On Sat, May 15, 2004 at 01:17:27PM -0500, Andy Lester wrote: > >Not having a threaded version of perl handy right now (I am home, it > >is at work), I cannot test this right now. The changes I had made were > >only to change a few tests to 'cmp_ok' and add test names, which > >should not affect things. However in the last update (prior to this > >one), I did change the order in which things were loaded when I > >converted this to use 'skip_all'. I did not think this would affect > >things, but without a threaded perl handy and limited knowledge of the > >details of threaded perl I suppose I should not have assumed.
I wouldn't have thought it would either. (And it's been a while since I tested with threaded perl so it's possible the problem predates your changes. I should check that at some point.) > It is a problem with Test::Builder and threading, and it relates to, I > believe, overloaded stringification of objects. At least, that's what > it was when I was having a similar problem in WWW::Mechanize's tests > and it was using an automagically stringified URI::URL object. I had > to explicitly call $obj->as_string and then it was happy. > > YMMV. I think our mileage does vary as I can't think of any overloaded stringification in the DBI or these tests. Tim.