On 12/12/2010 07:41 AM, Jim Meyering wrote: > That sounds good, assuming it triggers the bug reliably for you. > I was hoping to find a way to reproduce it without relying on gensort, > but won't object if you want to do that.
In my attempts to reproduce the problem, it's pretty flaky. I think it depends on how busy the operating system is. Sometimes I'd get failures all the time; sometimes, almost never. (This was with valgrind; I had much less luck without valgrind.) Anyway, I pushed this, which seemed to work well enough on my host. It prefers gensort if available, but falls back on seq+shuf if not. >From 63d1b425976ccc0b89159d743e33eb5da634de3c Mon Sep 17 00:00:00 2001 From: Paul Eggert <egg...@cs.ucla.edu> Date: Sun, 12 Dec 2010 13:38:19 -0800 Subject: [PATCH] tests: test for access to stale thread memory * tests/misc/sort-stale-thread-mem: New tests. * tests/Makefile.am (TESTS): Add it. --- tests/Makefile.am | 1 + tests/misc/sort-stale-thread-mem | 44 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 45 insertions(+), 0 deletions(-) create mode 100755 tests/misc/sort-stale-thread-mem diff --git a/tests/Makefile.am b/tests/Makefile.am index b573061..f7a8af8 100644 --- a/tests/Makefile.am +++ b/tests/Makefile.am @@ -238,6 +238,7 @@ TESTS = \ misc/sort-month \ misc/sort-rand \ misc/sort-spinlock-abuse \ + misc/sort-stale-thread-mem \ misc/sort-unique \ misc/sort-unique-segv \ misc/sort-version \ diff --git a/tests/misc/sort-stale-thread-mem b/tests/misc/sort-stale-thread-mem new file mode 100755 index 0000000..c4f4fcb --- /dev/null +++ b/tests/misc/sort-stale-thread-mem @@ -0,0 +1,44 @@ +#!/bin/sh +# Trigger a bug that would cause 'sort' to reference stale thread stack memory. + +# Copyright (C) 2010 Free Software Foundation, Inc. + +# This program is free software: you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation, either version 3 of the License, or +# (at your option) any later version. + +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. + +# You should have received a copy of the GNU General Public License +# along with this program. If not, see <http://www.gnu.org/licenses/>. + +# written by Jim Meyering and Paul Eggert + +. "${srcdir=.}/init.sh"; path_prepend_ ../src +print_ver_ sort + +expensive_ + +valgrind --help >/dev/null || skip_ "requires valgrind" +test "$(nproc)" = 1 && skip_ "requires a multi-core system" + +# gensort output seems to trigger the failure more often, +# so prefer gensort if it is available. +(gensort -a 10000 in) 2>/dev/null || + seq -f %-98f 10000 | shuf > in || + framework_failure_ + +# With the bug, 'sort' would fail under valgrind about half the time, +# on some circa-2010 multicore Linux platforms. Run the test 10 times +# so that the probability of missing the bug should be about 1 in +# 2**100 on these hosts. +fail=0 +for i in $(seq 100); do + valgrind --quiet --error-exitcode=3 \ + sort -S 100K --parallel=2 in > /dev/null || + { fail=$?; echo iteration $i failed; Exit $fail; } +done -- 1.7.2
