FvwmPager crashes at restart

[dirteat]

Hi there,

• Fvwm3 version (run: fvwm3 --version)
  fvwm3 1.1.1 (released)
  with support for: ReadLine, XPM, PNG, SVG, Shape, XShm, SM, Bidi text, XRandR, XRender, XCursor, XFT, NLS

• Linux distribution or BSD name/version
  Mageia, devel version Cauldron

• Platform (run: uname -sp)
  x86_64

Expected Behaviour

My config Swallow the FvwmPager into a button, works fine when fvwm3 starts, but the pager disappear when hitting the "restart" menu and the logs show that it crashed.

Enabling logging

pkill -USR2 fvwm3

Nothing in the logs appears.

Steps to Reproduce

How can the problem be reproduced? For this, the following is helpful:

• Reduce the problem to the smallest fvwm configuration example (where
  possible). In progress!

• Does the problem also happen with Fvwm2?
  The problem does not appear with fvwm2, but it was also not present with the previous version of fvwm3, 1.0.6a

Include your configuration with this issue.
It is very long, I'll provide a thin one once I have isolated the minimal config triggering it. Possibly the swallowing?

Test (EnvIsSet FVWM_IS_FVWM3) *DeskButtonsBox: (Swallow (NoHints) "Desk 0" "FvwmPager")

Does Fvwm3 crash?

No, but the Pager module does:

: [🡕] Process 2649567 (FvwmPager) of user 1000 dumped core.

Stack trace of thread 2649567:
#0 0x0000000000408d4d set_vp_size_and_loc (FvwmPager + 0x8d4d)
#1 0x000000000040e198 initialize_fpmonitor_windows.part.0 (FvwmPager + 0xe198)
#2 0x000000000040ed81 init_fvwm_pager (FvwmPager + 0xed81)
#3 0x000000000040755b main (FvwmPager + 0x755b)
#4 0x00007fa3be2d1737 __libc_start_call_main (libc.so.6 + 0x23737)
#5 0x00007fa3be2d17f5 __libc_start_main@@GLIBC_2.34 (libc.so.6 + 0x237f5)
#6 0x0000000000407981 _start (FvwmPager + 0x7981)
ELF object binary architecture: AMD x86-64

Cheers,
Chris.

[somiaj]

One thing I do notice is your FvwmButtons configuration shouldn't use the name of the page, if you add a name to Desk 0, the swallow command may not be able to match. Use the class/resource. Something like this would be better.
Also can you share your full FvwmPager configuration.

Test (EnvIsSet FVWM_IS_FVWM3) *DeskButtonsBox: (Swallow (NoHints) "FvwmPager" "FvwmPager")

[somiaj]

@ThomasAdam the only thing I can see is the method set_vp_size_and_loc does divide by fp->virtual_scr.{VxPages,VyPages,VWwidth,VHeight}, is there any chance one of those is zero. This isn't something I've ran into in the numerous times I restart fvwm though. Wonder if there is some issue with their monitor setup?

[ThomasAdam]

@somiaj

Maybe. But the stacktrace would indicate a SIGFPE -- although I can't tell from the informatioon we've been provided if that's true.

@dirteat -- I'm going to need to see a backtrace with debug symbols, please. I'd also want to see a fuller log from $FVWM_USERDIR/fvwm3-output.log. If you're able to reproduce this, you can add -v to the fvwm3 commandline, and send that log file through.

[dirteat]

All right, thanks for the feedback, I'll fix my swallow! I'll provide a bt, but the segfault is occuring one over many restarts (I can tell now, I've been toying with that since a few hours), and at all restarts the pager is not restarted. So the segfault and my issue here are not in one-to-one correspondence.

BUT, I finally found out the culprit in my configuration:

*FvwmPager: HilightColorset * 8

The number "8" is just a user-defined colorset:

Colorset 8 fg white, bg black, RootTransparent buffer, Tint green 35, \
	 bgTint green 35 

The Pager not restarting is occurring as soon as I set this option. Reading the man, I've just discovered it was deprecated, so this might be my fault indeed. That config is following me since decades :)

I think you should be able to reproduce. First start with that option set, everything works fine, a restart and the pager is gone without any log.

Let me know if you need me to do other tests. I'll backtrace the segfault if I can trigger it in a rational way.

Thanks!!

NB: The monitor setup is unusual indeed (beamer on HDMI0, DP-4 main screen. Due to their different aspect ratios and resolution, HDMI0 is centered within DP4).

HDMI-0 connected 1920x1080+320+260
DP-4 connected primary 2560x1600+0+0

[ThomasAdam]

@dirteat

I really don't think the HilightColorset * 8 config line is the problem.

Given:

Stack trace of thread 2649567:
#0 0x0000000000408d4d set_vp_size_and_loc (FvwmPager + 0x8d4d)

What I need from this, is a reliable stack trace to tell me where in set_vp_size_and_loc the issue is occurring. I really don't want to guess either, and with a stacktrace with debug symbols, I won't have to.

[ExecutorElassus]

@dirteat

The Pager not restarting is occurring as soon as I set this option. Reading the man, I've just discovered it was deprecated, so this might be my fault indeed. That config is following me since decades :)

Can you tell me what man page you're looking at, and what needs changing? I also have a very old config, and this week's update has caused the colors on FvwmPager to act weird (the active highlight color is now default light gray, not the colorset I specified).

[somiaj]

@ExecutorElassus Please don't piggy back on other issues, it distracts from the issue at hand which is FvwmPager crashing. If the FvwmPager manual page doesn't help you figure out how to set the colors you want for the pager, you can ask for help on the forums.

[dirteat]

@ThomasAdam
I will provide a backtrace.

This is weird, I can reproduce 100% of the time on my machine, without even using a Swallowing, that the Pager disappear at restart only when I set:

*FvwmPager: HilightColorset * 8

@ExecutorElassus
Yes, the Pager is now default light gray, without HilightColorset, I have no idea how to color it (man fvwm3all).

[dirteat]

@ThomasAdam

I really don't think the HilightColorset * 8 config line is the problem.

You're probably right, I've tested on another computer running fvwm3 1.1.1 within mono screen output (DP-0 connected primary 5120x2160+0+0), I don't get the bug anymore even though I am setting the *FvwmPager: HilightColorset * 8... Backtrace coming....

[dirteat]

gdb -p $pidofpager running and waiting... (the segfault is not systematic).

Still, I've narrowed down the issue a bit more. In addition to HilightColorset, my screen configuration has to be 2 screens without a match on the upper right corner!?

For instance, this configuration works (Pager restart fine with HilightColorset set)

Any other configuration I've tried with 2 screens not touching in the upper right corner fail, like this one (the default):