hdajackretask wont start: error 4 in libc

[Gladdle]

Maybe the topic is better in Multimedia, i dont know, so i wrote it down here. In the German Forum i have allready asked, but they didnt know an answer. The problem is simple: to configure my onboard sound card i need the software "hdajackretask", its a part of "media-sound/alsa-tools" (Currently installed: 1.0.29:0.9:gentoo) if i want to start the Software (doesnt matter if i start under kde plasma or cinnamon), the Software wont start and "dmesg" says this:

[Hu]

That message means that hdajackretask crashed. You need to determine why it crashed. Start by running it under gdb and examining the backtrace when the program receives a SIGSEGV. If you have not already, you will need to build the package with debug symbols enabled and prevent Portage from stripping those symbols prior to install.

[Gladdle]

Okay, i am sorry, i didnt understand everything. I have edited make.conf and expand CFLAGS="... -ggdb". Then i have done a

[Hu]

You might find the Introducing gdb subsection of the QA Backtraces document helpful. You can debug a core file generated by a prior run (assuming one was generated and that you can find it) or you can use gdb -q --args hdajackretask --arguments-to-hdajackretask to run a new instance and let it crash.

If you still need more help, please post back. I am not familiar with this tool, so I do not know specifically what text you should use for --arguments-to-hdajackretask. Perhaps some other forum user can tell you what parameters your chosen desktop environment normally uses when it runs this program. If you are out of ideas, you could leave the arguments blank and hope that it crashes when run with no arguments. This might or might not be the same crash that you are trying to debug, though.

[Gladdle]

seems like it found something:

[Hu]

I think your Google result is misleading. The output you showed says that a crash happened in libc. From your dmesg output, we expected to see that result. The "File or directory not found" (canonically, this is ENOENT: "No such file or directory" when printed as an error in English) is that gdb cannot find the file that defines the function that faulted, so it cannot show you the source. This is true, but not important. Neither of us wants to look at that file. It is just a utility. To use a carpentry analogy, we know that the hammer hit the wrong spot, but we do not need to examine the hammer. We need to examine who wielded the hammer and why the wielder aimed it at the wrong spot. For that, we want to see the backtrace that shows why the program passed invalid inputs to glibc. Please recreate the situation you showed, then run bt full and post its output.

The Google hits for vlc, KDE, and so on are likely because all those projects have, at various times, had bugs that led to them passing invalid input to glibc, which led to a fault when glibc trusted that the input was valid.

Fixing this may be too complicated for you, but providing the information necessary to help an expert fix it is just a matter of finding the right commands to run. We can help you find those commands and direct you to someone who can use that information to fix the root cause.

[Gladdle]

Finaly i did this:

[Hu]

You may need to build some of those components with a lower optimization level. The compiler seems to be optimizing the generated code to the point that gdb cannot find important variables. Try first with -Og instead of -O2. If you still see <optimized out>, you may need to use -O0.

When you rebuilt gtk+, did you also have FEATURES=nostrip? If Portage is allowed to strip the generated files, your use of -ggdb will not matter.

[Gladdle]

The errors get worser, not better. First i have changed the make.conf:

[Hu]

You should usually not set USE=debug. That enables package-specific code that may be debugging related. It is not what you want here. Please remove USE=debug and add FEATURES="nostrip". That is a valid setting, but may not be listed in your make.conf by default. After making those changes, rebuild again the packages you rebuilt for your most recent post: emerge --newuse -1av media-sound/alsa-tools sys-libs/glibc x11-libs/gtk+. Then retry the gdb command you showed in your most recent post.

[Gladdle]

I am really sorry, i missunderstand you. My english isn't the best. But this time it seems to be good:

[Hu]

There is no need to apologize. I appreciate you using English for this, since it makes your post available to a wider audience.

Unfortunately, that output is not as helpful as I had hoped. That is not your fault. You did as I asked. I think we can still make some assumptions about what is happening, but I also think we may be at the point that you need to involve someone familiar with this code. I had hoped it would be a simple error that could be spotted by inspection and resolved by adding error checking. It now looks like a more subtle error that will require an understanding of the affected program in order to fix it. Looking at main-gtk.c:

[tsiru]

Hello!

If someone still need workaround, I found one (tested on latest Clevo notebook):

Create/append following files:

/etc/portage/bashrc:

[Hu]

That patch is dangerous and wrong. The modified program conditionally initializes s, but does not set any state to record whether s has a valid value. It then unconditionally uses s.

It might be sufficient to ensure s is initialized to nullptr on the else path.

[ivan2k]

Just for other users who get there with the same error: i think you miss the CONFIG_SND_HDA_RECONFIG on your kernel configuration.