Re: Re: Possible explanation for readline configuration problems

Peter Eisentraut <peter_e@gmx.net> writes:
> (b) the readline dudes commented out (#if 0) the declaration of
> filename_completion_function in the header in favour of the new
> rl_filename_completion_function (but left the symbol in the library).

The "#if 0" silliness is not what's confusing configure however,
because what it's grepping is preprocessor output; it will not see the
commented-out declaration of filename_completion_function.  What it
*does* see is the declaration of rl_filename_completion_function, and
because of the lack of defense against partial-word matches, mistakes
that for filename_completion_function.  Our code would actually work if
AC_EGREP_HEADER weren't so nonrobust about what it will take as a match.

> Needless to say, the readline developers deserve to be shot for doing this
> in a minor release.

I don't suppose our life would be any simpler if they'd called it 5.0
instead of 4.2, though.  Visions of mayhem aside, we ought to try to
persuade our code to work cleanly with 4.2 as well as earlier releases.
Do you have time to work on that for 7.1.1 (ie, end of the month or so)?
        regards, tom lane