SYSARG_1 and SYSARG_RESULT are the same on ARM and AArch64

[yuyichao]

And I assume this is a problem on sh4 as well for SYSARG_5 and SYSARG_RESULT.

This cause the syscall exit handler to get the wrong arg1 value, in particular, for PR_getsockname. This causes the socket03 test to fail.

A similar issue exists on x86 between SYSARG_NUM and SYSARG_RESULT but the user_regs_struct have a orig_eax/orig_rax there that stored the old value.

[oxr463]

@yuyichao is this fixed with the latest v5.3.0 release?

[yuyichao]

The linked PR worked around this for the specific case. I didn’t see any built in use that can cause problem but I also don’t think there’s a generic solution atm.

[oxr463]

Hmm, okay. I'll leave this open for now then.