Re: configure problem on HP-UX 11

[scott douglass]

At 12:48 PM 29-06-02 -0600, Bob Proulx wrote:
>> and since the test doesn't work properly, it's likely that it could
>> be improved.
>
>The original poster reported that the test failed with the HP native
>ANSI C compiler version B.11.11.04.  I tested this on my system which
>has B.11.11.02, only slightly different, and the supplied configure
>test worked fine for me.  I don't have the patch to go from .02 to .04
>or I would apply it and check on the same system that the original
>poster was running.  If I can locate an identical system I will try to
>test and recreate the problem.
>
>The check seems okay.  I am suspicious that something has been broken
>on a specific particular system and not in general.

I may not have made myself clear.  Or perhaps I'm misunderstanding you.

The orginal test did (correctly) identify that the HP compiler understands 
'inline'.  The problem is that the HP compiler fails to parse some uses of 
inlines and typedefs.  I suggested that the test case be changed to one of 
these failing cases so that inline would not be used with the HP compiler so 
that the failing cases would compile.

At any rate Kevin Ryde had already made almost exactly the change I suggested.  
[But I don't think it's released yet and I don't know how often gcc/bash/etc 
update their autoconf's.]

Do you mean that B.11.11.02 compiles this without error?

    typedef unsigned nls_uint32;
    static inline nls_uint32 static_foo () {return 0; }

If so there must be a bug introduced between B.11.11.02 and B.11.11.04.

>It would be useful to have specific test cases to run so that details
>of the problem could be revealed.

I'd be happy to compile a few small test cases for you with B.11.11.04.