cross armhf tangle

[Hal Murray]

>> That was my first try.  It didn't work on BSD.
> Care to share what the failure was?

It didn't know what a struct timeval was.


> Where can I find that man page?

man timeval on this Linux box says:
SYNOPSIS
       #include <sys/time.h>

Without the sys, it worked on Linux but not on BSD.
With the sys it works on everything I've tried so I moved on to other 
things.



> Oh, now I have context.  The only extra code for cross builds would be
> the --march.  When you use --march then /usr/include/sys may not be used
> for <sys/time.h>.  cc swaps sys directory to one approriate to the
> target. 

Are you sure of that?  I don't understand this area.  That's what I was 
expecting, but look in .gitlab-ci.yml in the cross-armhf area.  There are 
a couple of runs that include:
  --cross-cflags "-I/usr/include/ -I/usr/include/python${pyver}"
Isn't that using the same headers?

My guess is that the same kernel header files work for all architectures 
and they are conditioned on #defines that the compiler provides.

/usr/include/bits/timex.h has stuff like this
  # if defined __USE_TIME64_REDIRECTS || (__TIMESIZE == 64 && __WORDSIZE 
== 32)
Where do they come from?

"gcc -dM -E -" tells you the compiler predefines.  397 of them on this box.  But no __TIMESIZE or __WORDSIZE

[After more grepping...]
/usr/include/bits/wordsize.h says
/* Determine the wordsize from the preprocessor defines.  */

#if defined __x86_64__ && !defined __ILP32__
# define __WORDSIZE     64
#else
# define __WORDSIZE     32
...


Similar stuff in /usr/include/bits/timesize.h


-- 
These are my opinions.  I hate spam.