CC @bogojevic @milanknezevic

cc @tklauser

Do I understand correctly that linux-mipsle-rtrk is linux/mipsle actually running on 64-bit MIPS hardware/kernel? Maybe that has something to do with it?

Correct.

On MIPS, the kernel returns whether the syscall has error in a separate register, R7 (https://go.googlesource.com/go/+/refs/heads/master/src/syscall/asm_linux_mipsx.s#107), instead of a negative result.

I'm not sure I understand why TestSyscallNoError expects syscall.RawSyscall returns an error while the syscall is not erred. It looks to me that the intent of TestSyscallNoError is that RawSyscallNoError does the right thing, not RawSyscall (incorrectly) returns an error.

It's not just the error value that's an unexpected value.

The first value is also unexpected. It's easier if it's reformatted like this:

expected 4294967295 / 2 / 4294967294,
     got 4294967294 / 0 / 4294967294

The first value, 4294967295, is just -1, as set by e.g. https://go.googlesource.com/go/+/refs/heads/master/src/syscall/asm_linux_386.s#79 when RawSyscall thinks it gets an error.

Assuming there is no error, 4294967294 is the expected result.

Change https://golang.org/cl/205898 mentions this issue: syscall: skip TestSyscallNoError on mips{,le}