This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Fix two more -Os strcoll / wcscoll build failures



On 26/02/2018 18:53, Joseph Myers wrote:
> Continuing fixes for -Os build issues shown with build-many-glibcs.py,
> this patch adds uses of DIAG_* to disable -Wmaybe-uninitialized in two
> more places where code inlined from strcoll / wcscoll is wrongly
> diagnosed as possibly using uninitialized structure fields.  (All
> these warnings in different places for these functions are I think
> essentially the same bug.)
> 
> Tested with build-many-glibcs.py for alpha-linux-gnu and
> mips-linux-gnu that this fixes the -Os build failures for those
> configurations with GCC 7.
> 
> 2018-02-26  Joseph Myers  <joseph@codesourcery.com>
> 
> 	* locale/weightwc.h (findidx): Ignore -Wmaybe-uninitialized for
> 	-Os in two more places.

Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>

> 
> diff --git a/locale/weightwc.h b/locale/weightwc.h
> index ac25ba9..36c65b5 100644
> --- a/locale/weightwc.h
> +++ b/locale/weightwc.h
> @@ -28,7 +28,15 @@ findidx (const int32_t *table,
>  	 const wint_t *extra,
>  	 const wint_t **cpp, size_t len)
>  {
> +  /* With GCC 7 when compiling with -Os the compiler warns that
> +     seq1.back_us and seq2.back_us, which become *cpp, might be used
> +     uninitialized.  This is impossible as this function cannot be
> +     called except in cases where those fields have been
> +     initialized.  */
> +  DIAG_PUSH_NEEDS_COMMENT;
> +  DIAG_IGNORE_Os_NEEDS_COMMENT (7, "-Wmaybe-uninitialized");
>    wint_t ch = *(*cpp)++;
> +  DIAG_POP_NEEDS_COMMENT;
>    int32_t i = __collidx_table_lookup ((const char *) table, ch);
>  
>    if (i >= 0)
> @@ -107,11 +115,18 @@ findidx (const int32_t *table,
>  	      continue;
>  	    }
>  
> +	  /* With GCC 7 when compiling with -Os the compiler warns
> +	     that seq1.back_us and seq2.back_us, which become usrc,
> +	     might be used uninitialized.  This is impossible for the
> +	     same reason as described above.  */
> +	  DIAG_PUSH_NEEDS_COMMENT;
> +	  DIAG_IGNORE_Os_NEEDS_COMMENT (7, "-Wmaybe-uninitialized");
>  	  if (cp[nhere - 1] > usrc[nhere -1])
>  	    {
>  	      cp += 2 * nhere;
>  	      continue;
>  	    }
> +	  DIAG_POP_NEEDS_COMMENT;
>  
>  	  if (cp[2 * nhere - 1] < usrc[nhere -1])
>  	    {
> 


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]