Re: strange gcc warning

[Steven Smith <sos22@cam.ac.uk>]

[-- Attachment #1: Type: text/plain, Size: 1091 bytes --]

> void
> my_print(int errno)
Defining a parameter to shadow a global variable is always a bad idea,
but it's an especially bad idea for errno. In multi-threaded programs,
and programs compiled with a certain set of flags, errno is actually a
preprocessor macro.  Usually, when using glibc, errno expands to
(*__errno_location()), and so your prototype expands to void
my_print(int (*__errno_location())).  gcc interprets this as a pointer
to a function returning an int pointer, which was probably not what
you expected.

> {
>     printf("ERRNO: %d\n", errno);
Now, this expands to ``printf("ERRNO: %d\n", (*__errno_location()));".
__errno_location is a function pointer returning an int *, and so
this calls the first argument, and then dereferences the returned
value.

> }
> 
> int
> main()
> {
>     int a = 10;
>     errno = 0;
>     my_print(a);
a is now implicitly converted to a pointer to function returning int
*.  When the code is run, my_print tries to call a function at address
10 in memory, and crashes.

> }

Steven Smith,
sos22@cam.ac.uk.

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]