Front page | perl.perl5.porters |
Postings from January 2018
# New Ticket Created by (Andreas J. Koenig)
# Please include the string: [perl #132790]
# in the subject line of all future correspondence about this issue.
# <URL: https://rt.perl.org/Ticket/Display.html?id=132790 >
Three little birdies twittered the story to me, I lend them my voice
to let you know about it.
This is not about a bug in Storable. It's about a bug in old threaded
perls and how this bug unfolded over time.
The bug in perl itself was fixed in v5.19.2-116-g82b84d0.
Test::More 1.001014 did not trigger the bug, it was only exposed
within dist/Storable/t/blessed.t through the changes in Test2.
But Test2 developed a workaround for the bug. Unfortunately that
workaround could only be applied to perls >= 5.14 not older ones. The
workaround is in EXODIST/Test2-0.000035-TRIAL.tar.gz. It can be
enforced by setting the environment variable T2_CHECK_DEPTH=1. So
older perls can run
T2_CHECK_DEPTH=1 make test
in order to pass all Storable tests.
There is a contrived workaround for blessed.t itself. It avoids
triggering the bug by changing compilation order. I didn't understand
the birdies' explanation on the inner workings of this solution. Here
it is:
% diff -U1 t/blessed.t{~,}
--- t/blessed.t~ 2018-01-30 20:44:59.935027614 +0000
+++ t/blessed.t 2018-01-30 20:51:01.133495863 +0000
@@ -18,4 +18,2 @@
-use Test::More;
-
use Storable qw(freeze thaw store retrieve);
@@ -28,2 +26,4 @@
+use Test::More;
+
{
If the patch would be applied, we could get rid of an inconvenience
for everybody who taps into this test failure with old perls in the
future.
--
andreas
Thread Previous