Use of `M_PI` in `stresstest.c` is not C99 and unfortunately breaks compilation with musl libc

[Sebastian Pipping]

# make -j16 
cc -O2 -march=x86-64 -pipe -pipe -frecord-gcc-switches -fno-diagnostics-color -fmessage-length=0 -Wall -Wextra `pkg-config --cflags ncursesw` -Wl,-O1 -Wl,--as-needed -Wl,--defsym=__gentoo_check_ldflags__=0 ttyplot.c `pkg-config --libs ncursesw` -lm -o ttyplot
cc -O2 -march=x86-64 -pipe -pipe -frecord-gcc-switches -fno-diagnostics-color -fmessage-length=0 -Wall -Wextra `pkg-config --cflags ncursesw` -Wl,-O1 -Wl,--as-needed -Wl,--defsym=__gentoo_check_ldflags__=0 stresstest.c `pkg-config --libs ncursesw` -lm -o stresstest
stresstest.c: In function 'main':
stresstest.c:74:69: error: 'M_PI' undeclared (first use in this function)
   74 | buffer_pos += sprintf(buffer + buffer_pos, "%.1f\n", (sin(n*M_PI/180)*5)+5);
      |                                                             ^~~~

stresstest.c:74:69: note: each undeclared identifier is reported only once for each function it appears in
stresstest.c:87:17: warning: implicit declaration of function 'usleep'; did you mean 'sleep'? [-Wimplicit-function-declaration[https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#index-Wimplicit-function-declaration]]
   87 |                 usleep(50);  // let ttyplot read this before proceeding
      |                 ^~~~~~
      |                 sleep

Also note the usleep issue further down.

Origin: https://bugs.gentoo.org/922285

Besides a fix, we could add CI using Docker with an image of Alpine Linux to cover compilation with musl, if that sounds acceptable. It would help protect against regressions once this is fixed.

[Edgar Bonet]

I suggest defining _XOPEN_SOURCE=500 in stresstest.c, as suggested by the glibc documentation. The Musl libc provides M_PI when this macro is defined, irrespective of its value.

[Sebastian Pipping]

I suggest defining _XOPEN_SOURCE=500 in stresstest.c, as suggested by the glibc documentation. The Musl libc provides M_PI when this macro is defined, irrespective of its value.

@edgar-bonet that's an interesting find and idea. I would want to test out this precise addition to the very top of stresstest.c then:

#if !defined(_XOPEN_SOURCE) || (_XOPEN_SOURCE < 500)
# define _XOPEN_SOURCE 500
#endif

Besides a fix, we could add CI using Docker with an image of Alpine Linux to cover compilation with Musl, if that sounds acceptable. It would help protect against regressions once this is fixed.

@edgar-bonet @tenox7 how do you feel about this direction in general?

[Edgar Bonet]

In order to avoid a “redefined” warning, I would undefine _XOPEN_SOURCE if it has an unsuitable value:

#if _XOPEN_SOURCE < 500
# undef _XOPEN_SOURCE
#endif
#ifndef _XOPEN_SOURCE
# define _XOPEN_SOURCE 500
#endif

Note that this is safe even if _XOPEN_SOURCE is initially undefined:

• an undefined macro evaluates as zero within a #if
• undefining an undefined macro is a no-op.

how do you feel about [a CI with Musl/Alpine/Docker] in general?

Great! I am not that comfortable with Docker, but I appreciate the great job you are doing on the CI front.

[Sebastian Pipping]

In order to avoid a “redefined” warning, I would undefine _XOPEN_SOURCE if it has an unsuitable value:

#if _XOPEN_SOURCE < 500
# undef _XOPEN_SOURCE
#endif
#ifndef _XOPEN_SOURCE
# define _XOPEN_SOURCE 500
#endif

Note that this is safe even if _XOPEN_SOURCE is initially undefined:

• an undefined macro evaluates as zero within a #if
• undefining an undefined macro is a no-op.

@edgar-bonet would you be okay with this alternative?:

#if !defined(_XOPEN_SOURCE) || (_XOPEN_SOURCE < 500)
# undef _XOPEN_SOURCE  // to address warnings about re-definition
# define _XOPEN_SOURCE 500
#endif

It seems to play by the rules you described and I would find it more "contained" and self-explaining, a bit easier to look at, but that's potentially all subjective. What do you think?

how do you feel about [a CI with Musl/Alpine/Docker] in general?

Great! I am not that comfortable with Docker, but I appreciate the great job you are doing on the CI front.

Thanks!

[Edgar Bonet]

Yes, that way of (re)defining _XOPEN_SOURCE looks good to me. I would even be happy with the “magic number” version:

#ifndef M_PI
# define M_PI 3.1415926535897932
#endif