Re: [Jack-Devel] jackd-logging -- was jackd2+ardour3 - tracking down "JackActivationCount::Signal value = 0 ref = INT"

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Nedko,

Thanks for all the feedback on the recent emails.

On 07/18/2011 08:30 AM, Nedko Arnaudov wrote:
> Robin Gareus <[hidden]> writes:
>> Maybe it's rather time for a smarter smart-logger :-)
>>
>> xruns are not the only issue. I just managed to fill
>> ~/.log/jack/jackdbus.log with 16MB of 4-line repeats:
>>
[snip]
> 
> Maybe, or have better logging in jack2. Inspecting several old lines on
> each new log line is probably waste of CPU. Do you have a motivation to
> implement such supersmart logger?

The motivation would be that JACK could use the limited manpower better
than re-working the logging mechanism.

I suppose several other projects could benefit from such a "buffering
smart-logger", too. I did not do my homework: maybe a tool/library like
this already exists?!

Waste of CPU.. Maybe. One certainly needs to limit the max number of
lines to buffer and search backwards. A sane value may be 8 and strcmp()
is fast. It bails out after the first mismatched char.

Certainly the bit of CPU overhead can be neglected compared to jackd
flushing 4+ log-lines for each process-cycle to disk otherwise.

but maybe there's a better solution than this pragmatic approach. Any ideas?

> If this error log line sequence is generated in single thread, error
> messages can be collected and then logged as a single line. Not sure how
> viable such solution is in the C++ do-it-right (tm) land. C++ exceptions
> are not universally accepted.

Also the messages are spread out through different source-files and
classes. For debugging it certainly helps to have a trace.

Sure, they could still be collated into a single line for a "simple
smart-logger"; but I estimate this to be more work and more tricky to
get right. or am I missing something?

>> Simply launching a new client (here jack_lsp) after supercollider
>> crashed made the issue go away. no jack-engine re-start needed.
> 
> This indicates issue with jack2 itself. Either genuinely wrong logic
> somewhere or it gets fooled by external events. IMO.

I sent the log-file and info off-list to Stephane.

Cheers!
robin
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iEYEARECAAYFAk4kB7oACgkQeVUk8U+VK0K9hgCfc0ZUEPc8HpH2WkCLTnnKYEih
0osAnRbEc1dlI00zPBkig23Xp4DaeguM
=Lvp3
-----END PGP SIGNATURE-----

1310984147.9418_0.ltw:2,a <4E2407BA.8050804 at gareus dot org>