Re: [9fans] Re: configure misery

Re: [9fans] Re: configure misery

[Li Yi]

Bushnell? He's been dismissed:
http://lists.softwarelibero.it/pipermail/discussioni/2003-November/00846
5.html

<mirtchov@cpsc.ucalgary.ca> wrote in message
news:<b911aa84520656750ef808a846d91661@plan9.ucalgary.ca>...
> > Really people, I think Plan 9 is great, I wish I had the time to
> > invest in working with it, and I'd love to see it take over the
> > world. But it's not true that the only source of decent software is
> > 1127, and it's also not true that there's nothing you can do about
> > GNUware that you don't like. (Yes, boyd, er, Susan, heresy. Oh
> > well.)
>
> There once was a man called Bushnell,
> Who posted to 9fans a knell:
> "Plan 9 is dead,
> Use GNUware instead"
> But we left him for Glenda to pell.
>
> :)
>

Re: [9fans] Re: configure misery

[Aharon Robbins]

I wrote this:

>> > Really people, I think Plan 9 is great, I wish I had the time to
>> > invest in working with it, and I'd love to see it take over the
>> > world. But it's not true that the only source of decent software is
>> > 1127, and it's also not true that there's nothing you can do about
>> > GNUware that you don't like. (Yes, boyd, er, Susan, heresy. Oh
>> > well.)

><mirtchov@cpsc.ucalgary.ca> wrote in message
>> There once was a man called Bushnell,
>> Who posted to 9fans a knell:
>> "Plan 9 is dead,
>> Use GNUware instead"
>> But we left him for Glenda to pell.
>>
>> :)

I never said Plan 9 was dead, or that all the world should switch to GNU,
or that everything about GNU is perfect.

Reread what I did say. It speaks for itself.
--
Aharon (Arnold) Robbins --- Pioneer Consulting Ltd.	arnold AT skeeve DOT com
P.O. Box 354		Home Phone: +972  8 979-0381	Fax: +1 530 688 5518
Nof Ayalon		Cell Phone: +972 51  297-545
D.N. Shimshon 99785	ISRAEL

Re: [9fans] Re: configure misery

[a]

// I never said Plan 9 was dead, or that...

for what it's worth, i didn't read mirtchov's verse as a serious critique
of your comment, but rather a light jab at a formerly-frequent poster. no
need to get defensive.

and in that vein:
// ...it's not true that the only source of decent software is 1127...

correct: Apple and Google also produce good code. And some folks at *NL
joints. and VN. and a few other 11* bits (although not most of them).
and... uh... yeah.

;-) (means i'm not *entirely* serious)
ア

Re: [9fans] Re: configure misery

[Jim Choate]

On Tue, 3 Feb 2004 a@9srv.net wrote:

> // I never said Plan 9 was dead, or that...
>
> for what it's worth, i didn't read mirtchov's verse as a serious critique
> of your comment, but rather a light jab at a formerly-frequent poster. no
> need to get defensive.

No need to be offensive either. The comment from Mirtchov was in poor
taste and confrontational.

Jeesh.

 -- --

Open Forge, LLC  24/365 Onsite Support for PCs, Networks, & Game Consoles
512-695-4126 (Austin, Tx.)  help@open-forge.com  irc.open-forge.com

Hangar 18  Open Source Distributed Computing Using Plan 9 & Linux
512-451-7087  http://open-forge.org/hangar18  irc.open-forge.org

James Choate  512-451-7087  ravage@ssz.com  jchoate@open-forge.com

Re: [9fans] Re: configure misery

[a]

// The comment from Mirtchov was in
// poor taste and confrontational.

maybe. but it was funny, too. :-)

and it wasn't confrontational, at least not to the person to
whom he was responding. what his intent re: bushnell was, i
couldn't really say.

and people tell me *i* have no sense of humour!
ア

Re: [9fans] Re: configure misery

[andrey mirtchovski]

>><mirtchov@cpsc.ucalgary.ca> wrote in message
>>> There once was a man called Bushnell,
>>> Who posted to 9fans a knell:
>>> "Plan 9 is dead,
>>> Use GNUware instead"
>>> But we left him for Glenda to pell.
>>>
>>> :)
>
> I never said Plan 9 was dead, or that all the world should switch to GNU,
> or that everything about GNU is perfect.
>
> Reread what I did say. It speaks for itself.

The limerick was about Thomas Bushnell, BSG, not about Arnold Robbins,
I'm sorry you took it personally -- it was never intended for you...
The only link between you and the the limerick is the word 'GNUware'
:)

andrey

Re: [9fans] Re: configure misery

[Jim Choate]

Andrey,

On Tue, 3 Feb 2004, andrey mirtchovski wrote:

> The limerick was about Thomas Bushnell, BSG, not about Arnold Robbins,
> I'm sorry you took it personally -- it was never intended for you...
> The only link between you and the the limerick is the word 'GNUware'
> :)

Then why did you send it to a stranger who wouldn't understand it? Nope,
me thinks there was something else going on there.

This is exactly the short of 'shit' on this list that is so damn aggravating.

 -- --

Open Forge, LLC  24/365 Onsite Support for PCs, Networks, & Game Consoles
512-695-4126 (Austin, Tx.)  help@open-forge.com  irc.open-forge.com

Hangar 18  Open Source Distributed Computing Using Plan 9 & Linux
512-451-7087  http://open-forge.org/hangar18  irc.open-forge.org

James Choate  512-451-7087  ravage@ssz.com  jchoate@open-forge.com

Re: [9fans] Re: configure misery

[boyd, rounin]

> Then why did you send it to a stranger who wouldn't understand it? Nope,
> me thinks there was something else going on there.

the limeric was posted to the list.

Re: [9fans] Re: configure misery

[Jim Choate]

On Tue, 3 Feb 2004, boyd, rounin wrote:

> > Then why did you send it to a stranger who wouldn't understand it? Nope,
> > me thinks there was something else going on there.
>
> the limeric was posted to the list.

No, it was posted in response to anothers comments. There is an implied
connection there.

Poor execution means poor results. In general if one wants to talk to a
general population you don't re: an individual.

 -- --

Open Forge, LLC  24/365 Onsite Support for PCs, Networks, & Game Consoles
512-695-4126 (Austin, Tx.)  help@open-forge.com  irc.open-forge.com

Hangar 18  Open Source Distributed Computing Using Plan 9 & Linux
512-451-7087  http://open-forge.org/hangar18  irc.open-forge.org

James Choate  512-451-7087  ravage@ssz.com  jchoate@open-forge.com

Re: [9fans] Re: configure misery

[David Presotto]

[-- Attachment #1: Type: text/plain, Size: 190 bytes --]

Must we start this all over again?  I'm not directing this to Jim alone
even though that's the message I'm replying to.  Can we squelch this
childish stupidity before it goes on for weeks?

