- for one-possible-value types, ir_analyze_struct_field_ptr()
no longer hardcodes const/volatile
- when slicing arrays, ir_analyze_instruction_slice()
no longer consults ConstValSpecialStatic
closes#5474
extracted function ir_try_evaluate_bin_op_const
extracted type_is_self_comparable function
renamed ir_try_evaluate_bin_op_const to ir_try_evaluate_bin_op_cmp_const
implemented analysis of ?T == T
added ir_set_cursor_at_end_and_append_basic_block_gen
use build_br_gen and ir_set_cursor_at_end_and_append_block_gen
added ir_append_basic_block_gen
removed include of all_types in ir.cpp
extracted compile-time and runtime evaluation of cmp_optional_non_optional to separate functions
closes#5390closes#1332
Consider a (legal according to the `@bitCast` rules) conversion from u16
to [2]u8: since the former is a scalar and the latter is a pointer
(arrays are represented at pointers in the codegen phase) we have to
allocate a temporary slot on the stack and then bitcast the resulting
pointer to the desired destination type.
Beware that this means the lifetime of the resulting value is the same
of the function it's contained in and for all intents and purposes
should be regarded as a local (eg. it should not escape).
Closes#4395Closes#5121
The codegen would sometimes change the LLVM type for some constants to
an unnamed structure in order to accomodate extra padding. This is fine
as long as the alignment of each field is still respected and it was not
the case for structure types, leading to ill-formed constants being
generated.
Optional types suffer from this to a lower extent as their layout is
quite lucky, the only missing piece was the tail padding.
Closes#4530Closes#4594Closes#4295
Closes my will to live
Before, type_has_one_possible_value would return false for the value
`.{1}`. But actually, that type is a tuple with a single comptime field.
Such a type, in fact, has one possible value.
This plus the corresponding adjustment to get_the_one_possible_value
solves #3878.
closes#4766
This commit also fixes the implementation of some utility functions for
adjusting properties of pointer types. Previously these functions would
incorrectly drop vector, sentinel, and inference metadata.
Constant-folding the pointers to the expected sentinel value have some
big problems: it hides the real content of the array, makes the pointer
to the sentinel point to a completely different memory region and treats
it like a const value even when the underlying array is mutable.
Fixes#4840
Jakub Konka
Michael Dusan
xackus
foobles
Veikka Tuominen
Timon Kruiper
Alexandros Naskos
Tadeo Kondrak
Andrew Kelley
LemonBoy
LemonBoy
elucent
Kelly Boothby