bug#12471: Avoid some signal-handling races, and simplify.

[Paul Eggert]

On 09/22/2012 01:02 AM, Eli Zaretskii wrote:

> What you suggest now is that I add to 'sys_kill' a lot of stuff for it to
> be able to replace 'raise'

Sorry, I don't follow.  I did not suggest that, at any rate.

Currently, emacs.c does the equivalent of this:

  kill (getpid (), sig);

which on Windows is macro-expanded to

  sys_kill (_getpid (), sig);

The proposed patch causes Emacs to do the
same Windows actions via the redefined 'raise'.  That is,
emacs.c will do 'raise (sig)', and the implementation
of 'raise (sig)' will do 'sys_kill (_getpid (), sig)'.

If there is a problem with the proposed patch with respect
to Emacs sending a signal to itself, why doesn't the
current Emacs trunk have the same problem?  And if so,
shouldn't that problem be addressed independently, regardless
of the patch proposed for Bug#12471?

> I will always prefer a small burden to a larger one.

I was referring to the overall burden, for all maintainers,
not just the burden for the Windows maintainers in particular.