[-- Attachment #2: Type: message/rfc822, Size: 2782 bytes --]

From: Jim Choate <ravage@einstein.ssz.com>
To: <9fans@cse.psu.edu>
Subject: Re: [9fans] Re: configure misery
Date: Tue, 3 Feb 2004 09:36:45 -0600 (CST)
Message-ID: <Pine.LNX.4.33.0402030935110.8049-100000@einstein.ssz.com>


Andrey,

On Tue, 3 Feb 2004, andrey mirtchovski wrote:

> The limerick was about Thomas Bushnell, BSG, not about Arnold Robbins,
> I'm sorry you took it personally -- it was never intended for you...
> The only link between you and the the limerick is the word 'GNUware'
> :)

Then why did you send it to a stranger who wouldn't understand it? Nope,
me thinks there was something else going on there.

This is exactly the short of 'shit' on this list that is so damn aggravating.

 -- --

Open Forge, LLC  24/365 Onsite Support for PCs, Networks, & Game Consoles
512-695-4126 (Austin, Tx.)  help@open-forge.com  irc.open-forge.com

Hangar 18  Open Source Distributed Computing Using Plan 9 & Linux
512-451-7087  http://open-forge.org/hangar18  irc.open-forge.org

James Choate  512-451-7087  ravage@ssz.com  jchoate@open-forge.com

Re: [9fans] Re: configure misery

[Aharon Robbins]

In article <8923b8c398ed8271a224b230c669a94c@plan9.ucalgary.ca> you write:
>The limerick was about Thomas Bushnell, BSG, not about Arnold Robbins,
>I'm sorry you took it personally -- it was never intended for you...
>The only link between you and the the limerick is the word 'GNUware'
>:)
>
>andrey

Apology accepted.  I'm sorry I took it personally too. I probably
shouldn't have.  -- Arnold

Re: [9fans] Re: configure misery

[Jim Choate]

Hi Aharon,

On Tue, 3 Feb 2004, Aharon Robbins wrote:

> I never said Plan 9 was dead, or that all the world should switch to GNU,
> or that everything about GNU is perfect.
>
> Reread what I did say. It speaks for itself.

Get used to it, it seems to be a MO for a considerable contingent on this
list.

My suggestion, just ignore them.

 -- --

Open Forge, LLC  24/365 Onsite Support for PCs, Networks, & Game Consoles
512-695-4126 (Austin, Tx.)  help@open-forge.com  irc.open-forge.com

Hangar 18  Open Source Distributed Computing Using Plan 9 & Linux
512-451-7087  http://open-forge.org/hangar18  irc.open-forge.org

James Choate  512-451-7087  ravage@ssz.com  jchoate@open-forge.com

Re: [9fans] Re: configure misery

[a]

// My suggestion, just ignore them.

mmm. the irony here is just delicious.

Re: [9fans] Re: configure misery

[Jim Choate]

On Tue, 3 Feb 2004 a@9srv.net wrote:

> // My suggestion, just ignore them.
>
> mmm. the irony here is just delicious.

Yep, there is a certain contingent that thinks they are above basic
courtesy.  They belive others should be 'courtious' to them but there is
no contra-coup upon themselves.

The real irony is that these folks can't even do what they expect others
to do.

Screw 'em.

 -- --

Open Forge, LLC  24/365 Onsite Support for PCs, Networks, & Game Consoles
512-695-4126 (Austin, Tx.)  help@open-forge.com  irc.open-forge.com

Hangar 18  Open Source Distributed Computing Using Plan 9 & Linux
512-451-7087  http://open-forge.org/hangar18  irc.open-forge.org

James Choate  512-451-7087  ravage@ssz.com  jchoate@open-forge.com

Re: [9fans] Re: configure misery

[andrey mirtchovski]

> Screw 'em.

You're now getting personal, which, in my opinion, is off-topic.  If
you refuse to accept my apology to Arnold Robbins (what's it to you,
anyways?  do you have a grudge or something?) make sure to vent your
frustration in a private email or on a different list.

andrey

Re: [9fans] Re: configure misery

[Jim Choate]

On Tue, 3 Feb 2004, andrey mirtchovski wrote:

> > Screw 'em.
>
> You're now getting personal, which, in my opinion, is off-topic.

Can't imagine how you'd take that personaly.

 -- --

Open Forge, LLC  24/365 Onsite Support for PCs, Networks, & Game Consoles
512-695-4126 (Austin, Tx.)  help@open-forge.com  irc.open-forge.com

Hangar 18  Open Source Distributed Computing Using Plan 9 & Linux
512-451-7087  http://open-forge.org/hangar18  irc.open-forge.org

James Choate  512-451-7087  ravage@ssz.com  jchoate@open-forge.com

Re: [9fans] Re: configure misery

[Wes Kussmaul]

> > mmm. the irony here is just delicious.
>
>Yep, there is a certain contingent that thinks they are above basic
>courtesy.  They belive others should be 'courtious' to them but there is
>no contra-coup upon themselves.
>
>The real irony is that these folks can't even do what they expect others
>to do.
>
>Screw 'em.

Jim, why not hang out on the OpenBSD lists. They just love these verbal
hockey games over there.

Re: [9fans] Re: configure misery

[Spamm Trapp]

> I will close with a single addemdum, when people
> start thinking about what they're saying -before-
> they post it I'll quit giving them crap for their
> crap. I'm simply tired of seeing it.

yup, me too.

can this thread die, please

-S.T.


__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free web site building tool. Try it!
http://webhosting.yahoo.com/ps/sb/

[9fans] Re: [hangar18-general] Frustration...

[Jim Choate]

Hi Brad,

I'm forwarding to 9fans as somebody there is probably already aware of
it...

This sounds like the problem that had me stalled last week. There were
some comments about it but I don't remember if it got fixed. Might want to
google the 9fans list...

I was going to take a look at it over the weekend as I have no other work on
deadline at the moment. I've got to get saucer back up because the
mixmaster port is on hold until that gets fixed. I'll be building pcauth
myself. Hopefully I won't have the problem (knock on wood).

