With this routine, we are now able to freely shift stub_helper
section in memory and in file since the VM addressing is now dynamically
dependent on the positioning of `__stub_helper` preamble and other
sections generated by the linker.
By incrementally I mean using the incremental linker machinery
and concepts. Currently, lots of repetition but already highlighted
a potential problem with resolving relocations for symbols that
weren't seen yet but wanting to write the atom to file (before
seeing the relevant atoms).
with relocations to `dyld_private` and `__dyld_stub_binder` symbols
making the routine properly dynamic (i.e., making it possible to call
the routine before VM allocation takes place).
After merging `populateMetadata` with `populateMissingMetadata`,
it is imperative we clear the number of symbols in `writeSymbolTable`.
However, this is hopefully just a temp measure until the convergence
of incremental with traditional is complete.
* AstGen: use coerced_ty ResultLoc on array types and rely on Sema
doing type coercion, to reduce the size of the ZIR for these
instructions.
* Sema: implement `@ptrCast`.
* Sema: implement coercion from `T` to `?T` with an intermediate
coercion rather than equality check.
instead of globally storing unresolved and tentative defs,
store indices to actual symbols in the functions that are responsible
for symbol resolution.
This is now no longer limited to targeting macOS natively but also
tries to detect the sysroot when targeting different Apple platforms
from macOS; for instance targeting iPhone Simulator from macOS. In
this case, Zig will try detecting the SDK path by invoking
`xcrun --sdk iphonesimulator --show-sdk-path`, and if the command
fails because the SDK doesn't exist (case when having CLT installed only)
or not having either Xcode or CLT installed, we simply return null
signaling that the user has to provide the sysroot themselves.
For example, in order to run a binary on an iPhone Simulator,
you need to specify that explicitly as part of the target as
`aarch64-ios-simulator` rather than `aarch64-ios-gnu` or
`aarch64-ios` for short.
* ensure we correctly transfer `-iwithsysroot` and
`-iframeworkwithsysroot` flags with values from `build.zig` and that
they are correctly transferred forward to `zig cc`
* try to look for `libSystem.tbd` in the provided syslibroot - one
caveat that the user will have to specify library search paths too
Previously, I have incorrectly assumed that with two-level namespace
we only need to link in dylibs/frameworks that actually export symbols
which are undefined in the linked image. Turns out, regardless of
whether we link with two-level namespace (default on macOS) or a
flat namespace (more common on other platforms), we always need to
put the dylibs/frameworks as specified by the user from the linker
line into the final linked image.
* Value: rename `error_union` to `eu_payload` and clarify the intended
usage in the doc comments. The way error unions is represented with
Value is fixed to not have ambiguous values.
* Fix codegen for error union constants in all the backends.
* Implement the AIR instructions having to do with error unions in the
LLVM backend.
* New AIR instructions: ptr_add, ptr_sub, ptr_elem_val, ptr_ptr_elem_val
- See the doc comments for details.
* Sema: implement runtime pointer arithmetic.
* Sema: implement elem_val for many-pointers.
* Sema: support coercion from `*[N:s]T` to `[*]T`.
* Type: isIndexable handles many-pointers.
* Introduce `ret_load` ZIR instruction which does return semantics
based on a corresponding `ret_ptr` instruction. If the return type of
the function has storage for the return type, it simply returns.
However if the return type of the function is by-value, it loads the
return value from the `ret_ptr` allocation and returns that.
* AstGen: improve `finishThenElseBlock` to not emit break instructions
after a return instruction in the same block.
* Sema: `ret_ptr` instruction works correctly in comptime contexts.
Same with `alloc_mut`.
The test case with a recursive inline function having an implicitly
comptime return value now has a runtime return value because of the fact
that it calls a function in a non-comptime context.
This commit solves the problem in a much simpler way: putting
runtime-known values in place of non-comptime arguments when
instantiating a generic function.
