Xpra: Ticket #2669: xpra-client cores considerably with RubyMine and SublimeMerge

I have seen a surge of 4.0 client cores these last few days. It could be that I'm using xpra more, or some changes are to blame for this.

I've sent you around 10~20 cores over e-mail; please process the application-specific cores first, and then descending on creation time.

Especially the RubyMine one comes for every other action.

Mon, 23 Mar 2020 12:21:53 GMT - stdedos: description changed

description modified (diff)

Wed, 25 Mar 2020 13:41:55 GMT - stdedos: summary changed

summary changed from xpra-client cores with RubyMine and SublimeMerge to xpra-client cores considerably with RubyMine and SublimeMerge

One core that's unrelated to RubyMine / SublimeMerge is this coming from Nemo file manager (Split from Nautilus):

"Xpra-Python3-x86_64_4.0-r25747\xpra_cmd" attach ssh://user@ip/20 --ssh="plink -ssh -agent" --modal-windows=no --title="@title@ on @hostname@/@server-display@"
2020-03-25 14:44:28,564 Xpra GTK3 client version 4.0-r25747 64-bit
2020-03-25 14:44:28,566  running on Microsoft Windows 10
2020-03-25 14:44:28,632 Warning: failed to import opencv:
2020-03-25 14:44:28,633  No module named 'cv2'
2020-03-25 14:44:28,633  webcam forwarding is disabled
2020-03-25 14:44:29,323 GStreamer version 1.16.2 for Python 3.8.2 64-bit
2020-03-25 14:44:29,726 keyboard layout code 0x409
2020-03-25 14:44:29,726 identified as 'United States - English' : us
2020-03-25 14:44:29,843 OpenGL_accelerate module loaded
2020-03-25 14:44:29,880 Using accelerated ArrayDatatype
2020-03-25 14:44:30,592 Warning: vendor 'Intel' is greylisted,
2020-03-25 14:44:30,593  you may want to turn off OpenGL if you encounter bugs
2020-03-25 14:44:30,608 OpenGL enabled with Intel(R) HD Graphics 4000
2020-03-25 14:44:30,818  keyboard settings: layout=us
2020-03-25 14:44:30,821  desktop size is 1600x900 with 1 screen:
2020-03-25 14:44:30,822   Default (423x238 mm - DPI: 96x96) workarea: 1600x860
2020-03-25 14:44:30,823     Generic PnP Monitor (309x174 mm - DPI: 131x131)
2020-03-25 14:44:35,189 enabled remote logging
2020-03-25 14:44:35,197 Xpra GTK3 X11 server version 3.0.8-r25722 64-bit
2020-03-25 14:44:35,201  running on Linux Ubuntu 16.04 xenial
2020-03-25 14:44:35,220 Attached to ip:22
2020-03-25 14:44:35,223  (press Control-C to detach)
(xpra_cmd:6620): Pango-WARNING **: 14:44:35.861: couldn't load font "Bitstream Vera Sans Not-Rotated 14.662109375", falling back to "Sans Not-Rotated 14.662109375", expect ugly output.
2020-03-25 14:44:36,715 sound output using 'opus' audio codec
2020-03-25 14:44:36,886 UI thread is now blocked
2020-03-25 14:44:36,908 UI thread is running again, resuming
2020-03-25 15:32:56,386 server is not responding, drawing spinners over the windows
2020-03-25 15:32:56,639 server is OK again

And the client died almost immediately after that. Attaching full server also; idk if it's wise to truncate it or not.

Wed, 25 Mar 2020 13:42:55 GMT - stdedos: attachment set

attachment set to redact-nemo-2669-display-_20_TIMESTAMP.log

Fri, 27 Mar 2020 07:38:59 GMT - stdedos:

Intellij Tools crashes may be related to autocompletion.

Thu, 02 Apr 2020 08:28:49 GMT - stdedos:

It turns out, that most of this (if not all) is mitigated by --opengl=no.

I might be nice if you opened them to verify, but I guess for heavy seamless usage, I have to go OpenGL-less

Thu, 02 Apr 2020 10:04:54 GMT - Antoine Martin: owner changed

owner changed from Antoine Martin to stdedos

Intellij Tools crashes may be related to autocompletion.

Tooltips are known to cause problems with Intel opengl...

It turns out, that most of this (if not all) is mitigated by --opengl=no.

I am very tempted to switch the Intel driver back to disabled by default.

The log you have added is a server log, so it won't have any client crash related info in it. Even the crash logs are usually useless. The only ways to debug opengl client crashes are:

run with -d all and collect the end of all the crash lots to spot a pattern
run the client with apitrace

Thu, 02 Apr 2020 10:15:57 GMT - stdedos:

Replying to Antoine Martin:

The only ways to debug opengl client crashes are:

run with -d all and collect the end of all the crash lots to spot a pattern
run the client with apitrace

If that's the only way, considering the Intel is already graylisted, it sounds too much hassle for both of us; taking also into account that maybe you cannot do anything about this Intel chipset just works badly with OpenGL.

Maybe I could try changing laptops altogether, but its battery life is just too amazing - even taking into account its quirks and issues.

You could re-use this ticket as a placeholder for opening the cores if you want; given all of what's said here, I don't see any meaningful way forward.

Wed, 08 Apr 2020 13:34:51 GMT - Antoine Martin:

#2721 may help.

Wed, 08 Apr 2020 13:48:39 GMT - stdedos:

Replying to Antoine Martin:

#2721 may help.

This can be done on window-by-window basis? Waaau :-D

In any case - I assume this is a solely v4.x fix which I cannot install on my server 😕

However, I am planning a Ubuntu 16.04 direct-to 20.04.1 upgrade for my server (when it lands), so ... Hopefully in a month's time or so.

Wed, 08 Apr 2020 14:15:24 GMT - Antoine Martin:

In any case - I assume this is a solely v4.x fix which I cannot install on my server 😕

This is scheduled for trac/milestone/3.1, but 3.1 is still a long way away.

Wed, 08 Apr 2020 14:44:21 GMT - stdedos:

Yeah, I think I may get to update my server first 😂

Sat, 11 Apr 2020 13:59:54 GMT - Antoine Martin:

r26044 is very likely to help with this.

Fri, 08 May 2020 14:17:40 GMT - Antoine Martin: status changed; resolution set

status changed from new to closed
resolution set to worksforme

Feel free to re-open if it still occurs with v4.

Sat, 23 Jan 2021 05:57:38 GMT - migration script:

this ticket has been moved to: https://github.com/Xpra-org/xpra/issues/2669