Hmm... good question. My guess is that if the executable is compiled
*without* the "-fomit-frame-pointer" option (which I guess is the case for
darktable), the "backtrace()" call should be really fast. In any case I'll
try to measure it and report back.
On Mon, Oct 6, 2014 at 1:20 PM, Tobias Ellinghaus <h...@gmx.de> wrote:
> Am Montag, 6. Oktober 2014, 00:08:23 schrieb Fernando:
> > So... I was investigating a mutex inter-lock situation and found it
> really
> > useful to add a new member to the "dt_pthread_mutex_t" structure to hold
> > the backtrace responsible for the latest successful lock operation.
> >
> > For this, you just need to do this:
> >
> > typedef struct dt_pthread_mutex_t
> > {
> > ....
> > void *last_lock_bt[10]; // <-- New member
> > }
> >
> > And then add this other line to both the
> > "dt_pthread_mutex_lock_with_caller" and
> > "dt_pthread_mutex_trylock_with_caller" functions:
> >
> > backtrace(mutex->last_lock_bt, 10);
> >
> > Thanks to this, whenever you arrive to a situation where a thread is
> > forever locked waiting for a mutex, you can use gdb to print the mutex
> > structure contents and find out WHO (and WHERE) took the mutex first.
> >
> > Of course, all of this only takes place when using the DEBUG build...
> >
> > Do you think it would be a good idea to add this to "upstream"?
> >
> > By the way, the problem I was debugging is
> > http://www.darktable.org/redmine/issues/10131 (in case you want to see
> how
> > this mutex backtrace would come handly). Oh! And also, please have a look
> > at the redmine bug description anyway, as right now any help would be
> > greatly appreciated. Thanks!
>
> That sounds useful indeed. Do you know how badly that affects the speed?
> Maybe
> it should be kept in for regular builds, too, as DEBUG is kind of useless
> to
> find real problems as it changes so many code parts that I am not sure if
> it
> couldn't confuse more than it helps (I never ever compile that way).
>
> Tobias
>
> ------------------------------------------------------------------------------
> Slashdot TV. Videos for Nerds. Stuff that Matters.
>
> http://pubads.g.doubleclick.net/gampad/clk?id=160591471&iu=/4140/ostg.clktrk
> _______________________________________________
> darktable-devel mailing list
> darktable-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/darktable-devel
>
>
------------------------------------------------------------------------------
Slashdot TV. Videos for Nerds. Stuff that Matters.
http://pubads.g.doubleclick.net/gampad/clk?id=160591471&iu=/4140/ostg.clktrk
_______________________________________________
darktable-devel mailing list
darktable-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/darktable-devel
