Eric Wong wrote:
> ST_DEBUG doesn't reveal anything nor does switching to
> QUADRATIC_PROBE.  Waiting for valgrind which is very slow
> and I need to go afk to do other things...

Could be a GC problem related to coverage (also [Bug #14334])
since there's minimal changes in st.c from v2_4_0..v2_5_0 and
I'm not hitting it with the 2.4 branch.

Nothing jumped out at me from reading diffs, I don't think
the guard I proposed in [ruby-core:85352] is effective, even

Bisecting now, and going mostly away from computers for a bit...

git bisect start v2_5_0 v2_4_0
git bisect run /tmp/bisect.sh

==> /tmp/bisect.sh <==
#!/bin/sh
git clean -qfdx || exit 125
autoconf >/dev/null || exit 125
pfx=/tmp/bisect
rm -rf $pfx || exit 125
./configure  \
	--prefix=$pfx --disable-install-doc >/dev/null || exit 125
make -j8 >/dev/null || exit 125
make update-gems || :
make -j8 install >/dev/null || exit 125
export PATH="$pfx/bin:$PATH"
(
	# copied a bunch of gems over to /tmp after
	# gem install rspec rspec-expectations coveralls
	cd /tmp/gems && gem install *.gem
)
gem unpack /tmp/gems/thread_safe-0.3.6.gem || exit 125
cd thread_safe-0.3.6 || exit 125

# run twice, sometimes won't crash
rspec -I lib spec || exit 1
rspec -I lib spec || exit 1
-- 
EW

Unsubscribe: <mailto:ruby-core-request / ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>