Re: [PATCH v2] gdb: replace alloca with gdb::unique_xmalloc_ptr in remote-fileio.c

[Pedro Alves <pedro@palves.net>]

On 2026-07-01 16:07, Luis Machado wrote:

>    char *c;
>    LONGEST retlong;
> @@ -211,6 +213,11 @@ remote_fileio_extract_ptr_w_len (char **buf, CORE_ADDR *ptrval, int *length)
>    *buf = c;
>    if (remote_fileio_extract_long (buf, &retlong))
>      return -1;
> +  /* Reject negative lengths and zero (unless the caller permits it for the
> +     Fsystem NULL-cmdline sentinel).  Oversized names are caught by the
> +     syscall.  */
> +  if (retlong < 0 || (!allow_zero_length && retlong == 0))
> +    return -1;

Given this no longer limits on PATH_MAX, RETLONG can be higher than INT_MAX
and still positive, and then the cast below yields bad values:

>    *length = (int) retlong;

Like, e.g., (int) 0x80000000, wraps to INT_MIN (a negative length).

So we should limit on INT_MAX.
> +  /* Valid: largest positive length accepted (technically INT_MAX).  */
> +  SELF_CHECK (parse ("1/7fffffff") == 0);
> +  SELF_CHECK (length == 0x7fffffff);

This test would have caught it, I think:

  SELF_CHECK (parse ("1/80000000") == -1);