Re: Elisp native profiler

[Stefan Monnier]

> But there are races if the SIGPROF handler runs in
> some other thread, no?

I guess there are races if the main thread is not suspended while the
handler runs, indeed (mostly because we try and read backtrace_list from
the handler which is allocated on the main thread's stack).

> That bug could be worked around by using thread-specific timers (which
> are available in recent GNU/Linux but not older versions or in other
> POSIXish platforms).

That sounds good, yes.  I don't think it's terribly important for this
profiler to work in "unusual" systems.

> A lesser question: why does the profiler use SA_SIGINFO?

No idea.  Better ask the author (Tomohiro Matsuyama <address@hidden>).


        Stefan