Re: Munge /r and /n in HTTP_HttpSendRequestA

[Rolf Kalbermatter]

Uwe Bonnes  wrote:
> This makes Xilinx webupdate.exe see the Servicepack
> - else
> -
>    /* remove \r and \n*/
> - {
> -
>    int nLen = strlen(lpwhr->lpszPath);
> -
>    while ((lpwhr->lpszPath[nLen-1] == '\r')||(lpwhr->lpszPath[nLen-1] == '\n'))
> -
>    {
> -
> nLen--;
> -
> lpwhr->lpszPath[nLen]='\0';
> -
>    }

Mmmh, what happens if the passed in path would only exist of \r\n? A for or while loop
without a guaranteed termination condition always seems suspect to me. I guess this
would crash with an access violation, and most probably overwrite a few bytes of memory
with 0 before that happens. Not good I think although I think it couldn't be effectively
exploited as buffer overrun!

while (nLen > 0 && ((lpwhr->lpszPath[nLen-1] == '\r') || (lpwhr->lpszPath[nLen-1] == '\n')))

might be the better solution. Though I still wonder about the [nLen-1] in above term.
It seems [nLen] instead would be actually the right thing to do here which would make it

while (nLen >= 0 && ((lpwhr->lpszPath[nLen] == '\r') || (lpwhr->lpszPath[nLen] == '\n')))

Rolf Kalbermatter