[llvm-branch-commits] [NFCI][ELF][Mips] Refactor MipsGotSection to avoid explicit writes (PR #150730)

[Jessica Clarke via llvm-branch-commits]

jrtc27 wrote:

> > The current code is quite crusty and divergent from non-MIPS in API use, but fixing it up like this is quite high-risk, especially given how weird the MIPS GOT is when it comes to the required initial memory state. Is anyone using LLD for MIPS these days who can test this in a wider context? I've tried to be very careful but I would not be surprised if I've made mistakes; our test coverage isn't great.
> 
> It's indeed very difficult to find folks still concerned with MIPS... I believe only one company is still actively contributing to the MIPS backend in LLVM... @wzssyqa
> 
> Besides the ClangBuiltLinux maintainer and the OpenBSD maintainer continue to support MIPS.

This one I don't hugely care about getting in soon as it's (a) meant to be NFC (b) limited to MipsGotSection. It makes the code clearer and more like the normal GOT, but in theory what's there is fine, and the wider LLD internals aren't affected by it.

https://github.com/llvm/llvm-project/pull/150730