Therein lies the problem. How do you identify which tests are standard,
Standard tests are those generated by Autoconf macros without
customization. You said that sometimes various packages use the same
name for different tests. That won't happen with standard test names,
because using a standard name for a nonstandard test is likely to fail
in other ways. Doesn't Autoconf have a naming convention to avoid
that?
vs. those tests that are dependent on compiler options currently in
effect or previously detected libraries in effect,
My previous suggestion was to record the options in the cache, so as
to reuse a cached result only when it is appropriate. There could thus
be several cached results for one test, with different options.
Maybe this could be done also with previously detected libraries.
Alternatively, it may be possible to determine a list of tests which
in practice depend on other libraries, and exclude them from caching,
so that we could cache teh rest.
when each package's
choice of earlier actions before running the same standard test may have
led to different results when run without the cache?
Is it possible to find bounds on which results can alter which results?
--
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
Use Ekiga or an ordinary phone call
_______________________________________________
Autoconf mailing list
Autoconf@xxxxxxx
https://lists.gnu.org/mailman/listinfo/autoconf
