Hello porters,
WWW::Mechanize::Shell has a failure on 5.10-to-be, but the
same test
works on 5.8.8 [1]. Unfortunately, I haven't been able to
whittle down
the test to something manageable that still exhibits the
failure
behaviour, and I've already spent the evening with that.
So, I beseech you to run a binary search when the test
starts failing in
the hope that the patch reason hints at the deeper cause of
the test
failures.
A bit of the situation:
The W:M:S code wraps LWP::UserAgent::request using
Hook::LexWrap.
The logging code in W:M:S is overwritten from the test
program, to
instrument it.
On the first round, it all works, LWP::UserAgent::request is
called and
the wrapping logging code in W:M:S resp. the instrumented
test code are
called. (Test 6)
On the second round, the wrapped LWP::UserAgent::request
still gets
called, but the test instrumentation code isn't, anymore.
(Test 14)
All my attempts to recreate this separately in a nice
self-contained
file have been in vain, and the same code has been working
for a long
time under 5.8.x, so I'm at a loss as to what's wrong with
it under
5.10-to-be ...
-max
[1] http://www.nntp.perl.org/group/perl.cpan.
testers/2007/10/msg671909.html
http://www.nntp.perl.org/group/perl.cpan.
testers/2007/10/msg671496.html
|