Undefine SYS_access so we only need __syscall_faccessat

[kripken]

This way we just need to define one syscall instead of two. (Musl will
call the proper syscall based on what it sees is defined.)

[sbc100]

Nice!

[kleisauke]

I think this PR broke access(2) with absolute Windows paths. See for example this test case on Windows:

#include <unistd.h>
#include <stdio.h>

int main() {
  printf("%d\n", access("C:/existingfile", F_OK));
  
  return 0;
}

# Make a new file named existingfile in C:/
$ type nul > C:/existingfile

# Ensure C:/existingfile does really exist
$ IF EXIST "C:/existingfile" (echo 1) ELSE (echo 0)
1

# faccessat(2) should ignore dirfd when pathname is absolute
# (i.e. it should check the presence of C:/existingfile and not ./C:/existingfile)
$ emcc test.c -o test.js -sNODERAWFS -sEXIT_RUNTIME

# This should print 0, indicating that C:/existingfile does exists
$ node test.js
-1

$ rem C:/existingfile

# This should always print -1, since we deleted C:/existingfile above
$ node test.js
-1

Absolute POSIX paths are already handled here:

emscripten/src/library_syscall.js

Lines 24 to 26 in acbf9e6

	if (path[0] === '/') {
	return path;
	}

So, it needs something similar to Node's Windows variant of path.isAbsolute:
https://nodejs.org/api/path.html#pathisabsolutepath

Perhaps we can just call that utility instead when ENVIRONMENT_IS_NODE? Since this is only a issue with -sNODERAWFS (AFAIK).

[sbc100]

Interesting.. I think maybe we need to create a PATH.isAbs and then have NODERAWFS override various part of the PATH object.

[kripken]

PATH.isAbs() sgtm, however I'm not sure why using faccessat() caused this to start failing on windows, so I'm missing the source of the problem here, and am not sure why PATH.isAbs() would fix it...

[sbc100]

PATH.isAbs() sgtm, however I'm not sure why using faccessat() caused this to start failing on windows, so I'm missing the source of the problem here, and am not sure why PATH.isAbs() would fix it...

Because then NODERAWRS under windows could override this method such that c:/foo is also reported as absolute.. not just /foo.

[kripken]

But why was such an overriding necessary after this PR (which does not modify JS code in a significant way AFAICT)?

[sbc100]

This change forces more code to through fassessat as opposed to the old faccess.. these means more code hits calculateAt than before which mean that a broken isAboslute check the top of this function effects more code.

[kripken]

Got it, thanks @sbc100 !