On Thu, 29 Jan 2004, Brad Davis wrote:

> A new kernel starts to load, then the system reboots. It's so fast
> I can't catch the error messages. :(
>
> I put a boot menu in plan9.ini, which makes it easy to test and
> recover, but no success yet.
>
> The default 9pcdisk kernel works fine, so I tried the default
> pcdisk config file but the new 9pcdisk kernel panics as well.
>
> The build always completes without errors. I still don't quite
> understand why the auth kernel loads from sdC0!9fat instead of
> /386... I tried it both ways just for kicks.
>
> Feel free to reply with "RTFM", or not at all. ;-)
>
> Just wanted to vent my frustration,
>
> --
> Brad
>

 -- --

Open Forge, LLC  24/365 Onsite Support for PCs, Networks, & Game Consoles
512-695-4126 (Austin, Tx.)  help@open-forge.com  irc.open-forge.com

Hangar 18  Open Source Distributed Computing Using Plan 9 & Linux
512-451-7087  http://open-forge.org/hangar18  irc.open-forge.org

James Choate  512-451-7087  ravage@ssz.com  jchoate@open-forge.com

[9fans] Re: configure misery

[rog]

[apologies to all for adding to this, but...]

Jim: have you ever considered that the maxim "be strict in what
you produce, but liberal in what you accept" can apply just as well
to mailing list interactions as it does to computer protocols?

Re: [9fans] Re: configure misery

[Jim Choate]

On Tue, 3 Feb 2004 rog@vitanuova.com wrote:

> [apologies to all for adding to this, but...]
>
> Jim: have you ever considered that the maxim "be strict in what
> you produce, but liberal in what you accept" can apply just as well
> to mailing list interactions as it does to computer protocols?

Absolutely, but one must draw a line somewhere. That somewhere in this
case is an 'institutionalized' bias in who can say what to whom.

May I ask, why you bothered to break your own suggestion?

I will close with a single addemdum, when people start thinking about what
they're saying -before- they post it I'll quit giving them crap for their
crap. I'm simply tired of seeing it.

 -- --

Open Forge, LLC  24/365 Onsite Support for PCs, Networks, & Game Consoles
512-695-4126 (Austin, Tx.)  help@open-forge.com  irc.open-forge.com

Hangar 18  Open Source Distributed Computing Using Plan 9 & Linux
512-451-7087  http://open-forge.org/hangar18  irc.open-forge.org

James Choate  512-451-7087  ravage@ssz.com  jchoate@open-forge.com

[9fans] Re: configure misery

[Aharon Robbins]

In article <81132473206F3A46A72BD6116E1A06AE479CA3@black.aprote.com>,
Tiit Lankots <9fans@cse.psu.edu> wrote:
>>Seeing as how this is grep, the + is NOT a metacharacter. Thus it's
>
>Not necessarily. GNU grep (by far the most widespread by now) is using
>the 'extended' syntax where + IS a metacharacter.
>IIRC, the configure-suite is assuming GNU tools.
>
>Tiit

Incorrect. The whole point of configure is to make it possible to configure
programs across the whole range of unix systems.  Particularly for standard
things like grep, the assumption is that GNU tools are NOT present. (Even there,
I'm pretty sure that plain grep uses \+ to make + a metacharacter.)

Autoconf requires GNU m4 to *generate* configure, but no non-standard tools
are required to run the generated script.

As for vitriol aimed at configure, come down from the Ivory Tower for a
little bit, people.  Consider what existed when it was first written in
the late '80s and early '90s.  A huge morass of different vendor systems
with a vast swamp of hodge-podge features.  The only way to know if you
have a library function or kernel feature is to try to test for it; you
can't rely on ifdefs.

The Autoconf machinery abstracts *a lot* of this away, where you, the
program author, don't have to deal with it.  *Switching* a program to
use Autoconf is a pain.  But the gain is well worth it. I went through this
with gawk in the mid-90s.  Look at gawk-2.15.6 (if you can find it) with
a bunch of little files for all the different kinds of systems out there.
ALL of which had to be maintained MANUALLY.  Now, I work with configure.ac
and Makefile.am, which are fairly easy to manage once setup, and the rest
of the headache is gone.

POSIX has helped considerably, but it hasn't made Unix systems homogenous.
Autoconf solved (and solves) a real problem (vendor Unix divergence)
for those of us without the luxury of working under Plan 9 all day.
And old systems (SunOS 4.x, Ultrix) *are* still in use.

Sure configure itself is huge.  It's object code though, not source code.
And IMHO the Makefiles generated by Automake are too complicated.  But
would I switch back to doing things manually?  No Way.

FWIW, gawk ships as standard on every Linux system and most or all BSDs,
and configures, compiles and runs out of the box on over a dozen different
kinds of Unix systems.  On all of them, ALL of them, the incantation is
ONE LINE:

	./configure && make all check install

Compare that to:

	for each vendor Unix system I happen to have
	do
		while it doesn't make
		do
			vi myprog.h Makefile
			make
		done
	done

And multiply that by all the different packages you'd like to install.

Now tell me what you, the poor sysadmin, would rather spend your time doing?

Flames to /dev/null.

Arnold

Re: [9fans] Re: configure misery

[Geoff Collyer]

configure is undeniably convenient when it works; it's the times it
doesn't work that are a pain.  If the configure authors haven't seen
your particular variant system, configure is very likely to either
generate an incorrect configuration or blow up and refuse to run.  The
incorrect configuration isn't too bad; you can just edit config.h.
But when it refuses to run, you then get to wade into the `object
code' shell script, which is dense and complicated, and try to
persuade it to do the right thing and continue to run.  I find that
deleting large sections of configure often accomplish this.  If that
fails, compiling by hand may be easier than coaxing configure into
working.  And you get to do this for every program you try to import
that uses configure.

And some of the things that configure probes for just aren't worth
probing for.  Every ANSI/POSIX C implementation has to provide a
compiler under the name `c89'.  So why does configure still probe for
compiler names?  If you have a very old system and still don't have an
ANSI C implementation, you're going to have a very hard time importing
software in general and you probably ought to go get one (lcc and gcc
are just two that come to mind).  configure also tries to verify that
the compiler it has found generates .o files.  I know this isn't
required by the 1989 C standard and I doubt that POSIX requires it,
and indeed some ANSI compilers we know don't generate .o files.
Similarly, configure probes for the system include-file directory; why
does it care?  The whole process would work *better* if configure just
stopped checking for things like this.  After all, how many systems
have commands named `c89' that don't generate locally-appropriate
object files?

And cross-compilation would seem to be out of the question with
configure, unless configure runs itself remotely on the target and
drags the answers back.

I think one can live in the real world and still find configure to be
a larger problem than the one it claims to solve.

