Re: [PATCH 1/3]: C++20 P0482R6 and C2X N2653: Fix for bug 25744, mbrtowc with Big5-HKSCS

[Adhemerval Zanella <adhemerval.zanella@linaro.org>]

On 07/01/2022 21:39, Tom Honermann via Libc-alpha wrote:
> Subject:
> [PATCH 1/3]: C++20 P0482R6 and C2X N2653: Fix for bug 25744, mbrtowc with Big5-HKSCS
> From:
> Tom Honermann via Libc-alpha <libc-alpha@sourceware.org>
> Date:
> 07/01/2022 21:39
> 
> To:
> libc-alpha <libc-alpha@sourceware.org>
> 
> 
> This patch for bug 25744 [1] updates the Big5-HKSCS converter to properly maintain the lowest 3 bits of the mbstate_t __count data member.  This change is necessary to ensure that state is correctly preserved when the converter encounters an incomplete multibyte character.  More details are available in bug 25744 [1].
> 
> The code changes are styled to match how these bits are maintained by converters such as iso-2022-jp.c, ibm930.c, and others.
> 
> Running 'grep __count' in the 'iconvdata' directory suggests that a number of other converters, euc-jisx0213.c for example, also fail to preserve these bits in some cases, though it may be that negative effects are not observed for those converters.  This patch does not attempt to address such issues with other converters.
> 
> This fix was previously posted to this mailing list on April 7th, 2020 [2], and again on June 7th, 2021 [3], but was not followed up on.
> 
> Tested on Linux x86_64.
> 
> Tom.
> 
> [1]: Bug 25744
>      "mbrtowc with Big5-HKSCS returns 2 instead of 1 when consuming the
>      second byte of certain double byte characters"
>      https://sourceware.org/bugzilla/show_bug.cgi?id=25744
> 
> [2]: "[PATCH] Correct the Big5-HKSCS converter to preserve low order
>      state bits (bug 25744)"
>      https://sourceware.org/pipermail/libc-alpha/2020-April/112595.html
> 
> [3]: "[PATCH 1/3]: C++20 P0482R6 and C2X N2653: Fix for bug 25744,
>      mbrtowc with Big5-HKSCS"
>      https://sourceware.org/pipermail/libc-alpha/2021-June/127231.html
> 
> n2653-1.patch

Next time please send the patch inline instead of as attachment (as git-send
does).  It is usually easier to reply with most email clients.

> 
> commit 900d307f4bfc5016e1a18711e8909dbce3488d3e
> Author: Tom Honermann <tom@honermann.net>
> Date:   Wed Jan 5 18:02:24 2022 -0500
> 
>     Correct the Big5-HKSCS converter to preserve low order state bits.
>     
>     BZ: https://sourceware.org/bugzilla/show_bug.cgi?id=25744

The fix make sense, although I am not familiar with the locate.  Some comments
on testcase.

