Several issues with pointer types are fixed:
Prior to this commit, Zig would not canonicalize a pointer type with
an explicit alignment to alignment=0 if it matched the pointee ABI
alignment. In order to fix this, `Type.ptr` now takes a Target
parameter. I also moved the host_size canonicalization to `Type.ptr`
since target is now available. Similarly, is_allowzero in the case of
C pointers is now treated as a canonicalization done by the function
rather than a precondition.
in-memory coercion for pointers now properly checks ABI alignment
of pointee types instead of incorrectly treating the 0 value as an
alignment.
Type equality is completely reworked based on the tag() rather than the
zigTypeTag(). It's still semantically based on zigTypeTag() but that
knowledge is implied rather than dictating the control flow of the
logic. Importantly, this fixes cases for opaques, structs, tuples,
enums, and unions, where type equality was incorrectly returning based
on whether the tag() values were equal.
Additionally, pointer type equality now takes into account alignment.
Because we canonicalize non-zero alignment which equals pointee type ABI
alignment to alignment=0, this now can be a simple integer comparison.
Type hashing is implemented for pointers and floats. Array types now
additionally hash their sentinels.
This regressed some behavior tests that were passing but only because
of bugs regarding type equality.
The C backend has a noticeable problem with lowering differently-aligned
pointers (particularly slices) as the same type, causing C compilation
errors due to duplicate declarations.
We were using the array type, not the element type. Also, we should do
the sentinel comparison after we verify that the element types of both
are compatible.
Now we handle -o /dev/null equivalent to -fno-emit-bin because
otherwise our atomic rename into place will fail. This also
makes Zig do less work, avoiding pointless file system operations.
Do the fallible logic in Sema where we have access to error reporting
mechanisms, rather than in Type/Value.
We can't just do the best guess when resolving queries of "is this type
comptime only?" or "what is the ABI alignment of this field?". The
result needs to be accurate. So we need to keep the assertions that the
data is available active, and instead compute the necessary information
before such functions get called.
Unfortunately we are stuck with two versions of such functions because
the various backends need to be able to ask such queries of Types and
Values while assuming the result has already been computed and validated
by Sema.
This also includes two other small fixes:
- Instantiate void TypeInfo fields as void
- Return error in `type.comptimeOnly` on unresolved comptime requirements
This adds a comptime result when comparing a comptime value to an
unsigned integer. For example:
( 0 <= (unsigned runtime value)) => true
(-1 < (unsigned runtime value)) => true
((unsigned runtime value) < -15) => false
This bug only causes a failure on my machine when running
test/behavior/eval.zig directly. If running the full behavior test
suite, std.builtin.TypeInfo will have already resolved its layout,
causing the test to pass.
I'd love to add a test that can reliably reproduce this problem,
but I'm afraid I'm not sure how to reliably create a union with
un-resolved layout.
The ZIR instruction `union_init_ptr` is renamed to `union_init`.
I made it always use by-value semantics for now, not taking the time to
invest in result location semantics, in case we decide to change the
rules for unions. This way is much simpler.
There is a new AIR instruction: union_init. This is for a comptime known
tag, runtime-known field value.
vector_init is renamed to aggregate_init, which solves a TODO comment.