Re: [9fans] Re: configure misery

[Charles Forsyth]

>>doesn't work that are a pain.  If the configure authors haven't seen
>>your particular variant system, ...

that gets to the heart of it: for all it claims to do, it really only works
well (and usually only works at all) as an enumeration of systems already
seen; in which case you might just as well have config-i386-linux.h
and be done with it.  for instance, when i looked at a QNX paper on
importing software i was impressed that configure worked for them
until i realised that was only because a qnx variant is written into configure!
what's the point?  it has clearly become someone's life work, heart and soul poured into it,
etc. but it's rather a waste of time.  its author(s) should get out a bit more.

Re: [9fans] Re: configure misery

[Bruce Ellis]

configure works on an x86 linux box. it certainly didn't work
on plan 9 or irix as dhog discovered. maybe it needs to be
more complex but still as unusable on a particular platform.

brucee
----- Original Message -----
From: "Aharon Robbins" <arnold@skeeve.com>
To: <9fans@cse.psu.edu>
Sent: Sunday, November 16, 2003 7:04 PM
Subject: [9fans] Re: configure misery

Re: [9fans] Re: configure misery

[Lyndon Nerenberg]

--On Sunday, November 16, 2003 10:04 AM +0200 Aharon Robbins
<arnold@skeeve.com> wrote:

> FWIW, gawk ships as standard on every Linux system and most or all
> BSDs,

Due to the AT&T lawsuit fiasco, and not because of the joys of
configure. (I'm not sure if that was the implication you were making.)
Also, FreeBSD has reverted to BWK awk for /usr/bin/awk.

--lyndon

Re: [9fans] Re: configure misery

[andrey mirtchovski]

On Sun, 16 Nov 2003, Lyndon Nerenberg wrote:

> Also, FreeBSD has reverted to BWK awk for /usr/bin/awk.

The BSD's would like to be completely free of all GNU software for pretty
much the same reasons people bash it on this list. Just ask Theo :)

andrey

Re: [9fans] Re: configure misery

[Dan Cross]

andrey mirtchovski <mirtchov@cpsc.ucalgary.ca> writes:
> On Sun, 16 Nov 2003, Lyndon Nerenberg wrote:
> > Also, FreeBSD has reverted to BWK awk for /usr/bin/awk.
>
> The BSD's would like to be completely free of all GNU software for pretty
> much the same reasons people bash it on this list. Just ask Theo :)

The thing is, the BSD source code isn't that much better than the GNU
stuff, especially now.  It seems to be growing without bound.

Maybe Theo et al would be more productive if they quit complaining and
actually set to work on some of the things they find so irksome.  For
instance, there's a BSD licensed compiler out there, and it's existed
for years; why haven't they switched to it?  All the energy they put
into reinventing the square wheel could be put into building new
backends for TenDRA, and getting it to generate correct code for their
build target.

	- Dan C.

Re: [9fans] Re: configure misery

[Lyndon Nerenberg]

--On Sunday, November 16, 2003 4:43 PM -0500 Dan Cross
<cross@math.psu.edu> wrote:

> The thing is, the BSD source code isn't that much better than the GNU
> stuff, especially now.  It seems to be growing without bound.

This isn't specific to just BSD and GNU. It's a general problem
attributable to the large number of new programmers who haven't been
mentored by people who understand the elegance of uncomplicated code
and tools. This feeds upon itself, and once it reaches critical mass....

Retreating into plan9 is a way to avoid dealing with this problem, but
if nobody teaches these folks the error of their ways, how can it ever
stop? I don't think most of us have the energy to attempt that any
more. (Or in my case, the patience.)

--lyndon

Re: [9fans] Re: configure misery

[mirtchov]

> This isn't specific to just BSD and GNU. It's a general problem
> attributable to the large number of new programmers who haven't been
> mentored by people who understand the elegance of uncomplicated code
> and tools. This feeds upon itself, and once it reaches critical mass....

I think the teachers themselves are of the new generation.  This
doesn't help much, especially when the educational system (I have
personal experiences with two canadian universities) is forcing the
students to cover a lot of ground very quickly, with emphasis on
quantity of the material submitted.

Even if a teacher comes along who is willing to mentor students in
"the right way", they simply refuse to learn (except the usual 5% who
are willing, but they usually know how to do code right already) --
their goal is to get through the course without a drop of sweat,
usually because there are 5 other software engineering courses waiting
for them to draw diagrams of the waterfall.  The complaints that the
particular course below was "too difficult" drowned any benefit the
students had from taking it.

This is an undergraduate course I TA'd for last year (abstract
yourselves from the fact that they used OpenBSD, it could've been any
other lunix OS):

	http://ezine.daemonnews.org/200306/obsd-classroom.html

And a quote from the article:

	The other part of the problem was the students' fledgling
	code-reading skills.  The TAs taught students how to use the
	OpenBSD machines, build kernels, and use tools like grep to
	search for things in the kernel source, but I asked them to
	give students limited guidance in the code itself.  I had high
	expectations of the students' ability to read code, perhaps
	too high -- students are not exposed to a lot of C code now,
	and the OpenBSD kernel documentation, written by experts for
	experts, was of little help to students.

I would've loved to see this course taught in Plan 9, if nothing else
for the simplicity of the code the students needed to get through.  I
doubt though it would've made much difference in the end.  Perhaps
Nemo can share his experience with teaching Plan 9 to undergrads...

Are we offtopic enough already? :)

Andrey

Re: [9fans] Re: configure misery

[boyd, rounin]

> Are we offtopic enough already? :)

maybe, but the problem is that the 'teachers' have been infected
by this 'bloat mentality' too.

Re: [9fans] Re: configure misery

[Dan Cross]

Lyndon Nerenberg <lyndon@orthanc.ca> writes:
> <cross@math.psu.edu> wrote:
> > The thing is, the BSD source code isn't that much better than the GNU
> > stuff, especially now.  It seems to be growing without bound.
>
> This isn't specific to just BSD and GNU. It's a general problem
> attributable to the large number of new programmers who haven't been
> mentored by people who understand the elegance of uncomplicated code
> and tools. This feeds upon itself, and once it reaches critical mass....

I agree with you completely.

> Retreating into plan9 is a way to avoid dealing with this problem, but
> if nobody teaches these folks the error of their ways, how can it ever
> stop? I don't think most of us have the energy to attempt that any
> more. (Or in my case, the patience.)

More to the point, I'm not sure it's possible.  People aren't interested
in learning how to program elegantly (try telling Theo Deraadt that he
could program more elegantly.  For that matter, try telling him he could
do *anything* differently), and it's difficult to persuade someone who's
convinced they are an expert on something to do that thing differently.
What's more, we're seeing a lot of people who confuse installed base and
popularity with quality.  Often times when I tell people I use Plan 9,
the first question they ask me is, ``why?''  When I say that it's better
than Linux in terms of the quality of implementation, the response I get
is ``then why don't more people use it?  If it's so good, why hasn't it
displaced Linux?''  Explaining the historical circumstances that's made
that the case rarely convinces people that Linux (or BSD) isn't all it's
cracked up to be by the hype.

