On Sat, Nov 22, 2014 at 05:00:20PM -0800, Paul Rogers wrote:
> Allow me to address your last point first, the scripts that I make/use:
> 
[...]
> As I mentioned initially, this is my fifth LFS build, these scripts have
> been used many times, just minor mods as are in the book, version to
> version.  There is one function that is very, very close to just what's
> in the book.  I'm very, very careful to not depart from what works in
> new and original ways.  My package manager uses before/after snapshots,
> totally out of the way during install.
> 

 OK, I was just pointing out that we cannot know what is in your
scripts.  I will assume you have made all the changes from the
previous version of the book (me, I often miss something here or
there - either it blows up and is obvious, or I can go for a long
time without noticing the difference.  So, it always pays to review
if there is any doubt.

> I registered my first LFS build in 2004.  I've seen some of the stuff
> you get on the mailing list.  You're right to be a little cautious, even
> if abi_check failures aren't a common problem.  I'll take responsibility
> for my scripts, but they DO work on the KISS principle. [Heck, that's
> why I use LFS!!! ;-) ]
> 
> > > I notice your test cases were build for x86_64, mine for i686.  As I
> > > said, I did a C&P into a wrapper script, but my error causes the
> > > "make check" to drop out.  That alone makes it seem pretty
> > > important.
> >
> > By "drop out" you mean, "only got this far, then stopped", or "failed
> > after reporting all the results ?"  The latter sounds like normal
> > behaviour for non-ignored errors.
> 
> I use this: (make -k check 2>&1 | tee log.check && exit $PIPESTATUS) &&
> The $PIPESTATUS, with "#!/bin/bash -e" means if the make produces an
> error code, bash drops the script, even if tee produces a 0.  That's
> what's happening--make check is erroring out, leaving a log file.
> 

 My current scripts (probably my third or fourth major rewrite - I
needed to improve my logging of what got installed - mostly fail on
any non-zero return (there are a few places where I had to turn that
off, for conditional commands).  And I also build lfs-dev, where
test failures often happen as versions change..  So, many of my test
invocations end ' || true'.  In other words, I do not regard a
failing test, even in the toolchain, as necessarily catastrophic -
for me, what matters is whether the resulting system does what I
want.  And if a test fails, a non-zero status is to be expected.

> > I will also point out that 7.2 was something like 2 years ago and few
> > people here are likely to remember any details.  On "one unexpected
> > failure" I agree with Bruce : Carry on.
> 
> I would have, but if the make check errors out, that's not a normal
> unexpected error.  make really wanted to quit!
> 

 It's your system, and your time.  I used to think that 'perfect'
test results were good.  But all the tests do is check for a
recurrence of past errors.  And with different hardware, different
things can cause tests to break (e.g. Bruce's toolchain results
often look better than mine).

 Also, I repeat what Bruce said - for one or two errors in the tests
for a toolchain package, carry on.

> > But two more comments below :
> >
> > > Using /usr/local/src/gcc-4.7.1/libstdc++-v3/testsuite/config
> > > /default.exp$
> >
> > /usr/local ??
> 
> I have chosen to do all my development in /usr/local/src.  That's OK
> with LSB, isn't it?  Once the target FS is created, that's where I
> move my scripts, tarballs, patches directories.  Is that any problem?
> Never has been.
> 

 Building there is ok, if somewhat unusual.  It just seems like an
odd place to build things which are part of the base system, and
will be installed in /bin or /usr/bin.  I did not pay too much
attention to the files, or I would have realised it was the build
tree.

> > If you are following the book (and the references to /usr/local in
> > what you pasted make me doubt that), people _might_ be able to help on
> 
> Not rigidly, no, but in all important respects.  (Look, I'm not out to
> waste my time or yours, and my profrssional career was in computing and
> computing management, beginning as a programmer in 1966.)
> 
> > the e2fsprogs issue.  But only if you explain the problem. Saying it
> > "died in a mysterious way" is not a useful error report.
> 
> The way it failed made me suspect the compiler.  I need a reliable
> compiler before wasting time with e2fsprogs.
> 

 Other things can happen (heat, component failure, transient errors).
A quick look at an x86_64 log for e2fsprogs-1.42.5 suggests that it
uses 'CC' as the compiler, not 'CXX', so a libstdc++ ABI change seems
unlikely to impact e2fsprogs.

ĸen
-- 
Nanny Ogg usually went to bed early. After all, she was an old lady.
Sometimes she went to bed as early as 6 a.m.
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

http://en.wikipedia.org/wiki/Posting_style

Reply via email to