[Scummvm-tracker] [ScummVM :: Bugs] #11528: ILLUSIONS: Duckman - Crash after option menu "times out"

[ScummVM :: Bugs]

#11528: ILLUSIONS: Duckman - Crash after option menu "times out"
--------------------+--------------------------------
Reporter:  raziel-  |       Owner:  (none)
    Type:  defect   |      Status:  new
Priority:  normal   |   Component:  Engine: Illusions
 Version:           |  Resolution:
Keywords:  Duckman  |        Game:
--------------------+--------------------------------
Comment (by dwatteau):

 @raziel-: Thanks for your new test!

 Indeed, there are various problems in my previous patches, but now I've
 think I've found the real culprit. And of course it was a one-liner :)

 Forget my Github fork, just go back to the official ScummVM sources, but
 apply this change in engines/illusions/resources/talkresource.cpp:

 {{{
  #if defined(SCUMM_BIG_ENDIAN)
 -       for (byte *ptr = (byte *)_text; ptr != _tblPtr; ptr += 2) {
 +       for (byte *ptr = (byte *)_text; *ptr != 0; ptr += 2) {
                 WRITE_UINT16(ptr, SWAP_BYTES_16(READ_UINT16(ptr)));
         }
  #endif
 }}}

 and I think that should be it.

 The reason is that, on big-endian systems, the original line above
 reversed the internal wide strings until reaching _tblPtr, but
 TalkEntry::load() is called for every piece of text in a talk entry, and
 _tblPtr is way past all the text parts:

 {{{
 TalkEntry::load() _talkId: 000F00C8; textOffs: 0000065C; tblOffs:
 0000285A; voiceNameOffs: 00003C04
 TalkEntry::load() _talkId: 000F00C9; textOffs: 000006E8; tblOffs:
 00002896; voiceNameOffs: 00003C0C
 TalkEntry::load() _talkId: 000F00CA; textOffs: 000007A8; tblOffs:
 000028E8; voiceNameOffs: 00003C14
 TalkEntry::load() _talkId: 000F00CB; textOffs: 0000080C; tblOffs:
 0000290E; voiceNameOffs: 00003C1C
 }}}

 so, every piece of text was reversed multiple times, sometimes being in
 native endianness, and sometimes not, hence the crashes (that's why c
 appeared as 22272 on big-endian PowerPC instead of 87 on Intel: 0x0057 vs.
 0x5700). Well, at least, that's what I understood from my debugging
 session :)

 I'll continue my testing a bit, but I think that's the fix I'm going to
 submit.
-- 
Ticket URL: <https://bugs.scummvm.org/ticket/11528#comment:20>
ScummVM :: Bugs <https://bugs.scummvm.org>
ScummVM