[scummvm-devel] ScummVM 2.3.0 proposed schedule

[Torbjörn Andersson]

On 08/08/2021 15.16, Eugene Sandulenko via scummvm-devel wrote:

> I plan to start looking through the buglist today and prioritize the 
> blockers and nice-to-get-fixed bugs.

One thing I was hoping to see fixed, even though it's not really a high 
priority, is that Lemonhead glitch in non-English versions of The Secret 
of Monkey Island.

For those who missed the background, it seems that when LucasArts 
upgraded the game to the CD version, they managed to introduce some bugs 
as well, including a couple of missing lines for Lemonhead. I recently 
submitted a pull request to fix this in the English version, and it was 
accepted a few days later. See 
https://github.com/scummvm/scummvm/pull/3239 for details.

But this only fixed the glitch in the English CD version, and apparently 
the bug is also present in localized versions as well?

 From what I understand, the game was translated - both on floppy and CD 
- into French, German, Italian and Spanish. There's also the "ultimate 
talkie" version, but I don't even know if those extra lines were ever 
recorded. Anyway, that version could change at any time so it's probably 
safer to leave it alone.

If you don't have any of these localized versions you can stop reading here.

Anyone still here? Good. To illustrate, here's how the English script is 
patched:

If you descumm the room-25-205 script in the floppy version (boot param 
408 is useful here), you'll see this:

[00A0] (14) print(3,[Text("Oooh, that's nice." + wait() + "Simple.  Just 
like one of mine." + wait() + "And little.  Like mine.")]);
[00F0] (AE) WaitForMessage();

In the CD version, this is inexplicably shortened to:

[009E] (14) print(3,[Text("Oooh, that's nice.")]);

The new ScummEngine::tryPatchCannibalScript() tries to patch the script. 
With the English CD version, the resource it gets to work with is 82906 
bytes. The Common::hexdump() function is useful for visualizing it. You 
have to locate the script within the resource:

012080: 20 68 61 76  65 20 6e 6f  74 68 69 6e  67 20 66 6f   | have 
nothing fo|
012090: 72 20 75 73  2e 00 c0 2a  69 ff a0 4c  53 43 52 00   |r 
us...*i..LSCR.|
0120a0: 00 02 5f cd  40 01 01 00  ff 62 cc 1a  61 80 01 00 
|.._. at ....b..a...|

...

012140: ae 02 14 03  0f 4f 6f 6f  68 2c 20 74  68 61 74 27   |.....Oooh, 
that'|
012150: 73 20 6e 69  63 65 2e 00  14 03 0f 41  6e 64 20 69   |s 
nice.....And i|

...

0122e0: 01 01 5d 33  01 01 06 00  ff 62 d4 07  2d 01 00 07 
|..]3.....b..-...|
0122f0: 2e 01 00 1a  20 00 04 00  c0 a0 4c 53  43 52 00 00   |.... 
.....LSCR..|
012300: 00 7f ce 48  00 40 02 00  62 00 48 6b  00 04 00 14 
|...H. at ..b.Hk....|

The LSCR tag at offset 0x01209b (decimal 73883). The LSCR tag at offset 
0x0122fa (decimal 74490) is the start of the next script. So the script 
is 607 bytes.

The print instruction (0x14) we want to change is at offset 0x12142 
(decimal 74050) or 167 bytes into the script. (Descumm said 158 bytes, 
but I guess there is some header information that it doesn't count.) The 
next print instruction starts at 0x12158 (decimal 74072) so there is 22 
bytes to patch.

So that's where most of this information comes from:

		expectedSize = 82906;
		scriptOffset = 73883;
		scriptLength = 607;
		expectedMd5 = "98b1126a836ef5bfefff10b605b20555";
		patchOffset = 167;
		patchLength = 22;
		lang[0] = 'E';
		lang[1] = 'N';
		lang[2] = 'G';

Just to be safe, the script will only be patched in the size matches the 
expected size, if there is an LSCR tag at the beginning of the script, 
if there is a print instruction at the patch offset, and if the MD5 sum 
for the script matches.

There isn't enough space to fit the corrected text, so it gets replaced 
by a unique marker. For the English script this is "/LH.ENG/". This is 
padded with spaces to the desired size, before inserting 
WaitForMessage() before the next print. Given the above information, 
this should all happen automatically.

This is then replaced with the desired string in 
ScummEngine_v5::printPatchedMI1CannibalString():

	if (strncmp((const char *)ptr, "/LH.ENG/", 8) == 0) {
		msg =
"Oooh, that's nice.\xFF\x03"
"Simple.  Just like one of mine.\xFF\x03"
"And little.  Like mine.";
	}

The \xFF\x03 escape sequence means that it waits before printing the 
next part of the message.

So that should give you all the information you need, I hope.

Torbjörn Andersson