Re: return values of test programs in *.m4 macros

[Ralf Wildenhues]

Hello,

* Bruno Haible wrote on Sun, Dec 05, 2010 at 02:12:49AM CET:
> > If the particular
> > combination of failures matches 63, 77, or 99 (as commonly used by
> > automake and autoconf to mean version mismatch, skip, or hard fail),
> > then the configure script might misbehave.
> 
> In the generated configure scripts, AC_RUN_IFELSE tests for an exit
> code equal to 0 and nothing else. And this cannot change, because it's
> documented behaviour of AC_RUN_IFELSE:
> 
>  -- Macro: AC_RUN_IFELSE (INPUT, [ACTION-IF-TRUE], [ACTION-IF-FALSE],
>           [ACTION-IF-CROSS-COMPILING])
>      If PROGRAM compiles and links successfully and returns an exit
>      status of 0 when executed, run shell commands ACTION-IF-TRUE.
>      Otherwise, run shell commands ACTION-IF-FALSE.
> 
> So, there is no problem now, and there cannot be a problem in the
> future.

Well, I think what Eric was hinting at was that, for example, some tests
in Autoconf's and Automake's own test suites interpret an exit status of
77 from a configure execution within that test run to infer that the
test should be skipped.

I don't think that the current gnulib testsuite invokes configure
programs from within, and I don't think Autoconf's nor Automake's test
suites currently use gnulib macros, but it wouldn't be a big leap to
think that some gnulib-using packages interpreted an exit from configure
with status 77 to infer some hint about skipping.

IMVHO a NEWS entry should be good enough to warn about this possible
semantic change, however.

Cheers,
Ralf