Re: about gawk |& operator and getline on HPUX11

[Stepan Kasal]

Hallo Paul, Aharon, Xiang Zhao and all readers of gnu.utils.bug!

> On Fri, May 24, 2002 at 10:53:19AM -0700, Paul Eggert wrote:
> > Wouldn't it be better to fix Autoconf's AC_STRUCT_ST_BLKSIZE so that
> > it reports that st_blksize does not exist on the buggy platforms?

On Sat, May 25, 2002 at 06:02:46PM +0200, I wrote:
> 1) Yes, that would be nice.  I'll do this, eventually, and ask Zhao Xiang
> to test it.

(Zhao Xiang has already confirmed that they will test what I try to write.
I'd like to say "thank you.")

But I've thought about the issue during Sunday.  If configure script detects
the bug, it should print a message telling the user to apply the HP patch
PHNE_26771 (or its successor), it should perhaps even wait for the user to
hit enter.

Well, if we start adding such autodetection of bugs into the configure script,
when will this all end?  How many bugs have already appeared in a release of
Linux kernel?

Is it right to convert configure script to an extremely huge beast which
tries to detect any defect which might appear on any platform?

Well, I see that argumenting with Linux (and other free systems which have
frequebt releases) is a bit demagogic, and that not every bug requires a
special treatment in configure script.  So, shouldn't we restrict to
commercial Unices with only a finite number of releases?  Should we also
cover the bugs introduced to these system with applying an older vendor 
patch, even though a newer patch removes the bug again?  Or should we just
cover the bugs in stock systems, for the convenience of those who don't patch
their system?  (Probably the later variant is bad, since professionaly
installed system are updated at the moment of purchasing, so people usually
won't get that old "stock system," am I right?)
Another question: shouldn't someone start actively monitor updates for every
system available and for (almost) every bug reported write a test for
autoconf?

I don't want to start making a collection, when there are no capacities to
hold the collection and the collection will be eventually thrown away.
And I cannot imagine a rules for the collection which wouldn't try to convert
autoconf configure scripts to kind of "universal knowledge base" for all
operating systems; and I think this goal is too unrealistic.

So the sad conclusion is: if a kernel (be it HP-UX or Linux) has a bug,
the user should solve it with the vendor (as Zhao Xiang has promptly done)
but GNU software should not try to tell people how to administrate their
systems.
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.

What do you think about this?

Looking forward to your suggestions,
                Stepan kasal