[Raion]

The reasons I decided not to look into these two (at least for now, maybe llvm as a whole becomes interesting down the line)

* "lld" - now you have to port the version of cmake, llvm build infra and related python version and tools, then actually port the linker itself which hasn't ever supported irix

Fair enough. Mach_kernel is working on LLVM in his spare time, so maybe hop in the discord and hit him up if you're curious.

* "gnu gold" - this linker also hasn't ever historically supported irix - sure it compiles, but that's quite different to it ever having actively supported or produced working binaries on irix. I'm willing to hear different. A warning bell for me - this is possibly being moved to separate support package from regular binutils due to lack of ownership / follow up by google (see https://www.phoronix.com/scan.php?page=n...agnate-F31)

So for me the "best bet" right now is to get regular gnu ld into a state where we can support building + using vaguely modern gdb and gcc, at which point building other modern software starts to be easier.

Don't get me wrong, diversity in our build utils are good, but the glaring issues with gnu ld make it a pretty bad choice compared to the default linker in IRIX. It's of course possible to fix those, but that means going through tens of thousands of lines of ld and bfd related code, with GNU never upstreaming those changes. So there's a lot that is good and bad here about them, just the unacceptable (in my view) issues with gnu ld means it's really only good for a gcc-only software distro such as the sgi dev pkgsrc stuff; that is, if they ever get around to packaging the software officially. It's this reason that I only am focusing on recipes for irisware that use the sgi linker/as; I think MIPSPro is worth trying for.