[Webkit-unassigned] [Bug 128115] [Win] LLINT is not working.

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=128115





--- Comment #41 from peavo at outlook.com  2014-02-13 12:25:39 PST ---
It seems Win 32-bit has 4 byte stack alignment, as I've seen both 4, 8, and C as the last nibble in esp when entering the callToJavaScript function.
I guess this means we cannot use fixed offsets of the stack pointer to find the function arguments on the stack.
I've tried aligning the stack to 16 byte as suggested.
I also stored the value (esp & 0xf) in a temp register, and used that to compute the offset of the function arguments.

This is the prologue code I've tried:

    ...
    pushCalleeSaves
    if X86
        subp 12, sp
    elsif X86_WIN
        subp 16, sp
        move sp, t4
        move t4, t0
        move t4, t2
        andp 0xf, t2
        andp 0xfffffff0, t0
        move t0, sp
        storep t4, [sp]
    elsif ARM or ARMv7 or ARMv7_TRADITIONAL
    ...

Code to get function arguments from stack:

    if X86
        loadp 36[sp], vm
        loadp 32[sp], entry
    elsif X86_WIN
        loadp 40[sp,temp3], vm
        loadp 36[sp,temp3], entry
    else
        move cfr, previousCFR
    end

    ...

    if X86
        loadp 28[sp], previousPC
        loadp 24[sp], previousCFR
    elsif X86_WIN
        loadp 32[sp,temp3], previousPC
        loadp 28[sp,temp3], previousCFR
    end
    storep previousPC, ReturnPC[cfr]
    storep previousCFR, CallerFrame[cfr]

    if X86
        loadp 40[sp], protoCallFrame
    elsif X86_WIN
        loadp 44[sp,temp3], protoCallFrame
    end

This works fine in both debug and release, no more asserts, and I can see that the value of the parameters picked off the stack are correct.

Is this an OK way to do it?

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.