Re: support for universal binaries on MacOS X (5/6)

[Bruno Haible]

Hi Jim,

> That looks like it'd work fine

It does work fine. I tested it :-)

> but I'd like it even more if it were 
> to set the cache variable to "no" *before* the use of AC_CACHE_CHECK.
> Then, this change would be more compact and would not affect the logical
> structure (i.e. no need for the else clause here

Well, the logical structure here is that we have a general case (the
AC_TRY_RUN code) and then one or several platform specific special cases.
Now there's only APPLE_UNIVERSAL_BUILD but I can well foresee that there
will be others, such as WINDOWS_UNIVERSAL_BUILD or AIX_UNIVERSAL_BUILD.
(The latter nearly already exists: Shared libraries on 64-bit AIX systems
contain a 32-bit .o file and a 64-bit .o file together. Someday we should
extend libtool to support this...) Therefore using the value of one of
these special cases as the default for all platforms is inappropriate.

> and no need for the indentation change in the nanosleep test).

The indentation change in the nanosleep test is because of the 'if',
which won't go away either way.

OK to commit the patch as proposed? Or should I undo the reindentation?

Bruno