> 
> diff --git a/iconvdata/big5hkscs.c b/iconvdata/big5hkscs.c
> index ed51412bb9..2455c9db53 100644
> --- a/iconvdata/big5hkscs.c
> +++ b/iconvdata/big5hkscs.c
> @@ -17769,7 +17769,7 @@ static struct
>     the output state to the initial state.  This has to be done during the
>     flushing.  */
>  #define EMIT_SHIFT_TO_INIT \
> -  if (data->__statep->__count != 0)					      \
> +  if ((data->__statep->__count >> 3) != 0)				      \
>      {									      \
>        if (FROM_DIRECTION)						      \
>  	{								      \
> @@ -17778,7 +17778,7 @@ static struct
>  	      /* Write out the last character.  */			      \
>  	      *((uint32_t *) outbuf) = data->__statep->__count >> 3;	      \
>  	      outbuf += sizeof (uint32_t);				      \
> -	      data->__statep->__count = 0;				      \
> +	      data->__statep->__count &= 7;				      \
>  	    }								      \
>  	  else								      \
>  	    /* We don't have enough room in the output buffer.  */	      \
> @@ -17792,7 +17792,7 @@ static struct
>  	      uint32_t lasttwo = data->__statep->__count >> 3;		      \
>  	      *outbuf++ = (lasttwo >> 8) & 0xff;			      \
>  	      *outbuf++ = lasttwo & 0xff;				      \
> -	      data->__statep->__count = 0;				      \
> +	      data->__statep->__count &= 7;				      \
>  	    }								      \
>  	  else								      \
>  	    /* We don't have enough room in the output buffer.  */	      \
> @@ -17878,7 +17878,7 @@ static struct
>  									      \
>  		/* Otherwise store only the first character now, and	      \
>  		   put the second one into the queue.  */		      \
> -		*statep = ch2 << 3;					      \
> +		*statep = (ch2 << 3) | (*statep & 7);			      \
>  		/* Tell the caller why we terminate the loop.  */	      \
>  		result = __GCONV_FULL_OUTPUT;				      \
>  		break;							      \
> @@ -17895,7 +17895,7 @@ static struct
>        }									      \
>      else								      \
>        /* Clear the queue and proceed to output the saved character.  */	      \
> -      *statep = 0;							      \
> +      *statep &= 7;							      \
>  									      \
>      put32 (outptr, ch);							      \
>      outptr += 4;							      \
> @@ -17946,7 +17946,7 @@ static struct
>  	  }								      \
>  	*outptr++ = (ch >> 8) & 0xff;					      \
>  	*outptr++ = ch & 0xff;						      \
> -	*statep = 0;							      \
> +	*statep &= 7;							      \
>  	inptr += 4;							      \
>  	continue;							      \
>  									      \
> @@ -17959,7 +17959,7 @@ static struct
>  	  }								      \
>  	*outptr++ = (lasttwo >> 8) & 0xff;				      \
>  	*outptr++ = lasttwo & 0xff;					      \
> -	*statep = 0;							      \
> +	*statep &= 7;							      \
>  	continue;							      \
>        }									      \
>  									      \
> @@ -17996,7 +17996,7 @@ static struct
>  	   /* Check for possible combining character.  */		      \
>  	    if (__glibc_unlikely (ch == 0xca || ch == 0xea))		      \
>  	      {								      \
> -		*statep = ((cp[0] << 8) | cp[1]) << 3;			      \
> +		*statep = (((cp[0] << 8) | cp[1]) << 3) | (*statep & 7);      \
>  		inptr += 4;						      \
>  		continue;						      \
>  	      }								      \
> diff --git a/iconvdata/tst-iconv-big5-hkscs-to-2ucs4.c b/iconvdata/tst-iconv-big5-hkscs-to-2ucs4.c
> index 91d7d3a552..12697ebe23 100644
> --- a/iconvdata/tst-iconv-big5-hkscs-to-2ucs4.c
> +++ b/iconvdata/tst-iconv-big5-hkscs-to-2ucs4.c
> @@ -128,6 +128,71 @@ check_conversion (struct testdata test)
>        printf ("error: Result of third conversion was wrong.\n");
>        err++;
>      }
> +
> +  /* Now perform the same test as above consuming one byte at a time.  */
> +  mbs = test.input;
> +  memset (&st, 0, sizeof (st));
> +
> +  /* Consume the first byte; expect an incomplete multibyte character.  */
> +  ret = mbrtowc (&wc, mbs, 1, &st);
> +  if (ret != -2)
> +    {
> +      printf ("error: First byte conversion returned %zd.\n", ret);
> +      err++;
> +    }

Although rest of the file does not use the libsupport checks, I think it is
expected the newer code does.  So change to:

  TEST_COMPARE (ret, -2);

Some for the rest of tests (there is no need to update err since TEST_COMPARE
will take care of setting the tests a failure for its parent).  If you want
you may also adapt all the previous tests to use TEST_COMPARE as well.

> +  /* Advance past the first consumed byte.  */
> +  mbs += 1;
> +  /* Consume the second byte; expect the first wchar_t.  */
> +  ret = mbrtowc (&wc, mbs, 1, &st);
> +  if (ret != 1)
> +    {
> +      printf ("error: Second byte conversion returned %zd.\n", ret);
> +      err++;
> +    }
> +  /* Advance past the second consumed byte.  */
> +  mbs += 1;
> +  if (wc != test.expected[0])
> +    {
> +      printf ("error: Result of first wchar_t conversion was wrong.\n");
> +      err++;
> +    }
> +  /* Consume no bytes; expect the second wchar_t.  */
> +  ret = mbrtowc (&wc, mbs, 1, &st);
> +  if (ret != 0)
> +    {
> +      printf ("error: First attempt of third byte conversion returned %zd.\n", ret);
> +      err++;
> +    }
> +  /* Do not advance past the third byte.  */
> +  mbs += 0;
> +  if (wc != test.expected[1])
> +    {
> +      printf ("error: Result of second wchar_t conversion was wrong.\n");
> +      err++;
> +    }
> +  /* After the second wchar_t conversion, the converter should be in
> +     the initial state since the two input BIG5-HKSCS bytes have been
> +     consumed and the two wchar_t's have been output.  */
> +  if (mbsinit (&st) == 0)
> +    {
> +      printf ("error: Converter not in initial state.\n");
> +      err++;
> +    }
> +  /* Consume the third byte; expect the third wchar_t.  */
> +  ret = mbrtowc (&wc, mbs, 1, &st);
> +  if (ret != 1)
> +    {
> +      printf ("error: Third byte conversion returned %zd.\n", ret);
> +      err++;
> +    }
> +  /* Advance past the third consumed byte.  */
> +  mbs += 1;
> +  if (wc != test.expected[2])
> +    {
> +      printf ("error: Result of third wchar_t conversion was wrong.\n");
> +      err++;
> +    }
> +
>    /* Return 0 if we saw no errors.  */
>    return err;
>  }
> 
> Attachments:
> 
> n2653-1.patch	5,5 KB
>