Re: Callbacks to Lisp from C (and SIGSEGV handling)

Necro'ing this thread...

I have been keeping an eye on releases for the removal of SIGSEGV from the
GC mechanism, but have not noticed any such?

Is it still on the timeline?

TIA,
Gareth

On Sat, Mar 22, 2025, 2:48 p.m. Douglas Katzman <[email protected]> wrote:

> There is restore_sbcl_signals() in the C runtime.  But using it sounds
> like a bad idea to me. Since signal handlers are not per-thread, there's no
> way to make this reliable. An SBCL-internal thread (of which there is at
> least one, but could be more) could do an operation that receives sigsegv
> while some random thread called the C thing that changed the handler. The
> SBCL thread will probably react badly to having the wrong handler called.
> On a more positive note, we are almost at the point where there is no need
> for sigsegv. Most of the so-called dynamic space does not use it.  I have
> to remove sigsegv for tracking writes to slots of symbols, and there's not
> a lot else. (One other object type, falling under "not a lot else").  Maybe
> by the next SBCL release (April, not the pending one), I can complete the
> work needed to confine use of sigsegv for actual faults and not "normal"
> activity.
>

_______________________________________________
Sbcl-help mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sbcl-help