Strange error on NetBSD

[Fred Wright]

On Fri, 21 Jun 2024, Hal Murray via devel wrote:

> Does anybody other than me run/test on NetBSD?

Ocasionally.  I have a NetBSD 6.1.5 VM.

> The test-all part of option-tester gets the following error.

Why only in option-tester?

> ntp_stdlib includes signal.h
> Do we have to do some magic to make that get siginfo_t?
>
> [ 43/136] Compiling libparse/clk_sel240x.c
> In file included from ../../include/ntp.h:15,
>                 from ../../libparse/clk_sel240x.c:10:
> ../../include/ntp_stdlib.h:118:53: error: unknown type name
> \xe2\x80\x98siginfo_
> t\xe2\x80\x99
>  118 | extern void signal_no_reset1(int, void (*func)(int, siginfo_t *,
> void *)
> );

Poke around with "grep -rInw <something> /usr/include" to see how it 
should get defined.

On Sat, 22 Jun 2024, Hal Murray via devel wrote:

> The NetBSD man page says:
>     The sigaction() function conforms to IEEE Std 1003.1-1990 ("POSIX.1").
>
> The Linux man page says:
>   Feature Test Macro Requirements for glibc (see feature_test_macros(7)):
>
>       sigaction():
>           _POSIX_C_SOURCE
>
>       siginfo_t:
>           _POSIX_C_SOURCE >= 199309L
>
> wscript has a comment discussing this area then does:
>      ctx.env.CFLAGS = ["-std=c99", "-D_GNU_SOURCE"] + ctx.env.CFLAGS
>
> It looks to me like c99 should be >= 1990, but I don't know much about
> that stuff.

Options like "-std=c99" only determine the language standard, not the OS 
or POSIX version.  They may affect headers, but only for headers 
considered part of the language, rather than part of the OS (e.g. 
stdint.h).

Fred Wright