Bug 48615

12:46 < jonnylamb> smcv: why does tp_presence_mixin_init (not class_init) call g_TYPE_set_data and not g_OBJECT_set_qdata?
12:49 < smcv> jonnylamb: good question
12:49 < smcv> jonnylamb: it's used in macros in the header, so can only be changed in next
12:50 < smcv> jonnylamb: tbh it's pretty harmless, every object of the same type ought to have the mixin at the same offset...
12:50 < jonnylamb> yeah, it shouldn't affect anything.
12:50 < jonnylamb> I'll open a bug.
12:50 < smcv> it should be set in class_init tbh
12:50 < smcv> unless you have a devious reason why you want it to vary by instance
12:51 < jonnylamb> g_type_set_qdata is right for class_init, but not for init..?
12:51 < smcv> but I hope you don't that'd be crack
12:51 < jonnylamb> if there's per instance data.
12:51 < smcv> g_type_set_qdata is right if there isn't per-instance data
12:51 < jonnylamb> right.
12:51 < smcv> class_init is also right if there isn't per-instance data, to enforce that it's always the same
12:52 < jonnylamb> oh there actually isn't for presence mixin.
12:52 < smcv> the data being set here is: given a MyConnection, what's the offset of the sub-struct with the per-instance data? 
12:52 < smcv> and *not*: what's the sub-struct for this instance?
12:52 < smcv> so there's no sane reason for it to vary per instance I don't think?