Re: msvcr90: Implement fread_s. (try 2)

[Piotr Caban]

On 12/23/11 08:44, Kusanagi Kouichi wrote:
> +  if (!buf)
> +  {
> +    _invalid_parameter(NULL, NULL, NULL, 0, 0);
> +    return 0;
> +  }
> You should set errno here.
> +  if (bufsize / size<  count)
> +  {
> +    fread(buf, bufsize, 1, fp);
> +    if (!feof(fp))
> +    {
> +      _invalid_parameter(NULL, NULL, NULL, 0, 0);
> +      memset(buf, 0, bufsize);
> +      fseek(fp, 1, SEEK_SET);
> +    }
> +    return 0;
> +  }
> This is still not working good. I don't know how it should be 
> implemented without doing much more testing. Probably printing a FIXME 
> message and returning an error will be acceptable.

If you want to fully implement it I guess it should do something like this:
  - execute _filbuf (this is probably causing the one character position 
change you're observing)
  - analyze internal FILE buffer and return error if needed
  - do the read
There are some problems if more data should be read then can be stored 
in internal FILE  buffer.

msvcr90: Implement fread_s. (try 2)

[Kusanagi Kouichi]

On 2011-12-23 12:41:56 +0100, Piotr Caban wrote:
> This is still not working good. I don't know how it should be
> implemented without doing much more testing. Probably printing a
> FIXME message and returning an error will be acceptable.
>
> If you want to fully implement it I guess it should do something like this:
>  - execute _filbuf (this is probably causing the one character
> position change you're observing)
>  - analyze internal FILE buffer and return error if needed
>  - do the read
> There are some problems if more data should be read then can be
> stored in internal FILE  buffer.

I don't want to fully implement fread_s. It is suffice for me to call
fread without paramater validation. I'd like to send reduced version
if it is acceptable.