maemo.org - Talk - [ANNOUNCE] CSSU-thumb thread

@yaliang: but the labels aren't translated at all... It reminds me an old problem with libc6 when I was trying to update it. This broke locales as well then.

@cssu devs: Ok, so we have a second locale based problem. Can we safely install previous locales, are they in the repos and what are the instructions to check whether it's the locales update to blame.

pali	2014-05-16 16:04

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900

Quote:

Originally Posted by pali (Post 1424769)

Crashing with last version from cssu-devel and also with last version with disabled ipv6 support.

It crashing randomly at any flac file which I tried to played. Sometimes it does not crash for long time and sometimes 2-3x in 5 minutes...

For few days I tested last OMP with and without last libc6. Conclusion: with last thumb libc6 PA often crash when playing any file (not only flac, mp3 or ogg too). Without thumb libc, there was no crash.

When I was testing I had connected bluetooth headset in A2DP mode. And I got about 1 PA crash per hour.

freemangordon

2014-05-18 07:58

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900

Quote:

Originally Posted by pali (Post 1425642)

Well, it seems that I have no option, but to revert to ARM-compiled glibc. Esp that I am unable to reproduce those crashes(and I tried hard, with both HF and BT Power amp receiver), so no way to fix them.

Will try today to fix h-d as well and to issue a new update.

pali	2014-05-18 08:19

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900

@freemangordon: maybe... can you try to compile libc6 again (with totally clean scratchbox)? maybe there could be bug in gcc/scratchbox/linker which cannot be easily detected...

And what makes even worse is that I have never saw that PA error:

Code:

May 7 19:07:20 Pali-Nokia-N900 pulseaudio[2732]: mainloop.c: poll(): Bad address

Maybe different behaviour in poll kernel system call? Btw, I tested only with kp52+git and kp53 versions. But I doubt that there is something between kp52 and kp53 which could change/break poll().

pali	2014-05-18 08:25

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900

Here is relevant PA code which starting crash log:

Code:



int pa_mainloop_poll(pa_mainloop *m) {

    pa_assert(m);

    pa_assert(m->state == STATE_PREPARED);



    if (m->quit)

        goto quit;



    m->state = STATE_POLLING;



    if (m->n_enabled_defer_events )

        m->poll_func_ret = 0;

    else {

        pa_assert(!m->rebuild_pollfds);



        if (m->poll_func)

            m->poll_func_ret = m->poll_func(m->pollfds, m->n_pollfds, m->prepared_timeout, m->poll_func_userdata);

        else

            m->poll_func_ret = poll(m->pollfds, m->n_pollfds, m->prepared_timeout);



        if (m->poll_func_ret < 0) {

            if (errno == EINTR)

                m->poll_func_ret = 0;

            else

                pa_log("poll(): %s", pa_cstrerror(errno));

        }

    }



    m->state = m->poll_func_ret < 0 ? STATE_PASSIVE : STATE_POLLED;

    return m->poll_func_ret;



quit:

    m->state = STATE_QUIT;

    return -2;

}

It looks like PA could have some poll() wrapper (no idea if maemo using it or not).

String Bad address is written when poll() returns EFAULT.

poll() manpage:

Code:

EFAULT The array given as argument was not contained in the calling program's address space.

So this looks like as there is memory corruption in PA.

Any idea how it is possible to debug? How PA can pass invalid pointer?

freemangordon

2014-05-18 09:28

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900

Quote:

Originally Posted by pali (Post 1425779)

Code:

May 7 19:07:20 Pali-Nokia-N900 pulseaudio[2732]: mainloop.c: poll(): Bad address

Maybe different behaviour in poll kernel system call? Btw, I tested only with kp52+git and kp53 versions. But I doubt that there is something between kp52 and kp53 which could change/break poll().

For sure there is some bug in gcc, see https://gitorious.org/community-ssu/...ad0cf41410ae23 , lines 39-40, without that patch locales does not work at all.

Unfortunately it seems we're hitting something else, and given that problems from thumb glibc prevail on the benefits (which are some couple of hundred KiB), I'd rather not waste time on chasing gcc/PA/glibc bugs.

freemangordon

2014-05-18 09:30

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900

Quote:

Originally Posted by pali (Post 1425780)

Here is relevant PA code which starting crash log:

Code:



...

}

It looks like PA could have some poll() wrapper (no idea if maemo using it or not).

String Bad address is written when poll() returns EFAULT.

poll() manpage:

Code:

EFAULT The array given as argument was not contained in the calling program's address space.

So this looks like as there is memory corruption in PA.

Any idea how it is possible to debug? How PA can pass invalid pointer?

Besides setting a breakpoint, I can't think of anything else.

But honestly, it does not worth the effort, I'll simply revert to ARM glibc. We'd better focus on other stuff than waste time on that.

freemangordon

2014-05-19 05:14

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900

Quote:

Originally Posted by Android_808 (Post 1424500)

i noticed a small issue.
lock orientation to landscape with keyboard closed.
open phone app. display switches automatically to portrait.
switch to menu, goes back to landscape.
open gallery, opens in landscape.
view any image fullscreen. switches to portrait whilst in landscape still.
going back to picture list is now in portrait but with list only covering half of the screen.

whilst viewing images fullscreen, turning to portrait causes image to display as in landscape.

http://talk.maemo.org/showpost.php?p...&postcount=383

Please check if that fixes the issue for you

pali	2014-05-19 07:21

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900

@freemangordon: same PA error occures when I call: sudo stop pulseaudio

Can you try to use thumb libc6 and last PA from cssu-devel and try to stop pulseaudio? I would like to see what happen in syslog...

All times are GMT. The time now is 04:01.

Page 209 of 227

209

Show 40 post(s) from this thread on one page

vBulletin® Version 3.8.8