Prior to this commit, the AIR arg instruction kept a reference to a ZIR
string index for the corresponding parameter name. This is used by DWARF
emitting code. However, this is a design flaw because we want AIR
objects to be independent from ZIR.
This commit saves the parameter names into memory managed by
`Module.Fn`. This is sub-optimal because we should be able to get the
parameter names from the ZIR for a function without having them
redundantly stored along with `Fn` memory. However the current way that
ZIR param instructions are encoded does not support this case. They
appear in the same ZIR body as the function instruction, just before it.
Instead, they should be embedded within the function instruction, which
will allow this TODO to be solved. That improvement is too big for this
commit, however.
After this there is one last dependency to untangle, which is for inline
assembly. The issue for that is #10784.
We return compare flags rather than a register which than wrongly
cheats the regalloc into thinking we carry the instruction in the
register which we do not.
If the function call is not extern and we are compiling in debug,
pass function params always on stack. This will improve debugging
capabilities since the params will not be volatile and possibly
clobbered by the procedure code.
Finish implementation of `imul_complex`.
using the provided -L directories before checking if we need to integrate
with system library paths. This prevents false positive of invoking
system cc to find native paths when in fact all dependencies are
satisfied by -L (or --search-prefix to zig build).
This reverts commit baead472d7641bdd96130354bafadc1fb1ed223b.
Let's go through the proposal process on this one. I want to push back
on this. My position is that, at the very least, a full trace of command
lines of sub-processes should be printed on failure, with the exception
of opt-in flags such as `--prominent-compile-errors`.
- Correctly get discard symbol by first checking if it was discarded or not.
- Remove imports if extern symbols were resolved by an object file.
- Correctly relocate data symbols by ensuring the atom is from the correct file.
- Fix the `Names` section by using the resolved symbols, rather than the ones defined in Zig code.
We now correctly allocate and create atoms for symbols from other object files.
Imports are now also resolved and appended when required.
Besides those changes, we now duplicate all symbol names, so we can correctly
generate unique names for unnamed constants.
TODO: String interning
This implements the merging of all sections, to generate a valid wasm binary where all symbols
have been resolved and their respective sections have been merged into the final binary.
This upstreams the object file parsing from zwld, bringing us closer to being able
to link stage2 code with object files/C-code as well as replacing lld with the self-hosted
linker once feature complete.
Address https://github.com/ziglang/zig/issues/3477
This provides a mechanism for builds to fully report an error to the user and prevent zig from piling on extra noise.
Non-Arm CPUs use u16 as the parameter to __extendhfxf2 and the return value of
__truncxfhf2, so insert appropriate bitcasts in gen_soft_f80_widen_or_shorten.
Otherwise, LLVM might crash because the functions are called in a different way
than its compiler-rt definition.
This fixes stage1 build on SPARCv9, and possibly other non-x86, non-Arm CPUs.
For some errors if the found token is not on the same line as
the previous token, point to the end of the previous token.
This usually results in more helpful errors.
This updates the C backend to use proper array types.
In order to do that, this commit also:
- fixes up elem_ptr and field_ptr handling
- adds `renderTypecast` (renders in C typecast format, e.g. "int* [10]")
- adds a bit special handling for undefined pointers, which is necessary
to support slice/elem_ptr to undefined decls
Now, the abstracted stack offsets grow in the same direction as
the real stack values in hardware, and allocating stack memory is done
by the taking the last stack offset, adding required abi size
and aligning to the required abi align. Stack handling is now more
natural as it aligns itself with how it works in hardware; hence
stepping through the debugger and printing out different stack
values is intuitive. Finally, the stack pointers are now correctly
aligned to the required (and not necessarily natural) alignment.
* move DWARF in file if LINKEDIT spilled in dSYM
* update VM addresses for all segments
* introduce copyFileRangeOverlappedAll instead of File.copyRangeAll
since we make lots of overlapping writes in MachO linker
This "feature" of gcc/clang means to treat this as a positional
link object, but using the library search directories as a prefix.
We solve this problem in the CLI layer, using a separate map for
the data since it is an uncommon case.
Closes#10851
