lib/perldebug.t: puzzling test failure

Yesterday evening, in:

commit b6e9e118a4133eb96eb5e5de209ea6a37fd15bfe

... I applied a set of 5 patches to blead.  Before I did so, I ran 'make 
test_harnes' on my Linux x86-64 machines and got PASS.  I pushed to 
origin.  I subsequently smoke-tested blead on my FreeBSD 10.3 VM and 
also got PASS.

Today I ran make test_harness on both machines at the point where HEAD is:

commit fcbb262bfa8a07198e3121b8daca0e18610694d2

I got PASS on FreeBSD, but FAIL on Linux.  Here is the failure on Linux:

#####
$ cd t;./perl harness -v ../lib/perlbug.t; cd -

ok 1 - config information dumped with -d
ok 2 - full config information dumped with -d -v

[snip]

ok 20 - attachment included in fake bug report
not ok 21 - [perl \#128020] long body lines are wrapped: maxlen 1133
ok 22 - long attachment lines are not wrapped: maxlen 1200
# Failed test 21 - [perl \#128020] long body lines are wrapped: maxlen 
1133 at ../lib/perlbug.t line 154
Failed 1/22 subtests
#####

My first hunch was that this failure was introduced into blead 
subsequent to b6e9e11.  However, it turns out that even when I *re-test* 
at that commit, I get the failures in lib/perldebug.t documented above.

At this point I can't explain the discrepancy in test results for blead 
in the two different environments, nor can I explain why am now getting 
a failure on Linux at a commit point at which I previously got no failures.

Is anyone else observing any problems with lib/perldebug.t?

Any suggestions?

Thank you very much.
Jim Keenan

• lib/perldebug.t: puzzling test failure by James E Keenan
• Re: lib/perldebug.t: puzzling test failure by James E Keenan
• Re: lib/perldebug.t: puzzling test failure by James E Keenan