LFS-6.6, Stage2, glibc, nscd.c:442

Paul Rogers paulgrogers at fastmail.fm
Wed Jun 2 11:39:18 PDT 2010


> > have one system that trails, i.e. has exactly the package versions
> > specified in the HSR, and verifies that each version of LFS does in
> > fact  install flawlessly with those prerequisites.
>
> Volunteers welcomed.

I reported it doesn't work with 6.1, that's as close as I had. And the
response I got was, paraphrasing, "use 6.3, it's known to work".  There
seemed to be a complete lack of recognition of it being a problem with
the book.  Frankly, the resistance I experienced getting general
recognition the problem should be taken seriously hasn't engendered much
desire to volunteer.  I don't care for this sort of welcome.

> I started in 1965.  I still have my FORTRAN (not II) manual.  I also
> used an IBM-1620.  The mass storage device was punched cards. Also IBM
> 7090/7094, IBM 360 Series 60, and CDC 6600.

Amazing what was done with those machines, the ground they broke! I used
to have a homage page up about the 1620. Nit: I don't believe there was
a model 60.  30, 40, 50, 65/67, 44, 75, 85, 91/95, 25.  The 20 & 195
were illegitmate children. http://en.wikipedia.org/wiki/IBM_System/360

> No, not petulance, frustration.  You report a problem doing things
> *your* way and we *did* try to help.  We ask for your help in
> confirming the problem in a way we can duplicate and you say no.  Who
> is being petulant here?

Really?  Who made a 6.1 system so they could duplicate my problem?  You,
collectively, told me to abandon 6.1, where the problem was. Use a
different kernel, different gcc, you said.  How does that duplicate the
problem?  I have YET to read that anybody successfully compiled 6.6 with
the HSR's.  Paraphrasing again, I heard, "Oh, those systems are too old
to consider.  It's a waste of time."  I was told the problem had been
reported before, but nobody knew why.  Was there an effort?  Pete Jordan
discovered a workaround and was criticized.  Do you know what gcc branch
he was using?  gcc-3.x!  It seems to me your belief the book was always
right, just following the book always leads to success was mistaken.
FBBG is good advice, but not infallible.

Well, last night I thought maybe I could easily try the Stage1/Pass2
gcc-
4.3.4 compiler in /tools to rebuild the host's linux-2.6.18 by prefixing
my path with /tools/bin, to see if that would bring up the CC-
STACKPROTECTOR option, but it didn't.

> My argument is that we will fix problems that we can confirm.  I still
> don't see why you won't help us help you.

How?  You don't want to solve the problem of building glibc with gcc-
3.x.  Use a workaround like Pete Jordan's until the gcc-4.x can be
installed, at which time glibc can perhaps (2.6.18 may not work) be
correctly recompiled?  You want to bypass it by using 6.3!

I'm still dealing with the fundamental problem.  Where is your help?
Your version of help reminds me of the IMF's--we'll loan you money, only
if your country will adopt the financial system we want.

> >> The Host System requirements may indeed be too low for LFS 6.6
>
> Updated to the packages in LFS-6.3, known to work for LFS-6.6  Added
> erratum to website.

Finally!

>
> I see Bruce has now changed the versions to those he knows to work.
> I'm slightly disappointed by this, because old versions were usually a
> lot smaller and faster, but they're still very old, so no real worries
> about treating them as minima.

Indeed, that's why I'm so slow to upgrade--I use old hardware.  I need
to avoid all the "creeping featuritis" and fluff that accumulates.  I
built this 6.1 system on a "classic" Pentium-MMX 233MHz.  I had a few
6hr compiles, two over 9hrs, and one 12hrs, IIRC.  But when I was done,
it worked.  It was "good enough".

> I'll repeat what I've said over the years - if you have a self-built
> host system (and ideally for any installed host system, but some
> distro setups can make it unnecessarily hard), the first step to
> installing a new system should be to use the kernel you are intending
> to boot.  It gives you a .config that you can test while you still
> have sufficient applications.

So I had this 2.6.17 system I couldn't build 6.6 on.  I needed a newer
kernel.  What do you suppose I found in the 6.6 tarballs?  Oh, it
compiled with my glibc-3.4.3, but it had too many errors in init to be
usable, so I was forced to patch to 18.  Which I saw in the book was
all I needed.

> we would probably tell you one adequate previous version from which to
> build - not very useful for anyone trying LFS for the first time.

Isn't that about what you just did with the 6.3 erratum?  I'm still
working my workaround path, with suggestions from Pete Jordan, not the
mailing list--who don't seem interested.

>  So, you subscribe to the idea that the better is the enemy of the good-
>  enough ?  I'm surprised you use the linux kernel.

I don't follow that line of reasoning.

> We do what we think is important.  Sometimes, LFS is a minor part of
> what matters to us.

I thought the purpose of a support mailing list was to help users solve
THEIR problems.  My mistake, I guess.

> However, my personal view is that you built LFS several years ago and
> have not kept in touch with what has changed.  You also seem to have a
> "not important for me" attitude to updating for vulnerabilities
> (evidence: your kernel version) which doesn't give me any confidence
> that you are liekly to do the right thing with regard to systems that
> people use - it might be an adequate view for your own system, but it
> smells of poor practice.

HAHAHAHA, that's rich!  It's all MY fault that I believed the HSR's!  Do
you get many volunteers when you treat people this way?

> You also seem to think paid-for and volunteer development is similar.
> It isn't.  In one, the person or organisation with the money can
> attempt to make the decisions.  In the other, people have to agree -
> those of us who decide we do not wish to spend our time in arguments
> that we don't find useful, may decide to reduce our presence here.

That's a two-edged sword, if you haven't noticed.

> If nobody is willing to test on multiple old hosts (and old distros)
> that used to be adequate, I'd rather see an updated book that we know
> works for most people, than one that will be released when it's passed
> QA but is already long out of date.

How far do you suppose we'd have to look to find someone with an old
6.1 system to try the 6.6 book.  Yeah, that's probably too much to
expect, ain't it?

> replying to Bruce), particularly the last sentence.  It's very easy to
> cause offence, and you seem to manage it well.

It has to be taken, doesn't it?

> I don't have problems with the idea of trying to keep old mainstream
> hardware working.  But part of the process for that is updating the
> software.

Darn, and here I thought trying to upgrade from 6.1 to 6.6 was exactly
that!  And the only reason I decided to do that was I found an
affordable hardware upgrade from my MMX-233's to these Tualatins.
-- 
Paul Rogers
paulgrogers at fastmail.fm
http://www.xprt.net/~pgrogers/
Rogers' Second Law: "Everything you do communicates."
(I do not personally endorse any additions after this line. TANSTAAFL :-)

  	

-- 
http://www.fastmail.fm - Faster than the air-speed velocity of an
                          unladen european swallow




More information about the lfs-support mailing list