[valgrind] [Bug 388017] Callgrind produces empty log for pulseaudio

[Josef Weidendorfer]

https://bugs.kde.org/show_bug.cgi?id=388017

--- Comment #4 from Josef Weidendorfer  ---
Good to see that `--trace-children=yes` worked!

Just as remark:

Callgrind opens a file descriptor at start to allow output
to be written to the directory which was the current working
directory at prgoram start. The motivation was that a program
could change to arbitrary directories during run, and it may be
difficult for an user to find the callgrind output.

Now if an application wildly closes all kind of file descriptors
before termination, it also may close this descriptor and leave callgrind
without writing output (this could be detected...).

However, you always can ask callgrind to dump output during the
program run using callgrind_control.

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 388017] Callgrind produces empty log for pulseaudio

[Philippe Waroquiers]

https://bugs.kde.org/show_bug.cgi?id=388017

Philippe Waroquiers  changed:

   What|Removed |Added

 Status|UNCONFIRMED |RESOLVED
 Resolution|--- |INVALID

--- Comment #3 from Philippe Waroquiers  ---
(In reply to Konstantin Kharlamov from comment #2)
> (In reply to Philippe Waroquiers from comment #1)
> > Several things can be tried to get some more info.
> > 
> > * I suppose pulseaudio will daemonize itself.
> >   So, you should run with --trace-children=yes
> >   Also, to be sure that the output goes to a known place, you should use
> >  --log-file=/some/abs/filename%p
> >(you should use %p to have filenames containing the pid)
> > 
> > * Try with other tools, e.g. --tool=none and memchec, and use from a shell
> >the command   'vgdb v.info scheduler'
> >   to check what is happening
> > 
> > * Run with -v -v -v -d -d -d --trace-syscalls=yes --trace-signals=yes
> 
> Amazing, thanks, the `--trace-children=yes` did the trick! I guess, this bug
> can be closed, not sure about the wording — "RESOLVED"? "NOTABUG"?
Closing the bug now as INVALID.

> 
> On an irrelevant note, I also tried running under valgrind and Wine a PE
> binary compiled with MinGW — it resulted in valgrind quitting with these
> lines:
> 
>   preloader: Warning: failed to reserve range 0011-6800
>   preloader: Warning: failed to reserve range 7f00-8200
> 
>   valgrind: m_debuginfo/debuginfo.c:551 (check_CFSI_related_invariants):
> Assertion 'cfsi_fits' failed.
> 
> …and then stacktrace follows. I'm wondering if it's something known or
> should it be reported.
That looks like a bug, yes.
So, better to report it as a new bug, with some more info.
Someone more knowledgeable about wine/windows/... might then jump on it.

E.g. attach the output of running the above with -v -v -v -d -d -d
(i.e. the general debugging of valgrind).

Possibly other debug --trace-cfi=yes --trace-symtab ... might give some hints
but these will produce a lot more info.

Thanks

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 388017] Callgrind produces empty log for pulseaudio

[Konstantin Kharlamov]

https://bugs.kde.org/show_bug.cgi?id=388017

--- Comment #2 from Konstantin Kharlamov  ---
(In reply to Philippe Waroquiers from comment #1)
> Several things can be tried to get some more info.
> 
> * I suppose pulseaudio will daemonize itself.
>   So, you should run with --trace-children=yes
>   Also, to be sure that the output goes to a known place, you should use
>  --log-file=/some/abs/filename%p
>(you should use %p to have filenames containing the pid)
> 
> * Try with other tools, e.g. --tool=none and memchec, and use from a shell
>the command   'vgdb v.info scheduler'
>   to check what is happening
> 
> * Run with -v -v -v -d -d -d --trace-syscalls=yes --trace-signals=yes

Amazing, thanks, the `--trace-children=yes` did the trick! I guess, this bug
can be closed, not sure about the wording — "RESOLVED"? "NOTABUG"?

On an irrelevant note, I also tried running under valgrind and Wine a PE binary
compiled with MinGW — it resulted in valgrind quitting with these lines:

preloader: Warning: failed to reserve range 0011-6800
preloader: Warning: failed to reserve range 7f00-8200

valgrind: m_debuginfo/debuginfo.c:551 (check_CFSI_related_invariants):
Assertion 'cfsi_fits' failed.

…and then stacktrace follows. I'm wondering if it's something known or should
it be reported.

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 388017] Callgrind produces empty log for pulseaudio

[Philippe Waroquiers]

https://bugs.kde.org/show_bug.cgi?id=388017

Philippe Waroquiers  changed:

   What|Removed |Added

 CC||philippe.waroquiers@skynet.
   ||be

--- Comment #1 from Philippe Waroquiers  ---
Several things can be tried to get some more info.

* I suppose pulseaudio will daemonize itself.
  So, you should run with --trace-children=yes
  Also, to be sure that the output goes to a known place, you should use
 --log-file=/some/abs/filename%p
   (you should use %p to have filenames containing the pid)

* Try with other tools, e.g. --tool=none and memchec, and use from a shell
   the command   'vgdb v.info scheduler'
  to check what is happening

* Run with -v -v -v -d -d -d --trace-syscalls=yes --trace-signals=yes

-- 
You are receiving this mail because:
You are watching all bug changes.