#39 lircd segfaults [duplicate of #19]

[Bit Twister]

lircd[31129]: segfault at 0 ip 00007fa327620a9e sp 00007fff91d5a8f8 error 4 in libc-2.19.so[7fa32758d000+1b0000]

systemctl start lircd

systemd[1]: Starting LIRC Infrared Signal Decoder...
systemd[1]: Started LIRC Infrared Signal Decoder.
lircd[31129]: Cannot open /sys/class/rc
kernel: lircd[31129]: segfault at 0 ip 00007fa327620a9e sp 00007fff91d5a8f8 error 4 in libc-2.19.so[7fa32758d000+1b0000]
systemd[1]: lircd.service: main process exited, code=killed, status=11/SEGV
systemd[1]: Unit lircd.service entered failed state.

[Alec Leamas]

Hm... Is this related to [tickets:#19] or [tickets:#21] ?

[Bit Twister]

Looking pretty close to [tickets:#21]
Commented out Connect, still seg faulted.

Install was

urpmi lirc

To satisfy dependencies, the following packages are going to be installed:
Package Version Release Arch
(medium "core")
lib64iguanair0 0.99 1.svn959.11.> x86_64
lib64irman0 0.4.5 5.mga4 x86_64
lirc 0.9.1a 1.mga5 x86_64

Guess I need 9.1b and the man page fix before I can proceed any farther.

[Alec Leamas]

0.9.1b is as of yet not released. It's just just the patches described in [tickets:#19] and [tickets:#21]. Could you apply both of them in a first step? If it still crashes, we'd need a stacktrace; there is some test on how to make that in CONTRIBUTE.md in current sources (not 0.9.1)

[Alec Leamas]

Had some private messages with reporter. We need more info to clarify if this is [tickets:#19], [tickets:#21] or something else.

[Bit Twister]

First, this seg fault problem will not start to get resolved by me until [tickets:#38] is fixed.

Downloaded your 91a tar file. "make install" locations were nowhere near Mageia's.

Have downloaded Mageia's src rpm to see if I can generate a debugable exe. Having no luck so far finding Mageia's lircmd make file. Going to have to learn how to modify your make file to fit Mageia install.

Found "man make", progress is going to be pretty slow.
Mageia's developers/packagers are pretty much buried in work trying to get all the packages to install let alone work application problems. Bug squad has not even assigned https://bugs.mageia.org/show_bug.cgi?id=13978 to anyone yet.

[Alec Leamas]

On 09/01/2014 02:38 PM, Bit Twister wrote:

First, this seg fault problem will not start to get resolved by me until [tickets:#38] is fixed.

This is a bug, agreed. That said, you can just run ./configure + make to get a complete set of manpages in doc/man and doc/html. So you don't really need those. I guess that's the reason no-one has spotted this before you.

Downloaded your 91a tar file. "make install" locations were nowhere near Mageia's.

This is normal. When packaging, packagers sets the paths explicitly. You could look at the spec file, a starter is probably to add --prefix=/usr to configure. On a 64-bit host you might also need to --libdir, not sure how Mageia handles that.

Have downloaded Mageia's src rpm to see if I can generate a debugable exe. Having no luck so far finding Mageia's lircmd make file. Going to have to learn how to modify your make file to fit Mageia install.

If mageia does the same as Fedora (which I really have no idea of) the rpm build might generate separate debug packages which you have to install to get debug symbols.

Found "man make", progress is going to be pretty slow.
Mageia's developers/packagers are pretty much buried in work trying to get all the packages to install let alone work application problems

I can understand that, I package this myself for Fedora....Again, that said, isn't there any forum where you can discuss this with other Mageia users which might have a little more experience? To apply a patch is actually a thing many people do.

I understand this is an uphill battle if you have to start with 'man make'... but from my perspective, having patches available is the normal way to handle this. This is how ArchLinux has solved it, and Fedora as well as soon I get some time...

Also, for obvious reasons I'd really have this sorted out before we release 0.9.1b (if it will be release at all).

[Bit Twister]

All right. You need to handle the problem where there is no /dev/lirc0.
I am using network tuners for lircd input and is working on 9.0 with no /dev/lirc0.

With lircd 9.0 I provided access to the tuner using udp on port 6546 via the LIRCD_OPTIONS="-H udp -d 6546" feature.

So far I have not found where/what I should set those values in /etc/lirc/lirc_options.conf. Current settings are:
$ grep -v ^# /etc/lirc/lirc_options.conf

[lircd]
nodaemon = False
permission = 666
driver = default
output = /var/run/lirc/lircd
pidfile = /var/run/lirc/lircd.pid
plugindir = /usr/lib/lirc/plugins
allow-simulate = No
repeat-max = 600

[lircmd]
uinput = False
nodeamon = False

I've done the .config, make, cd daemons, ddd lircd
and as far as I can tell, seg fault is at lircd.c line 2337
if (hw.device != NULL && strcmp(hw.device, lircdfile) == 0) {

hw.device=/dev/lirc0
lircdfile=0x0

[Alec Leamas]

Lomg story short this looks like [tickets:#19]. If so, you can walk-around it by changing the "output = /var/run/lirc/lircd" in lirc_options.conf to "lircdfile = /var/run/lirc/lircd". This is not the final solution, though.

As for your other problems -H corresponds to --driver and -d to --device, both of which are available in the config file. You can use lircd --help to find this (or of course the manpage).

The LIRCD_OPTIONS "feature" has nothing to do with the upstream lirc distribution. It's added by the packaging folks, in this case Mageia.

Cheers!

--alec

[Bit Twister]

Alec wrote

Long story short this looks like [tickets:#19]. If so, you can walk-around
it by changing the "output = /var/run/lirc/lircd" in lirc_options.conf
to "lircdfile = /var/run/lirc/lircd".

That works for me. Since #19 appears closed, looks like the segfault is not going to be fixed.

You can use lircd --help to find this

That works.

(or of course the manpage).

Not really.

[root@tb ~]# man lircd
fgets: No such file or directory
Error reading man page /usr/share/man/man8/lircd.8.xz
No manual entry for lircd

ll /usr/share/man/man8/lircd.8.xz

-rw-r--r-- 1 root root 32 Aug 23 15:12 /usr/share/man/man8/lircd.8.xz

man lircd

fgets: No such file or directory
Error reading man page /usr/share/man/man8/lircd.8.xz
No manual entry for lircd

I assumed when you fixed [tickets:38] you would also fix /usr/share/man/man8/lircd.8.xz

Thank you for your time and your patience.

[Alec Leamas]

So, this is basically a duplicate of [tickets:#19]. Closing.

As for your manpage problem, lircd installs a ~5 kb manpage in something like $(DESTDIR)$(prefix)/share/man/man8/lircd.man8. That you havn't isntalled this but rather a very small (empty?) .xz file looks to me like some kind o f problem with your box and/or distribution.

Until Mageia sorts this out, you can always use

$ nroff -man doc/man/lircd.man.8 
$ firefox doc/html/lircd.html

to view the pages in the source directory (after 'make')