This is the mail archive of the
mailing list for the Cygwin project.
Re: Signal handling tune up.
On Tue, Aug 19, 2003 at 10:34:22PM -0400, Pierre A. Humblet wrote:
>I got blocked from the list for the second time today,
>so I changed my address. Mail sent to the old one will
>still get to me.
What a pain. It's been a horrible day. I've been trying to do my
real job while attempting to deal with the massive flood of email resulting
from that (&*! virus. I don't know if the virus is causing your problems
or not but I suspect it is. I have to change the subscription mechanism
on sources.redhat.com so that it ignores virus-like email.
>At 09:16 PM 8/19/2003 -0400, Christopher Faylor wrote:
>>I think we're talking about different things. The only thing I did was
>>move the existing code out of setup_handler and into wait_sig. There
>>should be very little functional difference in doing this.
>Right. You did 50% of that particular change and I got worried. I will
>try to relax.
I did a little more just now but it is still not right. I was going to
check in your recursive elimination patch to sigdelayed but noticed that
my simple test case to check how recursion worked doesn't work. Sigh.
I should just go to bed.
>>>>Not necessarily. It depends on who is calling sig_dispatch_pending. An
>>>>outer sigframe user wins. This guarantees that signals will be
>>>>dispatched in sig_dispatch_pending.
>>>Thanks, but it's still greek to me. What is an "outer sigframe user".
>>Someone who calls sigframe earlier in the call stack.
>Still desperately trying to understand (I should stick to bicycles,
>sig_dispatch_pending builds a sigframe. It will be found by
>setup_handler, which will call interrupt_on_return,
>which will spoof the return address and call interrupt_setup to
>If nothing special is done, the handler will start when
>By calling sigframe::call_signal_handler () the handler gets
>called just before the return. What's the gain?
sig_dispatch_pending is supposed to guarantee that signals have been
If you do something like:
then the signal dispatch will happen when foo returns, not when
sig_dispatch_pending returns. The goal is that, in most cases, the
function closest to the user should be the one that gets "interrupted".