|
From: | Matthew Woehlke |
Subject: | Re: a coreutils release is imminent |
Date: | Wed, 21 Mar 2007 15:52:39 -0500 |
User-agent: | Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.0.10) Gecko/20070221 Thunderbird/1.5.0.10 Mnenhy/0.7.4.0 |
Matthew Woehlke wrote:
Jim Meyering wrote:Matthew Woehlke wrote:sort-compress still fails with a low process limit (e.g. my OSF system)... I thought this was fixed? It also failed on ia64/hpux. There is (still) a really annoying problem with the perl detection that leads to "can't find strict.pm" problems; maybe we could check that perlThis is the first I've heard of such a problem. What version of Perl is installed there? I may simply raise the required version number to one for which "use strict;" works. In the mean time, you can work around that by manually removing the "use strict" lines.That's what I was thinking. Hmm, 'perl --version' hangs waiting for input, and 'perl -V' is a usage error. So... how do I *tell* what version? (Clearly I'm not a perl expert :-).)Hmm, ok, I see this in the error: /freeware/lib/perl5/alpha-dec_osf/5.00401 ...so maybe it is 5.00401? [snip] methinks my perl is b0rk.
Ok, I finally bit the bullet and figured out how to build perl. (Talk about ugly build systems... I feel unclean.) It croaked once on tty-eof but skipped it after I nuked the dir and started clean from the tarball, so that was probably because I installed a working perl after running configure.
Anyway, sort-compress is still "broken". Wasn't there talk of limiting the number of forks based on getrlimit?
It passed on another re-run, then failed again... which as I recall is expected; the failure is intermittent because it depends on *which* fork fails. (There were still plenty of /fork/ failures* on the "successful" run, just not a "critical" one.)
So I guess sort-compress is the only remaining severe failure, failing on OSF (due to the extremely low 64-process-limit) and at least once on ia64/HPUX. Unfortunately this test seems to fail irregularly.
However, I am seeing one more failure on OSF: touch/empty-file. Both the local clock and the NFS clock incorrectly report PST, but otherwise appear to be in sync:
$ touch fubar ; ls --full-time fubar ; date ; rm -f fubar -rw-r--r-- 1 user group 0 2007-03-21 12:22:59.009964000 -0800 fubar Wed Mar 21 12:22:59 PST 2007Increasing SLEEP_SECONDS as suggested didn't help (although I only tried 5, not 3602...).
Here's the verbose output (which I admit I am looking at, and don't understand...):
+ DEFAULT_SLEEP_SECONDS=2 + SLEEP_SECONDS=3 + fail=0 + : . + framework_failure=0 + rm -rf ./a ./b ./c + 1> ./a + test -f ./a + 1> ./b + test -f ./b + 1> ./c + test -f ./c + test 0 = 1 + echo sleeping for 3 seconds... sleeping for 3 seconds... + sleep 3 + touch ./a + ls -t ./a ./b + set x ./b ./a + test x ./b ./a = x ./a ./b + fail=1 + echo sleeping for 3 seconds... sleeping for 3 seconds... + sleep 3 + touch ./b + ls -t ./a ./b + set x ./b ./a + test x ./b ./a = x ./b ./a + touch - + 1< ./c 2> /dev/null + rm -rf ./a ./b ./c + test 1 != 0 + cat + 1>& 2 0<< -- MatthewCongratulations! You've won a free trip to the future! All you have to do to claim your prize is wait five minutes...
[Prev in Thread] | Current Thread | [Next in Thread] |