12:13 <antocuni> fijal: another possibility is to completely change the semantics of HPy_AsStruct and say that writes are not immediately visible

12:13 <antocuni> and that e.g. you need to "commit" them before they are visible to other places

12:14 <antocuni> this would allow us to return a scratch memory which will be copied to the "real" memory upon commit, so we don't require pinning

12:15 <antocuni> but I'm not sure it's a good idea because 1) you end up doing lots of copies and 2) on CPython changes are immediately visible, so there will be code which works fine on CPython but not on ther implementations

12:15 <antocuni> the debug mode might help with (2), but still

12:17 <antocuni> fangerer: are we going to have a meeting on thursday?

12:18 <fangerer> yes, of course

12:18 <fangerer> I'll send out a reminder today

12:23 <fijal> I'm not sure I have time to talk today

17:10 <Hodgestar> Hmm. squeaky_pi went away before we could answer.