Re: GtkWindow fails to pass make test

[Ross McFarland <rwmcfa1 neces com>]

On Fri, 2003-08-15 at 18:23, muppet wrote:
> i guess it was a different version of Test::More on RH9 than i was 
> using on RH8 and Win2k, and that's why i never saw the error.  in fact, 
> i now recall the output showing Gtk2::Window=HASH(0xgunk) as a test 
> name, which i didn't notice at the time.  the first form would've 
> resulted in that since ok() prints its second arg as a string.  what 
> actually caused it to barf on your systems?  could you track that down? 
>   i would suspect that $win would still be stringified, why isn't it?

output:

t/01.GtkWindow............ok 43/87Invalid value for shared scalar at
/usr/lib/perl5/5.8.0/Test/Builder.pm line 319.
WHOA!  Somehow you got a different number of results than tests ran!
This should never happen!  Please contact the author immediately!


this is going to be a wild guess, but here goes. i'd venture that the
version of Test::More on redhat 9.0 uses perl ithreads in the course of
testing. 

i've gotten this "Invalid value for shared scalar ..." error message
before when i was working on another project involving trying to using
threads sharing blessed objects (that are really C underneath). anyway
if you read all of the details about the ithreads stuff you'll find out
that you can't share objects that have C things in them, or maybe even
can't share blessed objects (i'm not remember exactly at the moment).
there would be some non-trivial issues to deal in such cases. anyway
that message happens when you use the shared attribute on such an
object. since $win is a blessed object with C underneath and you used it
where Test::More was expecting a plain old scalar with a string it it
you get the above message. (i'm making this up as i go, but it's at
least an educated lie^Wstory)

now as for why Test::More got the error message about different number
of results than test i don't know i've seen that one several times when
things really die, perhaps even core dump.

-rm