Re: broken `sigblock' defns

[Eli Zaretskii]

On Wed, 31 Oct 2001, Golubev I. N. wrote:

> Version: 21.1
> 
> "syssignal.h" may disregard `sigunblock' being defined by `s/*.h'
> (`s/sco5.h' in particular) and incorrectly define it to nothing.
> 
> However "sysdep.c" defines `sigprocmask_set' var unconditionally,
> "syssignal.h" sometimes does not declare it, even if it is used in
> `sig{un,}block' definitions supplied by `s/*.h'.  This causes
> compilation failure.

Thank you for your report.

However, I'm afraid that I don't understand all the details well
enough to devise a solution.  Please help me fill in the blanks.

I need to know what systems are affected by this problem.  Is SCO 5
the only one, or do you know about others?

What compilation failure did that cause?  What were the error
message(s)?  Could you please post the relevant portions of
the build session transcript?

AFAICS, `sigunblock' is defined to nothing on systems which define
USG, but do not define POSIX_SIGNALS.  (This is done because old USG
systems cannot sigunblock.)  Is it possible that the specific
platform(s) where you saw the problem could define POSIX_SIGNALS in
their s/*.h headers, and avoid messing with this issue altogether?