a bug hunt opportunity

The change #19450 introduced Tie::File 0.95.

The change #19498 introduced a fix from Sarathy to plug a filehandle
leak (beneficial for all platforms, not just win32) in this Tie::File.

Unfortunately, the fix also seems to have unearthed some sort of bug
(in glob refcounting, maybe, possibly?)  The DESTROY now (because of
the #19498) closes the filehandles opened by Tie::File, but that made
threaded builds spew black smoke because the tests 51-58 in Tie::File's
09_gen_rs.t test trying do a seek on an already closed filehandle
(or, at least, that is that what *seems* like to be happening.)

Therefore I now (change #19509) have those tests marked as "not ok TODO".

Your mission, should you choose to accept it, is to distill down the
bug (whereever it is) unearthed by Tie::File + #19450 into a simple
test case that doesn't require all of Tie::File around.  Once found,
*fixing* the bug wouldn't be a bad idea, either :-)

A possible lead in the case is that only threaded builds started
failing, not singlethreaded ones.

-- 
Jarkko Hietaniemi <jhi@iki.fi> http://www.iki.fi/jhi/ "There is this special
biologist word we use for 'stable'.  It is 'dead'." -- Jack Cohen

• a bug hunt opportunity by Jarkko Hietaniemi
• Re: a bug hunt opportunity by Enache Adrian
• Re: a bug hunt opportunity by Gurusamy Sarathy