	- Dan C.

Re: [9fans] Re: configure misery

[boyd, rounin]

> ... and it's difficult to persuade someone who's convinced ...

next to impossible.

Re: [9fans] Re: configure misery

[Russ Cox]

> > Also, FreeBSD has reverted to BWK awk for /usr/bin/awk.
>
> The BSD's would like to be completely free of all GNU software for pretty
> much the same reasons people bash it on this list. Just ask Theo :)

Actually, only OpenBSD, and for completely different reasons.

OpenBSD wants to be free of GNU software for purely ideological reasons.
If the same software were available under the BSD license they would have
no problem whatsoever keeping it around.  Listen to what Theo has said
here -- if 8c were available under a BSD license, he'd use it.  Otherwise not.
Regardless of its technical merits or not vs. gcc.  I'm not aware that
any of the other BSDs are itching to get rid of all their GNU software.
Judging from the CVS and other logs, it appears that FreeBSD switched
to BWK awk mainly because gawk had problems on the Sparc64.

On this list, my impression is that people bash the GNU software because
the canonical examples (gcc, configure, etc.) are overly complicated, slow,
and hard to understand.  These are very practical reasons.  If software
written like Plan 9 were available under a GPL, I would use it and happily
incorporate it into the distribution if it were useful.

Ideology vs. a certain kind of software quality.

Finally (and this is more directed at Lyndon's post than yours) not all
GNU software is crap.  Some of it is quite good, and I believe that gawk
and groff are examples of this.  Arguably groff is built better than troff.
(I'm not as familar with gawk vs. awk so don't construe my lack of comment
as anything other than ignorance.)  Assuming that GNU software has uniform
software quality is simply flawed.

Getting back to the original topic, I think forsyth summed up well.
It's not clear to me that configure buys you much more than #ifdefs
nowadays, though in the days of many varied vendor systems I can
believe it had more utility.  I am particularly amused by the configure
scripts that have begun appearing recently, which churn for a while
and then say something approximating "config.sub: I've never heard
of a system called `plan9' before, so I'm not continuing."

Russ

Re: [9fans] Re: configure misery

[mirtchov]

> Actually, only OpenBSD, and for completely different reasons.

I agree with you on all other counts, but to be fair to Theo, he
wanted the compilers because they were fast, not because they weren't
GNU:

	http://www.monkey.org/openbsd/archive/tech/0302/msg00064.html

My original statement that BSDers hate GNU was more of a joke, not a
fact.  I myself "hate" GNU because of what they did to Hello World:

	ftp://ftp.gnu.org/gnu/hello/hello-2.1.1.tar.gz

;)

andrey

Re: [9fans] Re: configure misery

[Russ Cox]

> I agree with you on all other counts, but to be fair to Theo, he
> wanted the compilers because they were fast, not because they weren't
> GNU:

Yes but even though they are fast, he wouldn't take them if they were GNU.
I'm not motivated enough to check whether he ever said this to 9fans, but
he certainly said it to me.

Re: [9fans] Re: configure misery

[Mike Haertel]

>My original statement that BSDers hate GNU was more of a joke, not a
>fact.  I myself "hate" GNU because of what they did to Hello World:
>
>        ftp://ftp.gnu.org/gnu/hello/hello-2.1.1.tar.gz

We originally did that purely as a joke.  Somewhere along the line
it took on an unholy life of its own...

Re: [9fans] Re: configure misery

[Lyndon Nerenberg]

--On Sunday, November 16, 2003 4:46 PM -0500 Russ Cox <rsc@swtch.com>
wrote:

> I'm not aware that any of the other BSDs are itching to
> get rid of all their GNU software. Judging from the CVS and other
> logs, it appears that FreeBSD switched to BWK awk mainly because gawk
> had problems on the Sparc64.

FreeBSD would like to replace GPL code with BSD licensed code, *when it
it practical to do so*. The (well, a) main reason for this is to make
life easier for the embedded systems folks, who need to worry about GPL
taint requiring them make their proprietary source available. There's
much less zealotry about this in the FreeBSD camp vs. the OpenBSD
folks. The reasons for replacing awk, in no particular order, were:
size of the gawk code, license, and compatibility with traditional awk.
The mailing lists describe the reasoning better than CVS does.

> Finally (and this is more directed at Lyndon's post than yours) not
> all GNU software is crap.  Some of it is quite good, and I believe
> that gawk and groff are examples of this.

I wasn't (necessarily) implying that. My biggest concern with GNU code
(and this is not a GPL issue) is the massive feature-itis in the
traditional UNIX tools. It's fine for GNU tools to grow new features if
that's what they want, but when those tools are shipped as replacements
for the traditional UNIX ones, we start seeing lots of portability
problems. gawk is one example of this. When it's installed and invoked
as awk it doesn't disable the GNU extensions. This results in awk
scripts that rely on gawk-specific features, and in most cases the
script author doesn't even realize this is happening. bash masquerading
as sh is another culprit here. I've spent a lot of time over the years
fixing portability bugs in software that was developed on GNUish
systems, but had to run on traditional UNIX or POSIX environments. The
only real fix for this is for the world to acknowledge that Linux just
ain't UNIX, and let's all stop pretending it is.

--lyndon

Re: [9fans] Re: configure misery

[boyd, rounin]

> The only real fix for this is for the world to acknowledge that Linux just
> ain't UNIX, and let's all stop pretending it is.

yup, the irony being that GNU's Not Unix.  QED

Re: [9fans] Re: configure misery

[Aharon Robbins]

In article <2147483647.1068997081@[192.168.42.6]> you write:
>I wasn't (necessarily) implying that. My biggest concern with GNU code
>(and this is not a GPL issue) is the massive feature-itis in the
>traditional UNIX tools. It's fine for GNU tools to grow new features if
>that's what they want, but when those tools are shipped as replacements
>for the traditional UNIX ones, we start seeing lots of portability
>problems. gawk is one example of this. When it's installed and invoked
>as awk it doesn't disable the GNU extensions.

It's against the GNU Coding Standards to have a program's behavior
depend on its name, and this makes some sense.  One easy solution
is to have /bin/awk be:

	#! /bin/sh

	exec gawk --posix "$@"

And maybe BSD and Linux distributions should do just that.
Alternatively, you can put

	export POSIXLY_CORRECT=1

into /etc/profile. (This latter is rather draconian, but it would work.)

