Re: SIGTRAP in kill emulation on Windows

[Alain Schneble]

Eli Zaretskii <address@hidden> writes:

>> From: Alain Schneble <address@hidden>
>> CC: <address@hidden>
>> Date: Fri, 26 Aug 2016 10:58:58 +0200
>> 
>> Eli Zaretskii <address@hidden> writes:
>> 
>> Yes, debugbreak and the proposed implementation behave identically.
>
> Then I guess that's a limitation we will have to live with.  It
> doesn't sound like a grave one to me: after all, what exactly SIGTRAP
> does when no debugger is attached is not very important, since that is
> not the primary use case for that signal, AFAIU.  If a Lisp program
> wants to kill the process, it doesn't need to use SIGTRAP.

I agree.

>> Yes, and on GNU/Linux I observed that it prints a backtrace prior to
>> termination.
>
> That depends on ulimit and suchlikes, I think: on a GNU/Linux system I
> tried that, the program was simply dumped to the shell prompt without
> printing anything.

Ok, thanks for the hint.

>> But that's kind of irrelevant here I guess.  Where would be the best
>> place to document this quirk (c, signal-process, info manual)?
>
> Actually, perhaps not even there.  In NEWS might be enough, I think.

Ok, I'll arrange a new patch with that added to the NEWS entry.

>> FWIW, there is probably something we could do about it -- query if the
>> process in question is attached to a debugger.  If not, we could
>> terminate it.
>
> That's for the application to decide, IMO.  Signal delivery is too
> low-level to replace one signal with another one.

That's a good point.  I agree with you.