[Bug c/98905] New: lstat() confused with wrong struct stat

[gcc at madler dot net]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98905

            Bug ID: 98905
           Summary: lstat() confused with wrong struct stat
           Product: gcc
           Version: 5.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: gcc at madler dot net
  Target Milestone: ---

When compiling on Solaris (SunOS gcc-solaris11 5.11 11.3 sun4u sparc
SUNW,SPARC-Enterprise) using gcc 5.5.0 with a particular set of macros defined,
lstat() thinks its writing to a different struct stat than the struct stat that
is declared. This result in bogus values for the struct elements. Here is a
sample program:

#define _FILE_OFFSET_BITS 64
#define _XOPEN_SOURCE 600
#include <stdio.h>
#include <sys/stat.h>

int main(int argc, char **argv) {
    for (int i = 1; i < argc; i++) {
        struct stat st;
        int ret = lstat(argv[i], &st);
        printf("%s: lstat() returned %d, mode = %lx, size = %llu\n",
               argv[i], ret, st.st_mode, st.st_size);
    }
    return 0;
}

This will generally print a value like 1 or 2 for mode (should be 0x8zzz for a
file), and a giant number for size. Three things are required together to
manifest the bug: the two #define's and the use of lstat(). If either of the
defines is removed or stat() is used, the bug does not manifest.

Example output:

lstat-bug: lstat() returned 0, mode = 1, size = 6923981885842223364
lstat-bug.c: lstat() returned 0, mode = 1, size = 6923981159521138040