[Devel] PCF Accelerator parsing (was: pcfdriver.c memory leak)

[Detlef Würkner]

I wrote:

> The accel values are negative sometimes, more in the next mail.

When comparing the pcfread.c from XFree86 with the pcfread.c from
FreeType2, I find a difference in the parsing of the accelerator
tables.

XFree86 version:
----8<----
static Bool
pcfGetAccel(pFontInfo, file, tables, ntables, type)
[...]
    pcfGetMetric(file, format, &pFontInfo->minbounds);
    pcfGetMetric(file, format, &pFontInfo->maxbounds);
    if (PCF_FORMAT_MATCH(format, PCF_ACCEL_W_INKBOUNDS)) {
        pcfGetMetric(file, format, &pFontInfo->ink_minbounds);
        pcfGetMetric(file, format, &pFontInfo->ink_maxbounds);
[...]
static
pcfGetMetric(file, format, metric)
    FontFilePtr file;
    CARD32      format;
    xCharInfo  *metric;
{
    metric->leftSideBearing = pcfGetINT16(file, format);
    metric->rightSideBearing = pcfGetINT16(file, format);
    metric->characterWidth = pcfGetINT16(file, format);
    metric->ascent = pcfGetINT16(file, format);
    metric->descent = pcfGetINT16(file, format);
    metric->attributes = pcfGetINT16(file, format);
}

static
pcfGetCompressedMetric(file, format, metric)
    FontFilePtr file;
    CARD32      format;
    xCharInfo  *metric;
{
    metric->leftSideBearing = pcfGetINT8(file, format) - 0x80;
    metric->rightSideBearing = pcfGetINT8(file, format) - 0x80;
    metric->characterWidth = pcfGetINT8(file, format) - 0x80;
    metric->ascent = pcfGetINT8(file, format) - 0x80;
    metric->descent = pcfGetINT8(file, format) - 0x80;
    metric->attributes = 0;
}
----8<----

pcfGetAccel() never calls pcfGetCompressedMetric(). Now the FreeType2
version:

----8<----
  static FT_Error
  pcf_get_accel( FT_Stream  stream,
                 PCF_Face   face,
                 FT_ULong   type )
[...]
    error = pcf_get_metric( stream, format, &(accel->minbounds) );
    if ( error )
      goto Bail;
    error = pcf_get_metric( stream, format, &(accel->maxbounds) );
    if ( error )
      goto Bail;

    if ( PCF_FORMAT_MATCH( format, PCF_ACCEL_W_INKBOUNDS ) )
    {
      error = pcf_get_metric( stream, format, &(accel->ink_minbounds) );
      if ( error )
        goto Bail;
      error = pcf_get_metric( stream, format, &(accel->ink_maxbounds) );
[...]
  static FT_Error
  pcf_get_metric( FT_Stream   stream,
                  FT_ULong    format,
                  PCF_Metric  metric )
  {
    FT_Error error = PCF_Err_Ok;


    if ( PCF_FORMAT_MATCH( format, PCF_DEFAULT_FORMAT ) )
    {
      if ( PCF_BYTE_ORDER( format ) == MSBFirst )
        error = pcf_parse_metric( stream, pcf_metric_msb_header, metric );
      else
        error = pcf_parse_metric( stream, pcf_metric_header, metric );
    }
    else
      error = pcf_parse_compressed_metric( stream, metric );
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^
----8<----

Here pcf_get_accel() can call pcf_parse_compressed_metric(). Without
XFree86 and bdftopcf running here I cant definitely tell whether the
metrics in the accelerator tables are always uncompressed, but a quick
test with an additional boolean variable passed to pcf_get_metric()
and stating whether it was called from pcf_get_accel() or not and if
yes always use pcf_parse_metric() was ok.

Problem font example:

lib/X11/fonts/75dpi/lutBS14-ISO8859-1.pcf seems to have a negative
face->accel.maxbounds.characterWidth,

lib/X11/fonts/75dpi/lutBS24-ISO8859-1.pcf has a positive
face->accel.maxbounds.characterWidth.

Ciao, Detlef
-- 
_ // address@hidden
\X/  Detlef Wuerkner, Langgoens/Germany