Iain Sandoe

Test results for the branch

166

I'll post test results onto this "issue" (it's not really appropriate to post them to gcc-testresults yet, since this is 'out of tree') [summ-aarch64-darwin20-branch-on-r11-4586.txt](https://github.com/iains/gcc-darwin-arm64/files/5501166/summ-aarch64-darwin20-branch-on-r11-4586.txt)

darwinpcs d.3 is not yet implemented.

3

from the README " D.3 From [2] : The general ABI specifies that it is the callee’s responsibility to sign or zero-extend arguments having fewer than 32 bits, and that...

Need a solution to cater for different stack layouts for normal c.f. variadic fns.

2

Although most backends work quite hard to determine the stack layout when deciding on register allocation etc, in practice that is ignored by the functions that actually layout values that...

Assembler/relocation that produces a malformed object file.

3

adrp x0, foo@PAGE-1 add x0, x0, foo@PAGEOFF-1 is a reasonably common thing for GCC to emit (corresponding to 'one position before' something). The assembler accepts the syntax but produces a...

Although this is not mentioned (that I've seen) in the available documentation. Darwin places the FP and LR at the top of the stack frame (immediately below the incoming SP...

Although the current implementation is sufficient to allow bootstrap, there are still quite a few mismatches between the CC implemented here and that implemented by clang. Since it's quite tricky...

For the case that the compiler does not support symmetrical transfer the task promise base final awaitable is missing a noexcept marker. Added here.

The current code assumes Linux and uses /proc/cpuinfo and/or a CPUINFO env var (but again parsing for Linux-specific content)

missing .tune and .arch entries.

this leads to the fail of mgeneral_regs_2.c