[Libcdio-devel] ack! another bug fixed in mmc_read_cd....

[Robert William Fuller]

Turns out mmc_read_cd() was ignoring the read_sector_type and treating 
everything as CDIO_MMC_READ_TYPE_ANY.  I do not know when this occurred 
but it seems like it happened at the introduction of the 
CDIO_MMC_SET_READ_TYPE macro.  Basically, the broken code looked like this:

  i_read_type = read_sector_type << 2;
  if (b_digital_audio_play) i_read_type |= 0x2;
  CDIO_MMC_SET_READ_TYPE(cdb.field, i_read_type);

where the macro looks like this:

#define CDIO_MMC_SET_READ_TYPE(cdb, sector_type) \
  cdb[1] = (sector_type << 2)

Because of the double shift, once in the macro, and once in 
mmc_read_cd(), the read_sector_type was being put into the reserved 
field of the cdb where it was ignored.  I fixed it with this code (and 
eliminated a local variable in the process:)

  CDIO_MMC_SET_READ_TYPE(cdb.field, read_sector_type);
  if (b_digital_audio_play) cdb.field[1] |= 0x2;

Of course the unfortunate side of this is that if somebody has broken 
code that depends on broken behavior, they'll now get an error instead 
of erroneous success.