You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We still had some discrepancies between what the code thought could be
computed for a layout, and what we actually could compute layout for.
This hopefully makes those two computations now always give the same
answers (with the assistance of the layout cache).
Fixes#43303
(cherry picked from commit 30fe8cc)
if (ty->name->mayinlinealloc&& (ty->isconcretetype|| ((jl_datatype_t*)jl_unwrap_unionall(ty->name->wrapper))->layout)) { // TODO: use jl_struct_try_layout(dt) (but it is a safepoint)
253
+
if (ty->name->mayinlinealloc&&jl_struct_try_layout(ty)) {
254
254
if (ty->layout->npointers>0) {
255
255
if (pointerfree)
256
256
return0;
@@ -264,7 +264,7 @@ int jl_datatype_isinlinealloc(jl_datatype_t *ty, int pointerfree) JL_NOTSAFEPOIN
0 commit comments