On Thu, Jun 27, 2013 at 6:18 AM, Tony Cook <tony@develop-help.com> wrote: > On Wed, Jun 26, 2013 at 06:15:34PM -0700, Tony Cook via RT wrote: > > What actually sets PL_sv_consts[SV_CONST_*] ? I assume the result of > > newSVpv_share() here is leaking, causing the op/svleak.t test to fail. > > Actually, the SVs stored in PL_sv_consts[] would produce similar leaks > until the tests were adjusted. > Thanks for looking into this. I didn't look at your other set of comments, but I suspect that leak tests started to fail since I switched to "on demand" mode. For sure in this mode first call that uses some PL_sv_const "leaks" one SV. It leaks it into PL_sv_consts array. Either test should do bogus run to prefill all slots in the array or it can ignore one SV. For sure I should take another look at the code, from your questions and with fresh mind I see code is wrong in a few places :/ I will take closer look during weekend, 4 July or my holidays that start on 7 july. > Tony > -- Best regards, Ruslan.Thread Previous | Thread Next