Agreed on the leaking. I've had a quick look through what is there at the moment and have concluded that I'm best to start by refactoring the test suite.
The current test suite has two separate copy and pasted test scripts for the XS and pure perl cases. If I diff between the various releases, it appears that they are being synchronised largely by hand and in a fairly patchy fashion. So I've created a t::lib::Test module and I'm extracting the test code for each function into it's own test_uniq, test_whatever function, and changing both test scripts to use the common testing code. Once that is done, it should simplify the process of adding new tests for the various problems in only one place. After that, I might look at re-applying some of the simpler fixes that don't require XS-fu and expanding the tests for the relevant functions to test against more cases. There's also at least one CPAN Tester throwing out-of-memory errors in the tests, so I'm going to audit the size of the lists it uses to test with to try and reduce that problem. I'll assume that early in the new year you'll swoop in and do the hard stuff, hopefully the house will be clean and ready for you by then. Adam K On Wed, Dec 1, 2010 at 5:13 PM, Jens Rehsack <rehs...@googlemail.com> wrote: > 2010/12/1 Adam Kennedy <a...@ali.as>: >> Since my releases seem to have "regressed" issues that were "fixed" in >> the no-longer-existing 0.23 (and created problems for some people) my >> plan to not have to touch the actual code seem to have been derailed a >> bit. >> >> I'm going to see if I can apply some of the fixes suggested in RT, or >> reapply some of the fixes from 0.23. If I can't knock out a decent >> test-passing dev release I'm then going to rip out the leaking XS >> versions of the functions as damage control. >> >> Alternatively, if Jens plans to work on it straight away then I'll yield. > > Current plan of order of tasks: > - finish current task (SQL::Statement / DBI-bundled Pure-Perl-DBD's) > - Proc::ProcessTable code merge from current $work on libstatgrab > (make it thread-safe, all fields available on all OS, ...) > - List::MoreUtils (first (re)write tests to cover leak-test, then fix > and go on business as usual) > >> I'll see how far I can get now, but won't push a stable release >> without Jens' approval. > > The meanwhile deleted releases had other issues - all XS related. > I studied a bit since I got maintainership - should be easy to fix >:-) > > From my point of view, nothing beyond 0.22 didn't work for real. > I hope I get P::PT managed this year, but could require some weeks > at start of 2011. > > So you have a rough time line. I suggest you don't try to fix the > leaks without tests :) > > Jens > >> Adam K >> >> On Tue, Nov 30, 2010 at 11:27 PM, David Golden <xda...@gmail.com> wrote: >>> On Mon, Nov 29, 2010 at 4:57 PM, Adam Kennedy <a...@ali.as> wrote: >>>> I'll try to replicate on a 5.8.8 system later today, but this should >>>> not change your priority (although the existence of a Critical >>>> Twiki-breaking bug might). >>> >>> Isn't it just so tempting to stall on 5.8.X bugs until April 20, 2011 ... >>> :-) >>> >>> -- David >>> >> >
