Replace ct_is_hidden with asserts #7
Conversation
(Kumar will fix this soon!)
|
Looks like CI passes - this is ready for review. |
src/c/realize_c_type.c
Outdated
|
|
||
| assert(!((ct->ct_flags & CT_IS_OPAQUE) || | ||
| (ct->ct_flags_mut & CT_UNDER_CONSTRUCTION))); | ||
| // should have been unset during init for the ctype |
There was a problem hiding this comment.
I'm not sure I understand this comment. I'd think that the assert below is because opaque structs never have lazy field lists.
There was a problem hiding this comment.
I'd think that the assert below is because opaque structs never have lazy field lists.
This is what I was going for. I'll edit the comment.
| /* This is called by force_lazy_struct() in _cffi_backend.c */ | ||
| assert(ct->ct_flags & (CT_STRUCT | CT_UNION)); | ||
|
|
||
| assert(!(ct->ct_flags_mut & CT_UNDER_CONSTRUCTION)); |
There was a problem hiding this comment.
This seems like it might be a temporary assert. I think we can end up with concurrent calls to do_realize_lazy_struct() on the same CTypeDescrObject
There was a problem hiding this comment.
I think @kumaraditya303 plans to turn the CT_UNDER_CONSTRUCTION flag into a uint8 member of the CTypeDescrObject struct to facilitate reading and writing to it using atomic operations implemented using something like npy_atomic.h, which is a bare-bones version of pyatomic.h.
When that refactor is ready this assert can be moved or deleted when we're sure concurrent calls to b_complete_struct_or_union are safe.
2 participants
Followup for #5 to more fully disentangle
CT_IS_OPAQUEfromCT_UNDER_CONSTRUCTION.