I'll agree that GNUware is too featureful.  This includes gawk, which
has suffered much from my then-youthful enthusiasm.  I currently try to
avoid feature creeep, but old habits die hard.  Naetheless, there are
lots of things people ask for that I say "no" to.  So I'm learning,
albeit slowly.  (A decent extension mechanism and standardize API
into the interpreter's internals would move the problem out of the core
code. Not enough time...)

Featuritis isn't unique to the GNU world though.  Anybody looked at
ksh93 recently?  Who needs awk or perl or sed anymore?

Arnold

Re: [9fans] Re: configure misery

[Russ Cox]

> Featuritis isn't unique to the GNU world though.  Anybody looked at
> ksh93 recently?  Who needs awk or perl or sed anymore?

We don't write anything in Perl anymore,
because [ksh93] has all the functionality built in.

 - David Korn, quoted in a Slashdot interview

Re: [9fans] Re: configure misery

[Joel Salomon]

>> Featuritis isn't unique to the GNU world though.  Anybody looked at
>> ksh93 recently?  Who needs awk or perl or sed anymore?
>
> We don't write anything in Perl anymore,
> because [ksh93] has all the functionality built in.
>
>  - David Korn, quoted in a Slashdot interview
>

There's not much Perl code written for plan9 either, is there... ;-)

--Joel

Re: [9fans] Re: configure misery

[Enache Adrian]

On Sun, Nov 16, 2003 a.d., Russ Cox wrote:
> Finally (and this is more directed at Lyndon's post than yours) not all
> GNU software is crap.  Some of it is quite good, and I believe that gawk
> and groff are examples of this.  Arguably groff is built better than troff.
> (I'm not as familar with gawk vs. awk so don't construe my lack of comment
> as anything other than ignorance.)  Assuming that GNU software has uniform

There's another GPL'ed awk there, called mawk.

|| $ time awk -f mc.awk mtrace
||          0          0
||          1       5898
||          2      66407
||          3      16749
|| 	...
|| gawk:
|| 	real	0m4.174s
|| 	user	0m4.078s
|| 	sys	0m0.080s
|| nawk:
|| 	real	0m6.078s
|| 	user	0m6.025s
|| 	sys	0m0.037s
|| mawk:
|| 	real	0m1.391s
|| 	user	0m1.320s
|| 	sys	0m0.055s
||
|| $ ls -l `which gawk nawk mawk`
|| -rwxr-xr-x    1 root     root       316530 Aug 11  2002 /bin/gawk
|| -rwxr-xr-x    1 root     root       106352 Jan  9  2003 /opt/bin/nawk
|| -rwxr-xr-x    1 root     root        94680 Jan  9  2003 /opt/bin/mawk
||
|| $ ls -l mtrace
|| -rw-r--r--    1 adi      adi      16956443 Nov 17 01:49 mtrace
||
|| where mc.awk is a silly script like that:
||
|| BEGIN { FS="[ \t()]+" }
|| /malloc/ { sz[$5]++ }
|| END { for(i=0; i<7; i++) printf "%10d %10d\n", i, sz[i] }
||
|| and mtrace contains lines of the form:
||
|| xxxx: xx.c:130	malloc	0x0805e3d8 (48)
|| xxxx: xx.c:135	free	0x0805e1f0
|| xxxx: xx.c:1635	malloc	0x0805e458 (2)
|| xxxx: xxxx.c:7183       realloc 0x08081de0 => 0x08081de0 (4)

And it seems not to have some of gawk's bugs:

|| $ echo | gawk 'function p() { print "hi" } BEGIN { gsub(/a/, p()) }'
|| $ echo | nawk 'function p() { print "hi" } BEGIN { gsub(/a/, p()) }'
|| hi
|| $ echo | mawk 'function p() { print "hi" } BEGIN { gsub(/a/, p()) }'
|| hi

Regards,
Adi

Re: [9fans] Re: configure misery

[Aharon Robbins]

In article <20031117002402.GA1301@ratsnest.hole> you write:
>There's another GPL'ed awk there, called mawk.

Yes, mawk is faster, it uses a different internal algorithm (byte codes
vs. recursive tree evaluation).  It is also, sadly, unmaintained; there
hasn't been a new release since 1996, and it does have a few bugs.

>And it seems not to have some of gawk's bugs:
>
>|| $ echo | gawk 'function p() { print "hi" } BEGIN { gsub(/a/, p()) }'
>|| $ echo | nawk 'function p() { print "hi" } BEGIN { gsub(/a/, p()) }'
>|| hi
>|| $ echo | mawk 'function p() { print "hi" } BEGIN { gsub(/a/, p()) }'
>|| hi

What version of gawk are you using?  I get the expected output on my
system with 3.1.3 (the current version).  (Let's take the rest of the
discussion off line.)

Arnold

Re: [9fans] Re: configure misery

[Taj Khattra]

On Mon, Nov 17, 2003 at 02:16:53PM +0200, Aharon Robbins wrote:
> Yes, mawk is faster, it uses a different internal algorithm (byte codes
> vs. recursive tree evaluation).  It is also, sadly, unmaintained; there
> hasn't been a new release since 1996, and it does have a few bugs.

not specific to mawk, and it's unfortunate that whatever bugs it has
haven't been fixed (very often i prefer mawk too, unless i require a
gawk extension :) ), but this touches on another pet peeve of mine wrt
what seems like the current modus operandi: unless you keep releasing
updates to a s/w package every so often, people automatically assume
that it's unmaintained!  bug fixes are always welcome, but more often
than not what you end up with is a continuous stream of updates with
dubious "enhancements".  there is no fixed-point in sight.

as Lamport might say (in a different context), it's as if program
maintenance is analogous to automobile maintenance, requiring constant
tweaking and "lubricating the branch statements and cleaning the pointers"
(http://research.microsoft.com/users/lamport/pubs/pubs.html#automobile)

-taj

Re: [9fans] Re: configure misery

[boyd, rounin]

configure has two fundemental problems:

    1) it only really knows about systems it's seen

    2) it goes hunting for all sorts of crap in totally bizarre ways
        and when that breaks you are in a world of pain

people who write code that use configure then decide to use
every possible facility provided that it may or may not find,
rather than a minimal subset that should work on most
systems.

i guess all that mmap() and shared memory junk is my fave.
i've never used it, never will.

isn't there come great horror, err war, story about those boys
from BSD who re-wrote cat to use mmap() which worked, as
it were, until cat & mmap() hit a pipe?

right, mmap me a socket ...

