inn-CURRENT: three building bugs

[Russ Allbery]

Heiko Schlichting <inn-bugs at FU-Berlin.DE> writes:

> trying to build inn-CURRENT-20020423 on an SGI system running IRIX 6.5.15f
> using MIPSpro Compiler 7.3.1.3 failed for the following reasons:

> 1) configure uses inn_cv_c99_vamacros instead of inn_cv_c_c99_vamacros
>    which probably is a typo.

>    Attachment 1 contains a patch to configure.in to correct this problem.

Thanks, applied.

>    BTW: it would be nice if configure.in could be made compatible 
>         with autoconf 2.53. At this time the older version 2.13 is needed.

The Autoconf people did a butcher job on backward compatibility, which
rather annoys me.  Because of that, I've not yet installed 2.53 for myself
because I'm going to have to install it under another name; it's not
usable for the majority of packages.  At some point I'll get around to
that, although doing the real conversion to 2.53 idioms isn't scheduled
until 2.5.

> 2) support/mksystem does not work as expected because awk tries to read
>    from stdio and not the '<<' input. It also does not honor the awk path
>    specified in Makefile.global.

That behavior of awk is stunningly broken.

>    Attachment 2 contains a patch for support/mksystem and include/Makefile
>    to correct the problem and it works on my system. I don't know how
>    portable it really is but I assume it cause much less problems than the
>    current code.

It uses a multiline single-quoted string, which is the reason why I didn't
do it that way in the first place.  I'm pretty sure that we're eventually
going to break the size limit for command lines somewhere.  The old method
really should have worked portably, but IRIX's userland tools are junk.

Anyway, what I'll probably do eventually is rewrite it as a regular awk
script, and your patch will work in the meantime.  Thanks, applied.

> 3) storage/buffindexed/buffindexed.c can't be compiled because of error
>    messages: "The expression must be a pointer to a complete object type."
>    The reason seems to be calculating with void-pointers which is not
>    allowed. Casting the variables to char-pointers fixes the problem and
>    should not cause any harm.

>    Attachment 4 contains a patch for storage/buffindexed/buffindexed.c

Thanks, applied.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>