Multiple symbols can point to the same function, this means that when we loop over
the symbol list, we must deduplicate those functions being added twice.
Additionaly, we must also ensure that when we append a new type and set the type
index on a function, we must not do this again for the same function.
This commit also implements sorting of code atoms to ensure their order matches
the order of the function section to ensure the function signature matches
that of the function body.
Implements the creation of an undefined symbol for a compiler-rt intrinsic.
Also implements the building of the function call to said compiler-rt intrinsic.
The for-loop in dump() would index out of bounds if `t.index` is greater
than size, because `end` is the maximum of `t.index` and `size` rather than the
minimum.
This approach is more inline with what LLVM/LLD does for testing
of their output, and seems to be more generic and easier to extend
than implementing a lot of repetitive and nontrivial comparison
logic when working directly on structures.
CheckMachOStep specialises CheckFileStep into directed (surgical)
MachO file fuzzy searches. This will be the building block for
comprehensive MachO linker tests.
If page aligned requested pagezero size is 0, skip generating
__PAGEZERO segment.
Add misc improvements to the pipeline, and correctly transfer the
requested __PAGEZERO size to the linker.
Pass `-pagezero_size` to the MachO linker. This is the final
"unsupported linker arg" that I could chase that CGo uses. After this
and #11874 we may be able to fail on an "unsupported linker arg" instead
of emiting a warning.
Test case:
zig=/code/zig/build/zig
CGO_ENABLED=1 GOOS=darwin GOARCH=amd64 CC="$zig cc -target x86_64-macos" CXX="$zig c++ -target x86_64-macos" go build -a -ldflags "-s -w" cgo.go
I compiled a trivial CGo program and executed it on an amd64 Darwin
host.
To be honest, I am not entirely sure what this is doing. This feels
right after reading what this argument does in LLVM sources, but I am by
no means qualified to make MachO pull requests. Will take feedback.
wasm32-wasi-musl wants the standard symbol names however Linux requires
the `__gnu_*` flavors. I did not find any authoritative source on what
decides which symbol flavors to use. If we run into more trouble in the
future we can go back to having both.
After doing performance testing, it seems that multi-compilation-unit
compiler-rt did not bring the performance improvements that we expected
it to. The idea is that it makes linking faster, however, it incurred a
cost in the frontend that was not offset by any gains in linking.
Furthermore, the single-object compiler-rt (with -ffunction-sections and
--gc-sections) ends up being fewer bytes on disk and so it's actually
the same or faster linking speed than the multi-compilation-unit
version.
So we are planning to keep using single-compilation-unit compiler-rt for
the foreseeable future, but may experiment with this again in the
future, in which case this commit can be reverted.
Finishes cleanups that I started in other commits in this branch.
* Use common.linkage for all exports instead of redoing the logic in
each file.
* Remove pointless `@setRuntimeSafety` calls.
* Avoid redundantly exporting multiple versions of functions. For
example, if PPC wants `ceilf128` then don't also export `ceilq`;
similarly if ARM wants `__aeabi_ddiv` then don't also export
`__divdf3`.
* Use `inline` for helper functions instead of making inline calls at
callsites.
