Created attachment 1575470 [details] Screenshot of typing booster with latin-pre filter enabled. Description of problem: With Typing Booster enabled (configuration provided in attached screenshot), pressing the ctrl+up_arrow key combination in a web browser (tested with both chrome and firefox) causes the operating system to hard freeze/lock up/hang. The problem does not occur in gnome-terminal. I haven't tested anything else. Version-Release number of selected component (if applicable): ibus-typing-booster-2.6.1-1.fc29.noarch How reproducible: Every time. Steps to Reproduce: 1. Enable typing booster (I'm not sure what configuration is required, but all I have enabled is the "Latin-pre" input method. 2. In Firefox's address/search bar, type "byebye", then press the ctrl+up_arrow key combination. Actual results: Hard freeze of operation system requiring a power-cycle to restore. Expected results: Cursor moves to beginning of line. Additional info:
Created attachment 1588402 [details] ibus-typing-booster-fedora-30-firefox-url-bar-control-up.webm I tried this on Fedora 30, both with Gnome on Xorg and Gnome Wayland and cannot reproduce it there. See the video.
Hi Mike: Thanks for checking on this. And thanks for developing typing booster. I find it extremely useful. Maybe the issue is something to do with the particular settings that I am using then. I hadn't thought of doing a video. I'll try to share my desktop out to another system and record the problem so you can see what is happening. I just started my work shift, so it will be eight hours at least before I can get the video posted. Thanks.
You also use Gnome? I have not tested on F29 yet, but I never saw something like this while I was using F29 either.
Yes, it is Gnome and it is F29. One difference between my setup and what I saw in the video that you posted is that I don't have the auto-complete menu set to display by default. What I have configured is just using the latin-pre filter to enable typing accents. I'll try to share a video demonstration of the problem later today (maybe in about 3 or 4 hours depending on how long it takes me to get VNC sharing and recording setup). Thanks.
Created attachment 1588605 [details] os hard-freeze demo Here is the promised video recording of this bug. Notice at the end, when I press ctrl-up_arrow, the cursor stops blinking and the animated gif stops. Even my mouse is unresponsive. I have to hard-powercycle the computer to get it going again. This problem affects both Firefox and Chrome. It does not occur if typing booster is disabled. Thanks.
Is there anything interesting in the log file?: ~/.local/share/ibus-typing-booster/debug.log One can get more debugging output by putting a number >= 2 in the "Debug level" entry in the setup tool.
Hi Mike: I'm out of town right now, but I should be able to check those logs Thursday evening or Friday. Thanks.
Created attachment 1592162 [details] typing-booster log Here is the log file you mentioned. Nothing in it looks particularly interesting to me. I'm going to try increasing the log level and triggering the bug next ...
Created attachment 1592164 [details] typing-booster log verbosity level 3 Here is the same log file after changing the verbosity level to 3 and triggering the bug. I still don't see anything interesting. I suspect that this bug might be preventing the logs from getting "flushed" to disk, in which case, the error may never show up regardless of the log level. If you cannot reproduce the bug, then I guess it isn't as big a deal as I thought. Still though, I find it a bit annoying. Thanks.
(In reply to Gregory Lee Bartholomew from comment #9) > Created attachment 1592164 [details] > typing-booster log verbosity level 3 > > Here is the same log file after changing the verbosity level to 3 and > triggering the bug. I still don't see anything interesting. I suspect that > this bug might be preventing the logs from getting "flushed" to disk, in > which case, the error may never show up regardless of the log level. > > If you cannot reproduce the bug, then I guess it isn't as big a deal as I > thought. Still though, I find it a bit annoying. Well, it is really a big deal if that happens, hard freezing the operating system sounds very serious. But it is hard to tell what is causing this if I cannot reproduce it. And you are right, there is nothing interesting in the log file, not even in the logfile with verbosity level 3.
OK. Let me know if you have any other ideas about how to go about troubleshooting this. In the meanwhile, I just searched bugzilla for any other references to "freeze" and I found this: https://bugzilla.redhat.com/show_bug.cgi?id=1626539 I see some references to the i915 video driver, which I believe my home system is also using. His report says "at any time using Gnome" though, which is a little different from my case. Also, his report is against xorg-x11-drv-intel, but I think my home system is using wayland (although I think I read that Gnome will fall back to the x11 driver under some circumstances, so I'm not completely sure about that).
Here is a more recent one that also reports OS freezes in Fedora 29 when running Firefox or Chrome: https://bugzilla.redhat.com/show_bug.cgi?id=1667550 That one is against gnome-shell though.
