Stefan Hajnoczi <stefa...@redhat.com> writes: > On Fri, May 05, 2017 at 04:59:58PM +0100, Alex Bennée wrote: >> >> Paolo Bonzini <pbonz...@redhat.com> writes: >> >> > On 05/05/2017 12:38, Alex Bennée wrote: >> >> Signed-off-by: Alex Bennée <alex.ben...@linaro.org> >> > >> > Can you look at the patch I just sent a pull request for? It only has >> > locked and unlocked trace events, you can add lock on top. >> >> Cool - great minds think alike ;-) >> >> I'll re-spin my trace analysis script and the lock trace point once that >> pull request is merged. I would be nice if we could associate locks with >> names though as the QemuMutex * is basically just an anonymous handle. >> Would it be overly extravagant to add a const char * to QemuMutex to can >> be init'ed with a human readable name? >> >> Stefan, >> >> Does the trace sub-system have any way to register a human readable >> string to a given pointer value? I guess this is something that could be >> added to the trace pre-amble? > > No, it doesn't. I would make the trace event take const char * and pass > in the string.
It would be nice to avoid having the string for non-trace builds. I was thinking of something like: trace_register_human_name(void *ptr, const char *name) Which compiles away to nothing when tracing is not enabled. > > Stefan -- Alex Bennée