Re: [9fans] Re: configure misery

[Aharon Robbins]

In article <2147483647.1068984706@[192.168.42.6]> you write:
>--On Sunday, November 16, 2003 10:04 AM +0200 Aharon Robbins
><arnold@skeeve.com> wrote:
>
>> FWIW, gawk ships as standard on every Linux system and most or all
>> BSDs,
>
>Due to the AT&T lawsuit fiasco, and not because of the joys of
>configure. (I'm not sure if that was the implication you were making.)

The point was twofold:  1. gawk isn't some dinky program that sees no use
except by its author, and 2. that autoconf makes the 1-line configuration
incantation work on dozens of Unix variants.

>Also, FreeBSD has reverted to BWK awk for /usr/bin/awk.

I believe this is for ideological reasons as well.

As for sparc64 problems, I fail to understand why people encountering
these problems FAIL TO REPORT THEM.   This past week was the first time I
got a bug report about 64-bit sparc using gcc and glibc (under Solaris),
and because the machine isn't ssh-able, I can't get in to try and fix it.

Is Autoconf perfect?  No.  Is it better than the alternative?
My experience is yes, Geoff's is no.  Fine.  If everyone is so concerned,
open up a dialogue with the maintainers.  Maybe you'll even get somewhere,
I can't promise anything.  But sitting off in the corner and doing
nothing but kvetching about it for sure won't see it make any progress.

Arnold

Re: [9fans] Re: configure misery

[John Stalker]

Gawk is an interesting test case.  I poked around a little in the
FreeBSD port.  It doesn't just ./configure && make && make install.
It has to pass two arguments to configure to tell it the location
of libraries it wouldn't otherwise find.  More worryingly it has
to patch both Makefile.in and configure.  In other words, some poor
soul had to read and understand the ``object code'' generated by
autoconf well enough to patch it.  In this case it appears that
autoconf has shifted some work from the program author to the port
maintainer.

I don't mean to flame either the developers of gawk, which I use, or
of autoconf, which I don't.  I do think autoconf was a great mistake,
however.  If you want a program which produces install scripts then
there are two sensible paths: (1) write something quick and dirty
which produces simple scripts which almost work on most systems or
(2) write something which produces horrible monsters which no one need
ever look at because they always work.  Autoconf clearly aims for
(2), but it doesn't quite get there.  A program which opted for (1)
would be more useful.

Note that gawk has far fewer dependencies than some other programs.
In fact the only library it needs which is not part of the base system
is the gettext library, which is the one autconf doesn't find.

> The Autoconf machinery abstracts *a lot* of this away, where you, the
> program author, don't have to deal with it.  *Switching* a program to
> use Autoconf is a pain.  But the gain is well worth it. I went through this
> with gawk in the mid-90s.  Look at gawk-2.15.6 (if you can find it) with
> a bunch of little files for all the different kinds of systems out there.
> ALL of which had to be maintained MANUALLY.  Now, I work with configure.ac
> and Makefile.am, which are fairly easy to manage once setup, and the rest
> of the headache is gone.
.
.
.
> Sure configure itself is huge.  It's object code though, not source code.
> And IMHO the Makefiles generated by Automake are too complicated.  But
> would I switch back to doing things manually?  No Way.
>
> FWIW, gawk ships as standard on every Linux system and most or all BSDs,
> and configures, compiles and runs out of the box on over a dozen different
> kinds of Unix systems.  On all of them, ALL of them, the incantation is
> ONE LINE:
>
> 	./configure && make all check install
>
> Compare that to:
>
> 	for each vendor Unix system I happen to have
> 	do
> 		while it doesn't make
> 		do
> 			vi myprog.h Makefile
> 			make
> 		done
> 	done
>
> And multiply that by all the different packages you'd like to install.
>
> Now tell me what you, the poor sysadmin, would rather spend your time doing?
>
> Flames to /dev/null.
>
> Arnold
>

--
John Stalker
Department of Mathematics
Princeton University
(609)258-6469

Re: [9fans] Re: configure misery

[Aharon Robbins]

In article <200311170042.hAH0gTnM028044@math.Princeton.EDU> you write:
>Gawk is an interesting test case.  I poked around a little in the
>FreeBSD port.  It doesn't just ./configure && make && make install.
>It has to pass two arguments to configure to tell it the location
>of libraries it wouldn't otherwise find.  More worryingly it has
>to patch both Makefile.in and configure.  In other words, some poor
>soul had to read and understand the ``object code'' generated by
>autoconf well enough to patch it.  In this case it appears that
>autoconf has shifted some work from the program author to the port
>maintainer.
>
> ...
>
>Note that gawk has far fewer dependencies than some other programs.
>In fact the only library it needs which is not part of the base system
>is the gettext library, which is the one autconf doesn't find.

That is surprising to me.  I haven't looked at FreeBSD, but this makes it
sound like they've put the library in some non-standard place, since the
machinery as distributed is supposed to be able to find it, AND, since
gawk ships with a copy of the gettext library, use the one in the source
distribution if it can't find one on the system.

So, this is a guess only, the FreeBSD people seem to have made this work
for themselves.

Arnold

Re: [9fans] Re: configure misery

[Lucio De Re]

On Mon, Nov 17, 2003 at 02:28:09PM +0200, Aharon Robbins wrote:
>
> That is surprising to me.  I haven't looked at FreeBSD, but this makes it
> sound like they've put the library in some non-standard place, since the
> machinery as distributed is supposed to be able to find it, AND, since
> gawk ships with a copy of the gettext library, use the one in the source
> distribution if it can't find one on the system.
>
> So, this is a guess only, the FreeBSD people seem to have made this work
> for themselves.
>
They may have done what NetBSD did, which is to use a directory
hierarchy for the distribution packages which is unique to them
(/usr/pkg).  You probably know this.

If you want to engage the autoconf people, then high on my wish
list is to be able to distribute the answers for one tarball to
all the others tarballs that can use them (can I say MS Registry?
Yes, I can) so that configure does not have to repeat itself ad
nauseam.

I actually think Geoff has a philosophically more sound approach,
where a single i386-lucent-plan9-config.h is more sensible than
figuring it out each time.  Again, I can't help thinking it ought
to be a hierarchical structure.

++L

Re: [9fans] Re: configure misery

[Lucio De Re]

On Mon, Nov 17, 2003 at 02:42:14PM +0200, Lucio De Re wrote:
> >
> They may have done what NetBSD did, which is to use a directory
> hierarchy for the distribution packages which is unique to them
> (/usr/pkg).  You probably know this.
>
> [ ... ]

Oops, I meant that to go to Aharon only, but forgot that Mutt has its
own mind.

