[issue5677] Serious interpreter crash and/or arbitrary memory leak using .read() on writable file

[Robert Xiao]

Robert Xiao <nneonneo at gmail.com> added the comment:

It seems like this is actually a problem in Windows libc or something (tested using MinGW on Windows XP):

#include <stdio.h>

main() {
    FILE *f = fopen("test", "wb");
    fwrite("test", 1, 4, f);
    char buf[2048];
    size_t k = fread(buf, 1, 2048, f);
    printf("%d\n", k);
    int i=0;
    for(; i<k; i++) printf("%02x", buf[i]);
}

This causes a lot of garbage to be printed out. Removing the fwrite causes "0" to be printed with no further output.

The garbage is not from the uninitialized buffer, since I've verified that the original contents of the buffer are not what is being printed out. Furthermore, adjusting 2048 produces a maximum output of 4092 bytes (even with 9999 in place of 2048).

Short of simply denying read() on writable files, I don't really see an easy way around this libc bug.

----------

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue5677>
_______________________________________