My guess was accurate. A couple of small patches[0] allow me to cross-compile a working hello-world. I'll see if I can turn this into a full bindist to make arm64 ghc self hosting.
Then there a question of building ghc 9.8.3 instead of 9.10.1 which we currently have. I imagine we can figure this out without much trouble. I also want the hammer flag to be focused on a single file instead of everything. Eventually I want to get this into the ports (of course!) > I gather retguard is a mitigation against return-oriented programming > attacks, which is about manipulating return addresses so functions > return to malicious code instead of the code that called them, to > subvert protections like W^X and code signing. This is from a quick > search. https://www.openbsd.org/papers/eurobsdcon2018-rop.pdf is probably of interest to you then. > (For some reason, earlier in the thread, I was referring to W^X in the > context of different partitions, is that (or something similar) a thing > there too or was I misremembering or imagining things?) > > That said, if you can explain how to mark a function with > no-retguard-please, I'm happy to do that. Googling “openbsd > "no-retguard-please"” gives no results. That's because it's spelled -fno-ret-protector :-) > Also, I reckon we're probably nearing the time to actually create a > ticket on the GHC issue tracker. What do you reckon? I don't understand the issue well enough to file a sensible ticket. So far we know that the assembly fragment is somehow not compatible with retguard. I don't expect anybody in GHC land to become excited about debugging it instead of us doing the work. [0] https://gitlab.haskell.org/blackgnezdo/ghc/-/commits/ghc-9.10.1-openbsd