++L

Re: [9fans] Re: configure misery

[Aharon Robbins]

In article <20031117145310.D16487@cackle.proxima.alt.za> you write:
>On Mon, Nov 17, 2003 at 02:42:14PM +0200, Lucio De Re wrote:
>> >
>> They may have done what NetBSD did, which is to use a directory
>> hierarchy for the distribution packages which is unique to them
>> (/usr/pkg).  You probably know this.
>>
>> [ ... ]
>
>Oops, I meant that to go to Aharon only, but forgot that Mutt has its
>own mind.

The list also sets the reply-to.  I didn't know about *BSD's /usr/pkg.
But again, they made their own work.

I work under Linux, and find it stable, reliable, fast and useful. Much
better than any proprietary Unix.  My Linux systems don't crash, and they
don't lose files. Is fossil/venti there yet?  It's hard to judge based
on the list traffic.  Granted, they're both quite young, but I need my
systems to make my living.

Maybe my using Linux makes me personna non grata here, if so too bad.
I'm also a GNU volunteer that doesn't touch Emacs, so there, nyah, nyah,
nyah.

If Geoff can tweak config.sub so that it recognizes plan 9, then please
do so and mail it in; it'll get added.

The Autoconf people are aware of autoconf's flaws; the shared database
thing is, I believe, on their list.  There is already configure.cache
so that reconfiguring the same package is sped up.

Again, if it matters to you, talk to them, don't just sit off in
the 9fans corner sulking.

Really people, I think Plan 9 is great, I wish I had the time to
invest in working with it, and I'd love to see it take over the world.
But it's not true that the only source of decent software is 1127,
and it's also not true that there's nothing you can do about GNUware
that you don't like. (Yes, boyd, er, Susan, heresy. Oh well.)

Arnold

Re: [9fans] Re: configure misery

[mirtchov]

> Really people, I think Plan 9 is great, I wish I had the time to
> invest in working with it, and I'd love to see it take over the world.
> But it's not true that the only source of decent software is 1127,
> and it's also not true that there's nothing you can do about GNUware
> that you don't like. (Yes, boyd, er, Susan, heresy. Oh well.)

There once was a man called Bushnell,
Who posted to 9fans a knell:
"Plan 9 is dead,
Use GNUware instead"
But we left him for Glenda to pell.

:)

Re: [9fans] Re: configure misery

[David Presotto]

[-- Attachment #1: Type: text/plain, Size: 10 bytes --]

well said.

[-- Attachment #2: Type: message/rfc822, Size: 3842 bytes --]

From: Aharon Robbins <arnold@skeeve.com>
To: 9fans@cse.psu.edu
Cc: 
Subject: Re: [9fans] Re: configure misery
Date: Mon, 17 Nov 2003 15:43:03 +0200
Message-ID: <200311171343.hAHDh3Eb001606@skeeve.com>

In article <20031117145310.D16487@cackle.proxima.alt.za> you write:
>On Mon, Nov 17, 2003 at 02:42:14PM +0200, Lucio De Re wrote:
>> >
>> They may have done what NetBSD did, which is to use a directory
>> hierarchy for the distribution packages which is unique to them
>> (/usr/pkg).  You probably know this.
>>
>> [ ... ]
>
>Oops, I meant that to go to Aharon only, but forgot that Mutt has its
>own mind.

The list also sets the reply-to.  I didn't know about *BSD's /usr/pkg.
But again, they made their own work.

I work under Linux, and find it stable, reliable, fast and useful. Much
better than any proprietary Unix.  My Linux systems don't crash, and they
don't lose files. Is fossil/venti there yet?  It's hard to judge based
on the list traffic.  Granted, they're both quite young, but I need my
systems to make my living.

Maybe my using Linux makes me personna non grata here, if so too bad.
I'm also a GNU volunteer that doesn't touch Emacs, so there, nyah, nyah,
nyah.

If Geoff can tweak config.sub so that it recognizes plan 9, then please
do so and mail it in; it'll get added.

The Autoconf people are aware of autoconf's flaws; the shared database
thing is, I believe, on their list.  There is already configure.cache
so that reconfiguring the same package is sped up.

Again, if it matters to you, talk to them, don't just sit off in
the 9fans corner sulking.

Really people, I think Plan 9 is great, I wish I had the time to
invest in working with it, and I'd love to see it take over the world.
But it's not true that the only source of decent software is 1127,
and it's also not true that there's nothing you can do about GNUware
that you don't like. (Yes, boyd, er, Susan, heresy. Oh well.)

Arnold

Re: [9fans] Re: configure misery

[John Stalker]

I don't know if anyone cares, but FreeBSD puts libintl and libiconv
in under /usr/local, as with everything not part of the base system.
I assume NetBSD does put them under /usr/pkg, but I am sitting at
a linux box at work, so I can't check.  I actually prefer the NetBSD
convention, but I fear straying further off topic.

What I really want to comment on is the sentence ``since gawk ships
with a copy of the gettext library, use the one in the source distribution
if it can't find one on the system.''  I think that is a recipe for
disaster.  The system has a working gettext.  I don't want each application
linked against its own version.  Yes, I know gawk is statically linked, so
it won't take up any extra space, but it is a messy solution and messy
solutions always sneak up and bite me in odd ways.  I noticed that gawk
is by no means the only program which comes with an included gettext.
Even freeciv comes with one, and a configure option to enable it.  At
least both gawk and freeciv have the decency not to link in their own
versions of libraries without being asked.  I would argue that the
fact that you need to include a private gettext with the gawk distribution
is a symptom of a wider problem, and that autoconf is part of that problem.
I do admit that the default LDPATH on FreeBSD is another part of the problem.

> On Mon, Nov 17, 2003 at 02:28:09PM +0200, Aharon Robbins wrote:
> >
> > That is surprising to me.  I haven't looked at FreeBSD, but this makes it
> > sound like they've put the library in some non-standard place, since the
> > machinery as distributed is supposed to be able to find it, AND, since
> > gawk ships with a copy of the gettext library, use the one in the source
> > distribution if it can't find one on the system.
> >
> > So, this is a guess only, the FreeBSD people seem to have made this work
> > for themselves.
> >
> They may have done what NetBSD did, which is to use a directory
> hierarchy for the distribution packages which is unique to them
> (/usr/pkg).  You probably know this.
--
John Stalker
Department of Mathematics
Princeton University
(609)258-6469

Re: [9fans] Re: configure misery

[Aharon Robbins]

In article <200311171631.hAHGVToY025874@localhost.localdomain> you write:
>What I really want to comment on is the sentence ``since gawk ships
>with a copy of the gettext library, use the one in the source distribution
>if it can't find one on the system.''

All the world doesn't come with gettext. Think Irix, Tru64, HP-UX.  The
Solaris implementation is poor.  It is for that reason that gawk ships
with its own, although in the next release this will probably change.

Arnold