[llvm-bugs] [Bug 35085] New: Clang seems to generate C code wrongly

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=35085

            Bug ID: 35085
           Summary: Clang seems to generate C code wrongly
           Product: clang
           Version: 3.8
          Hardware: PC
                OS: FreeBSD
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: LLVM Codegen
          Assignee: unassignedclangbugs at nondot.org
          Reporter: voidbrainvoid at gmail.com
                CC: llvm-bugs at lists.llvm.org

Created attachment 19344
  --> https://bugs.llvm.org/attachment.cgi?id=19344&action=edit
Artifacts to reproduce the bug and some infos from my environment

Hi there,

I am was facing some issues with Clang when trying to compile an ARC4
implementation that was working pretty well in GCC.

Inspecting my code a little more I discovered that maybe Clang is generating
wrong code for the following code construction:

#include <stdio.h>

struct wtf_ctx {
    unsigned char S[256];
    int i;
};

int main(int argc, char **argv) {
    struct wtf_ctx wc;

    for (wc.i = 0; wc.i < 256; wc.S[wc.i] = wc.i++) // <-- This exact one.
        ;

    for (wc.i = 0; wc.i < 256; wc.i++) {
        printf("S[%.2X] = %.2X\n", wc.i, wc.S[wc.i]);
    }

    return 0;
}

What I have noticed is that GCC code sets the array positions with the expected
values: [0] = 0, [1] = 1 .. [255] = 255.

Unlike, Clang code seems to use undefined values.

I abstracted all the ARC4 algorithm complications to isolate what could be a
"bug". The code above is able to reproduce those
differences between GCC and Clang.

The compilation command lines are simply: "gcc weird.c -oweird" and "clang
weird.c -oweird".

Originally I have ran this code in FreeBSD but I also have the same issue in a
Linux 32-bit SMP box (distro Slackware 14.00)
with the same Clang version and the "bug" is recurrent there too.

I am also attaching some information that maybe could be relevant and useful to
the development team.

Thanks in advance!

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20171025/f8874d49/attachment-0001.html>