sectionsd: segmentation fault

[JOCKYW2001]

I compiled a YADD from this morning's CVS.
Following appears in the bootlog when neutrino is selected:

Code: Alles auswählen

$Id: sectionsd.cpp,v 1.171 2004/02/25 00:28:12 thegoodguy Exp $
caching 504 hours
events are old 60min after their end time
Segmentation fault

I also tried it with sectionsd.cpp v1.170, same segmentation fault.

Then I tried booting enigma, it also crashes with a segmentation fault. Don't know if both problems are related but it looks like it.

I have no clue how to further diagnose or even fix this.

[rasc]

There were changes on the API.
Ensure that all applications and tuxbox libraries were re-compiled.

[JOCKYW2001]

Hmm, did a complete checkout not more than 13 hours ago ..
Anyhow, some minutes ago I checked out again on another PC and now there is this when running make all:

make: *** No rule to make target `directories', needed by `core'. Stop.

Update: That's because 2 hours ago Waldi changed cdk/Makefile.am I checked out previous version and it's building all. Let's see if it works in a couple of hours

[Sat_Man]

Hmm, did a complete checkout not more than 13 hours ago ..
Anyhow, some minutes ago I checked out again on another PC and now there is this when running make all:

make: *** No rule to make target `directories', needed by `core'. Stop.

Update: That's because 2 hours ago Waldi changed cdk/Makefile.am I checked out previous version and it's building all. Let's see if it works in a couple of hours

http://cvs.tuxbox-cvs.sourceforge.net/l ... 00526.html

[JOCKYW2001]

Yes, that make problem is solved.

But the segmentation faults are still there
Both in sectionsd when booting neutrino and also when booting enigma (system goes down right after it finds STV6412)

My build procedure in short:
1. checkout CVS completely
2. autogen.sh & configure with flash rules
3. make all
4. autogen.sh & configure *without* flashrules
5. rebuild u-boot & kernel-cdk (make core; make boot)

[JOCKYW2001]

So I just tried it again. Started from scratch and checked out CVS. Compiled it and tried the YADD on a Philips. Not my own Philips, but this time from a friend. Enigma boots fine now, but if you boot into Neutrino there is still this shitty segmentation fault. See the bootlog below. If you don't look at the bootlog you will not be aware that something is wrong since the box boots up and seems to work normal.

I can reproduce the error by telnetting to the box, kill sectionsd and restart sectionsd.

Would someone be so kind and try the same?

Please press Enter to activate this console. [LCDFONT] initializing core...
[LCDFONT] adding font /share/fonts/micron.ttf...OK (Micron/Regular)
[LCDFONT] Intializing font cache...
[LCDFONT] FTC_Face_Requester (Micron/Regular)
$Id: sectionsd.cpp,v 1.171 2004/02/25 00:28:12 thegoodguy Exp $
caching 504 hours
events are old 60min after their end time
Segmentation fault
[ConfigFile] Unable to open file /var/tuxbox/config/timerd.conf for reading.
[camd] ca system id: 1702
$Id: zapit.cpp,v 1.344 2004/02/25 10:45:35 thegoodguy Exp $
/var/tuxbox/config/zapit/services.xml: No such file or directory
$Id: controld.cpp,v 1.115 2004/02/19 23:00:33 zwen Exp $

[controld] Boxtype detected: (2, Philips D-BOX2)
[nhttpd] Neutrino HTTP-Server starting..
[neutrino] frameBuffer Instance created
812k video mem
avia_gt_gv: set_input_mode (mode=2)
avia_gt_gv: set_input_size (width=720, height=576)
[lcdd] time-skin not found -> using default...
[lcdd] weekday-skin not found -> using default...
[lcdd] date-skin not found -> using default...
[lcdd] month-skin not found -> using default...
[LCDFONT] initializing core...
[LCDFONT] adding font /share/fonts/micron.ttf...OK (Micron/Regular)
[LCDFONT] Intializing font cache...
[LCDFONT] FTC_Face_Requester (Micron/Regular)

[Npq]

Yep, same Revision, no problems.

Guess it's your turn now to whip out the almighty gdb

[JOCKYW2001]

Yes, I'm afraid so

Thanks for testing, with what box did you test it?

[Npq]

Sagem Sat 1xI, eNX/600

[JOCKYW2001]

Finally I found what is causing the problem. It is sectionsd. There is nothing wrong when it is compiled *without* flashrules (1.7mB), but when it is compiled *with* flashrules (195kB) it produces the segmentation fault.

I hope Ralf Gandy can fix this, cuz' I can't.

Edit:
For those who do not believe this and do not want to build from CVS: download and flash latest snapshot from http://dboxupdate.berlios.de/
Then telnet into the box, "killall -9 sectionsd", "/bin/sectionsd" and voila:

Code: Alles auswählen

/bin > sectionsd
$Id: sectionsd.cpp,v 1.171 2004/02/25 00:28:12 thegoodguy Exp $
caching 504 hours
events are old 60min after their end time
Segmentation fault

[Soli]

Ich habe das gleiche Problem
CVS Stand 05.03.2004

$Id: sectionsd.cpp,v 1.171 2004/02/25 00:28:12 thegoodguy Exp $
caching 504 hours
events are old 60min after their end time
Segmentation fault

Das war früher nicht.

Habe ne Sagem 2x Avia600

[HELLz]

ich hab das auch...
was für auswirkungen hat das denn überhaupt, denn soweit ich das bis jetzt beurteilen kann läuft mein image als solches fehlerfrei, nur ebend dieses bekackte
"Segmentation fault"

p.s.: sagem 1x sat

[obi]

looks like a bug in g++-3.3.3's optimizer when used with -Os. Try g++ 3.2.3 or use gdb with an unstripped binary to produce a backtrace.

[Npq]

The crash occurs after the parent forks when on its way out of the door the destructors for the two static demux-objects are called.

It seems to be the STL-Routines responsible for cleaning up the s_filters-vector that get "overoptimized", maybe the crash occurs because the vector is still empty at the time, I don't know.

It doesn't harm the operation of the sectionsd (the parent was going to finish anyway).

Hope this doesn't cause other nasty bugs in the flash image.