[Pkg-sympa-devel] Bug#711604: Hash randomization test failures

[Niko Tyni]

On Wed, Jan 08, 2014 at 10:02:13PM +0000, Dominic Hargreaves wrote:

(cc'ing the sympa maintainers; this bug in libmime-lite-html-perl
is keeping sympa out of testing.)

> > So this package needs to be checked to see whether the test failures
> > are definitely random, and that the random breakage is purely in the test
> > suite. If so, the severity of this bug can be downgraded to important.
> 
> This doesn't seem to be the case here; this package fails to build
> for me 12 times in a row. Updating the package to 1.24 (which looks
> like it would be a worthwhile update anyway) doesn't change anything.
> And I've just re-verified that the package does build on wheezy.

It's certainly hash randomization as it builds with PERL_HASH_SEED=0. 
The test suite relies on the old hash ordering in several places.

I started fixing libmime-lite-html-perl by adding things like

--- a/t/50generic.t
+++ b/t/50generic.t
@@ -36,6 +36,7 @@ foreach my $it ('cid', 'location', 'extern') {
     my $url_file = "file://$f";
     print $url_file,"\n";
     my $rep = $mailHTML->parse($url_file, "A text message");
+    $rep->field_order('from', 'to', 'subject');
     $rep = $rep->as_string;
     $rep =~s/^Date: .*$//gm;
     my @bound;

but ran into trouble with t/ref/index.cgi.eml2.cid which has an empty
line between headers that MIME::Lite::field_order() loses (intentionally
or not.)
 
> This is currenly keeping sympa out of testing.

Not sure if the sympa maintainers are aware of that, so cc'ing them.
-- 
Niko Tyni   ntyni at debian.org