Re: Inline detection problem

[Akim Demaille]

>>>>> "Daniel" == Daniel Sands <address@hidden> writes:

Daniel> I recently ran into a problem where one building package
Daniel> detects "inline" and then does a recursive configure on
Daniel> another package if it exists in the same source tree, which
Daniel> also detects "inline" but uses the same cache file.  The
Daniel> problem is that the "inside" package uses gcc -ansi, which now
Daniel> removes the old "inline", but it was already detected as
Daniel> "inline" and cached into the other one.  So is there a reason
Daniel> to "do nothing" if "inline" works in one case, when ANSI
Daniel> removes this definition?

This is a bug in the use of configure.  If you share the cache, you
are asking for troubles.  Just don't, or keep things consistent.

Autoconf 2.50 should do a better job at this, and reject the cache for
the slave configure.