Not sure if this will be helpful to finding the common cause of these freezes, but here is an excerpt from the #fedora IRC channel where another user reported hard freezes in Fedora. He is using Fedora 30 though and he is running the Cinnamon spin, not Gnome. It looks like he is also using the i915 driver: T 1562769374 1sdgathman1> I have a T61 running Fedora 30 Cinnamon spin fresh install, fully updated. On normal startup to GUI, it hangs in a hard CPU loop (no dmesg, no nothing) shortly after starting lightdm. T 1562769427 1sdgathman1> However, if I start up in mode 3, then do "telinit 5", it works fine (after systemctl restart NetworkManager - which stops working on switching modes). T 1562769467 1sdgathman1> I don't even know what component to report the bug against. T 1562769575 1sdgathman1> It also works if I add nomodeset to kernel cli, booting directly to mode 5 with no system freeze. But then there is no video acceleration, and cinnamon runs in software rendering mode, and videos require 2x the CPU. T 1562769665 1sdgathman1> This system worked in F29. ... T 1562769925 1sdgathman1> The T61 is from LiveCD fresh install. ... T 1562770014 1dowdle1> sdgathman: For your laptop, I'd recommend trying a different spin as Cinnamon wants accellerated 3D... although it is supposed to switch to software rendering if it has trouble with it... but it would be interesting to see how, say, XFCE works or not. T 1562770034 1sdgathman1> I hangs before starting cinnamon. T 1562770046 1sdgathman1> Only lightdm is running when it freezes. ... T 1562770107 1dowdle1> sdgathman: Well, it is most likely one of the core components like the kernel... or an xorg driver. T 1562770133 1sdgathman1> All those are working fine - if they are started in proper sequence. ... T 1562771648 1sdgathman1> There is something that freezes the kernel if it runs after starting lightdm ... T 1562771900 2gregbartholome30 It is a pain, but you can also try "dnf downgrade <package-name>" on whatever packages you suspect might have a new bug to see if you can narrow down the problem. T 1562771938 1sdgathma> gregbartholomew: all packages seem to be working great - as long as they startup in the right sequence. T 1562771963 1sdgathman1> Once system is up, video, wireless, and sound, and all are quite stable. ... T 1562772121 1sdgathman1> Unfortunately, unlike the Dell, it has never worked with default boot sequence. T 1562772167 1sdgathman1> It freezes system, cpu loop, keyboard dead, about 3 secs after loading lightdm. T 1562772181 1sdgathman1> bugs_: yes, it works when I do that T 1562772195 1sdgathman1> Also works with "telinit 5" after multi target T 1562772210 1sdgathman1> So it is clearly a startup timing problem again. T 1562772224 2gregbartholome30 Oh, it completely freezes the OS? Requiring a hard power-cycle to restart? T 1562772231 1sdgathma> gregbartholomew: yes T 1562772247 1sdgathman1> And with CPUs and fan going full tilt ... T 1562772372 2gregbartholome30 I am getting a hard os freeze on one of my home systems as well, but in my case, it is triggered by a weird combination of gnome plugins and key presses. I filed a bug report for it and I am waiting for someone more talented than I to trace down the real problem. T 1562772411 1sdgathman1> I can't tell what to file a bug report against. T 1562772427 2gregbartholome30 I think anytime the keyboard/mouse stops responding, the cursor stops blinking etc. it is a kernel-level bug. T 1562772464 1sdgathman1> Ok, that's helpful ... T 1562772509 1sdgathman1> I can see that as something starting about the time it freezes... ... T 1562772696 1sdgathman1> bugs_: again, cinnamon never runs T 1562772707 1sdgathman1> the *only* gui running is lightdm ... T 1562772866 1sdgathman1> Nope. Still freezes 3 secs after lightdm loads T 1562772887 1sdgathman1> You can still switch console, start to login, etc during that 3 secs... T 1562772907 1sdgathman1> I can type over 100wpm, but still, that is pretty tight... T 1562772918 2gregbartholome30 sdgathman: you narrowed it down to the video card earlier for this system right? T 1562772944 1sdgathma> gregbartholomew: no. T 1562772970 2gregbartholome30 Did you say that nomodeset made this one work or was that the other one? T 1562772999 1sdgathman1> We tried using an xorg.conf, but that just made me notice that it works if you don't start in mode 5. T 1562773029 1sdgathma> gregbartholomew: nomodeset makes the system start without freezing in mode 5 - but then there is no GPU accel. T 1562773064 1sdgathman1> And when you start cinnamon, it then uses software rendering. T 1562773075 1sdgathman1> bugs_: local auth only T 1562773130 2gregbartholome30 still, that might hint at where the problem could be. T 1562773131 1sdgathman1> Ah, I should try starting in single user mode (is that still gotten by appending 's' ?) T 