Re: 3-bigendian-tests.patch

[akim]

On Sun, Aug 12, 2001 at 02:24:06PM +0200, Alexandre Duret-Lutz wrote:
> 
> But this time we use _AT_CHECK_AC_MACRO which perform the same
> as the four macro calls from the first snippet, plus a calls to
> AC_CHECK_ENV.  AC_CHECK_ENV checks that the files present at the
> start of the second snippet are the same as those present at the
> end.  In other words, it checks that AC_C_BIGENDIAN did not
> create a new file (meaning a different endianess result).
> 
> I'll reword the comments to make this clearer before checking in.

Wow.  IMHO this is a bit too smart.  I would rather propose that
the first run creates at-endianness containing explicitly big, little
or whatever is humanly readable, and that the second configure loads
this value and explicitly `test . != . && AC_MSG_ERROR(unexpected
endianness: first run found `...', second found `...').

The result should be understandable by novice or people who did not
fully catch all of the test suite subtleties.

BTW, are you sure it works?  My memory (and the sources :) seem
to reveal that at-* files are considered as by products of the test suite,
hence are irrelevant for AC_STATE_SAVE.

In fact, if you intend to change anything about this issue, I'd like to
ask you to move state- into at-state-something (by be sure to remain
distinguishible (don't know if this is actual English, but you understand
me :) in 8.3).