics 4600 * rgba : True * safe : True * shading language version : 4.20 - Build 10.18.10.3412 * texture-size-limit : 16384 * transparency : False * vendor : Intel * zerocopy : True
Is opengl enabled or disabled by default on your system?
Opengl is enabled by default.
Is this why you force it on?
Yes, as I said. I think if I force opengl on, the client will be using GPU and improve Xpra's performance. It's right, or I misunderstand something?
If you force something on, please mention it early. wiki/ReportingBugs clearly states that you should include the command lines for example and also clearly states: changes made to the default configuration, if any.
Had I known that you were forcing opengl on from the start, this bug would have taken minutes to close rather than the 9 months it has now taken.
Closing as invalid: we disable opengl on intel chipsets because the drivers are not reliable. Anyone second guessing the software and forcing this on should be prepared to deal with the fallout, such as this one.
See wiki/ClientRendering/OpenGL
this ticket has been moved to: https://github.com/Xpra-org/xpra/issues/809