the problem of test.c about memtest86

[MoreAndMore34545]

line 380 is here /* Original C code replaced with hand tuned assembly code /
/
for (; p <= pe; p++) {
*p = rand_32bits(me);
}
*/
line 386 is here asm volatile (
"jmp L200\n\t"
".p2align 4,,7\n\t"
"L201:\n\t"
"addl $4,%%edi\n\t"
"L200:\n\t"
"pushl %%ecx\n\t"
"call rand_32bits\n\t"
"popl %%ecx\n\t"
"movl %%eax,(%%edi)\n\t"
"cmpl %%ebx,%%edi\n\t"
"jb L201\n\t"
: : "D" (p), "b" (pe), "c" (me)
: "eax"
);

[anphsw]

Original code was written for very old GCC version (and not by me), and you compiling on the modern version of GCC, which may optimize code and can give incorrect results.
You can compile test.c with assembler output: in makefile add "-S" option:
$(CC) -S -c -Wall -march=i486 -m32 -O0 -fomit-frame-pointer -fno-builtin -ffreestanding -fno-pie -fno-stack-protector -std=gnu99 test.c

Compile will fail, but will produce test.s file with assembler source. You can compare then assembler results of original C code optimized by GCC and hand-tuned code.

Second, if you remove all assembly code from test.c, you can compile it without "-fno-pie -fno-stack-protector" options (as they added to support modern GCC versions) and see if it works.

[MoreAndMore34545]

thanks a lot.
I have fix the problem.
the replaced c code should be as below:
/* Original C code replaced with hand tuned assembly code */
*p = rand_32bits(me);
for (; p < pe; ) {
p++;
*p = rand_32bits(me);
}

[MoreAndMore34545]

the above c code could replace the assembly code
my GCC version is 5.4.0 on ubuntu 16.04

[anphsw]

Yes, it makes sense.
This code is unused, but if you will make commit fixing it, I will merge it. Or I can make it by myself, as you wish.