1562773137 1sdgathman1> And then telinit 5 T 1562773155 1sdgathman1> That will bisect the problem T 1562773172 1bugs_1> sdgathman have you ever had to put vmalloc=256M in your grub? T 1562773225 2gregbartholome30 maybe run lsmod to see which driver your video card is using and then modinfo <modname> to see if it takes any parameters. T 1562773290 1sdgathman1> starting in single user works fine, switching to mode 5 from single user still freezes. So it is something that happens between single user and graphics. T 1562773303 1sdgathman1> But before mode 3. T 1562773313 2gregbartholome30 it sounds like a problem with the video driver then T 1562773313 1sdgathman1> Because starting in 3 and going to 5 works. T 1562773353 1sdgathman1> Since nomodeset "fixes" the startup freeze, it does seem like video driver. T 1562773374 1sdgathman1> But since 3 -> 5 works, there shound be a work around. T 1562773403 2gregbartholome30 it may just be initializing the video card in a slightly different way when you do that sequence. T 1562773434 2gregbartholome30 how about if you disable rhgb? T 1562773596 1sdgathman1> removed rhgb and quiet - it worked going directly to graphics T 1562773617 2gregbartholome30 weird T 1562773772 1sdgathman1> still freezes going directly to 5 with quiet, but not rhgb T 1562773858 1sdgathman1> Well, it is stable with both rhgb and quiet removed. T 1562773905 1sdgathman1> This is a new linux user, and seeing startup messages is part of the look and feel. :-) T 1562773931 1sdgathman1> It is actually helpful when something goes wrong: "What's the last line on the screen?" T 1562773946 2gregbartholome30 maybe file a bug report against the video driver then. T 1562773973 1sdgathman1> That would be the xorg-x11-drv-intel ? ... T 1562774380 1sdgathman1> I spoke too soon. Went to pee, and when I came back T61 was frozen. It just takes more than 3 secsc now. T 1562774396 1sdgathman1> But booting to mode 3 first still works... T 1562774539 1sdgathman1> bugs_: takes 5 secs to freeze without quiet... T 1562774556 1sdgathman1> So it does change the timing T 1562774583 1sdgathman1> Still not long enough to switch console and grab a systemd-analyze ... T 1562774608 2gregbartholome30 sdgathman: it sounds like a pretty low-level driver/firmware problem in any case. about all you can do is look for firmware updates and try fiddling with some driver parameters (if it takes any). T 1562774652 1sdgathman1> ok T 1562774764 1sdgathman1> Still freezes 5 sec after starting lightdm T 1562775099 1sdgathman1> There is no i915.conf in /etc/modprobe.d T 1562775153 1sdgathman1> Does that turn off accel ? T 1562775175 1sdgathman1> So makes cinnamon slow T 1562775247 1sdgathman1> The accel is working fine once startup is finished. At this point, I am trying to automate the delay introduced by starting in mode 3 first. T 1562775528 1sdgathman1> Is there another DM to try? T 1562775640 1sdgathman1> lightdm already runs After=rc-local.service T 1562776600 1sdgathman1> bugs_: switched to lxdm - and it worked! T 1562776606 1sdgathman1> rebooting again just to make sure ... T 1562776711 1sdgathman1> didn't work that time - froze T 1562776940 1sdgathman1> Did it again, pressed escape to display startup messages, and didn't try to login until disk stopped blinking - worked. ...
I just found a work-around for this bug. This bug does not occur if I select "GNOME on Xorg" from GDM. I am content with this work-around, so you can close the bug if you think it is appropriate to do so. Another aspect of my system that I had forgotten about is that I had added "intel_idle.max_cstate=1" to my kernel command line years ago to work around "the intel c-state bug": https://bugzilla.kernel.org/show_bug.cgi?id=109051 According to the posts in the above thread, this old work-around is no longer working for some users on more recent kernels. I suspect that this is my real problem and that the freeze can be triggered consistently with typing booster + cntl + up-arrow + ( firefox or chrome ) + Wayland is just a really odd coincidence. Thanks, and sorry for troubling you with something that probably isn't directly related to your plugin. Thanks also for Typing Booster -- it is a great plugin!
(In reply to Gregory Lee Bartholomew from comment #14) > I just found a work-around for this bug. This bug does not occur if I select > "GNOME on Xorg" from GDM. Yes, unfortunately there are a lot of problems when using ibus-input methods on Gnome Wayland which do not occur on Gnome Xorg. This must be fixed of course as Wayland is the future, but it doesn’t seem easy. > I am content with this work-around, so you can > close the bug if you think it is appropriate to do so. I am closing it as WORKSFORME because I cannot reproduce it. I think you are right in guessing that it is caused by something else and not ibus-typing-booster.