Re: about gawk |& operator and getline on HPUX11

[Paul Eggert]

> From: Stepan Kasal <address@hidden>
> Date: Mon, 27 May 2002 11:20:45 +0200
> 
> Thus I think that the attempt to autodetect the bug not only doesn't
> fit to gawk but that it doesn't fit to autoconf either.

It's a judgment call.  If the bug is common and can be detected easily
at configure-time, then we might as well add it to Autoconf.  But if
if the bug is rare and cannot reliably be detected at configure-time,
then perhaps it's better to use a run-time test, or failing that to
tell users to get a working system.

This seems to be the latter case.

It's easy and cheap to test for (0 < st_blksize && st_blksize <=
PTRDIFF_MAX) as a sanity check, since no block size should ever be out
of that range.  Diffutils does that, so it's a reasonable precedent;
it might be good enough in practice, since nobody is reporting this
sort of problem about diffutils on HP-UX hosts.