apple syscall touches registers for no reason

[tmcarthur]

In syscall.c, _start, or somewhere else? It uses same code as Linux unless I'm mistaken, right?

[lunixbochs]

OS X uses the stack for syscalls iirc, only uses eax/rax for the syscall number.

[tmcarthur]

I think that's true for i386, though I can't test since I don't have an i386 box right now - I'll try to grab one - in the case of x86_64 though I'm fairly sure it uses the same ABI for system calls as Linux (System V w/ AMD64 support). Maybe I'm mistaken though? I know general function calls pass parameters either in GP registers and/or stack though - and reserves stack space if you use GP registers (I think?).

[tmcarthur]

More info: https://developer.apple.com/library/mac/documentation/DeveloperTools/Conceptual/LowLevelABI/140-x86-64_Function_Calling_Conventions/x86_64.html#//apple_ref/doc/uid/TP40005035-SW1

[lunixbochs]

I'm in the process of improving processor configuration at build time (and adding preliminary mips support) so even if it only applies to i386, it still matters here as I'm actually able to build/run 32-bit lib43 on OS X now.

[tmcarthur]

Okay cool - was just curious :)

[lunixbochs]

i386 syscall prologue can probably be changed to pop eax?

[tmcarthur]

Agreed. Will also need to make sure OS X sys calls using i386 ABI don't add the 2 << 30 offset we add to 64-bit - it's only used for 64-bit sys calls

[lunixbochs]

grr, i386 is broken on OS X. Try the latest master with -DCPU=i386

[tmcarthur]

Fun! Do you want to fix it, or do you want me to submit a pull req to?

[lunixbochs]

I'm not sure what exactly is breaking (I already fixed the _start), it does make it to syscall() before it hangs.

I'm in the middle of moving syscalls from C to assembly (now that the arch stuff is more fleshed out) so you shouldn't pull req just yet if it comes to that.

[tmcarthur]

Sounds good - I'll have a look and then need to run some errands - but I'll check out where things are this evening and can submit pull req to fix it then unless you already have it taken care of if that works?

Oh also let me know if you have any issues w/ arm pull req - I can fix merge conflicts once you update stuff on that one if you'd prefer as well - thanks.

[lunixbochs]

After my updates today (especially with mips support), the arm pull request will hopefully just need to be syscall and crt0 _start asm blobs (besides the div stuff).

You should take a look at my updates to CMakeLists, as I did change it a bit for sanity.

[tmcarthur]

Yeah - I figured as much - that's a good thing! The div stuff I'm going to do in separate commit in day or two because I want to optimize it so it isn't slow/crappy and it handles all the corner cases (meaning all the variants that ARM expects, idiv, uidiv, etc), for now just planning to send pull req for armv7+

Yeah I saw that - it's a lot cleaner now and it's going to be easier to support multiple architectures which will be nice.

Cool - I'll hold off on making any more pull reqs for now then and just wait for you to finish your refactoring. Could you just let me know over this or twitter once you're done so I know when I'm good to go? Thanks!

[lunixbochs]

Fixed by e07c45de562cd550c1cbd0b6820c66bbebee19ec