17 Jul
2023
17 Jul
'23
6:41 p.m.
On Mon Jul 17 16:44:37 2023 +0000, Jacek Caban wrote:
That works too, yes, but is there something wrong with my approach?
Not that it matters, but there's 1 less flag to worry about, and same amount of checks (check liveness of key or map, instead of the new flag). It should possible yes, but that would need a pointer to WeakMap instead of the flag. Anyway, the part I don't like is the extra pass (which should be possible to avoid with reference tracking).
I tried with the flag, but it looks like we will need pointer to WeakMap anyway, for jsdisp_free to be able to remove the entry (which needs access to the WeakMap's rb_tree).