Re: [PATCH v2] ftell: seek to end only when there are unflushed bytes (BZ #17647)

[Siddhesh Poyarekar <siddhesh at redhat dot com>]

On Fri, Dec 05, 2014 at 10:05:44AM -0200, Adhemerval Zanella wrote:
> The 'data_len' is calculated with wsclen and allocated as 'char'. The subsequent fgetws
> will then try to write at most 'data_len' wchar_t in a buffer with just data_len 'char'.
> The following patch fix the issues I am seeing on powerpc64:
> 
> diff --git a/libio/tst-ftell-active-handler.c b/libio/tst-ftell-active-handler.c
> index f69e169..44a4fac 100644
> --- a/libio/tst-ftell-active-handler.c
> +++ b/libio/tst-ftell-active-handler.c
> @@ -84,6 +84,7 @@ static const char *char_data = "abcdef";
>  static const wchar_t *wide_data = L"abcdef";
>  static size_t data_len;
>  static size_t file_len;
> +static size_t char_len;
>  
>  typedef int (*fputs_func_t) (const void *data, FILE *fp);
>  typedef void *(*fgets_func_t) (void *ws, int n, FILE *fp);
> @@ -364,11 +365,11 @@ do_ftell_test (const char *filename)
>              reading.  */
>           if (test_modes[i].fd_mode != O_WRONLY)
>             {
> -             char tmpbuf[data_len];
> +             char tmpbuf[data_len * char_len];
>  
>               rewind (fp);
>  
> -             while (fgets_func (tmpbuf, sizeof (tmpbuf), fp) && !feof (fp));
> +             while (fgets_func (tmpbuf, data_len, fp) && !feof (fp));
>  
>               write_ret = write (fd, data, data_len);
>               if (write_ret != data_len)
> @@ -656,6 +657,7 @@ do_test (void)
>    fgets_func = (fgets_func_t) fgets;
>    data = char_data;
>    data_len = strlen (char_data);
> +  char_len = sizeof (char);
>    ret |= do_one_test (filename);
>  
>    /* Truncate the file before repeating the tests in wide mode.  */
> @@ -678,6 +680,7 @@ do_test (void)
>    fgets_func = (fgets_func_t) fgetws;
>    data = wide_data;
>    data_len = wcslen (wide_data);
> +  char_len = sizeof (wchar_t);
>    ret |= do_one_test (filename);
>  
>    return ret;
> 
> 
> What do you think?
> 

Ah yes, thanks for debugging this.  Looks good to me.

Siddhesh

Attachment: pgpvO5DNOvPvI.pgp
Description: PGP signature