#2466 closed defect (fixed)
FT: Invalid Actor when
| Reported by: | SF/jorpho | Owned by: | eriktorbjorn |
|---|---|---|---|
| Priority: | normal | Component: | Engine: SCUMM |
| Version: | Keywords: | script | |
| Cc: | Game: | Full Throttle |
Description
I have observed this error in both ScummVM 0.8.0 and the daily CVS build (Win32 versions in both cases) from January 15, 2006.
Go to the Junkyard with the gas can, lock pick, and hose in your inventory. After trapping the dog in a suspended car, go to the area where the dog is trapped. The attached save game already has everything in position.
Right-click to bring up the inventory, click on the hose, and then click on the gas can. ScummVM will crash with Error (28:103:0x12E): Invalid Actor 108 in o6_isActorInBox.
I am not sure if this bug can only be triggered in this location, but this is where I have observed it.
I am using an English version of Full Throttle (specifically, the Creative Labs-issued CD with the 3- level Special Edition of Dark Forces).
Ticket imported from: #1407789. Ticket imported from: bugs/2466.
Attachments (5)
Change History (42)
by , 18 years ago
comment:1 by , 18 years ago
| Summary: | Invalid Actor when "Combining" Two Items → FT Invalid Actor when "Combining" Two Items |
|---|
comment:2 by , 18 years ago
A few random observations:
* It seems to happen regardless of which inventory objects I use on each other.
* It does not happen if I first leave the room.
* It appears to be the "sentence script" (in this case, script 103) that's triggering the error message. The parameters to the script are 'verb', 'objectA' and 'objectB'. In this case, 'objectB' is 108, so that's probably where it gets the invalid actor id from.
It could be a script bug that happened to work with the original interpreter, but that ScummVM considers to be a fatal error. But I'm far from sure.
If it is a script bug, I guess the only thing we can do is to add a special case to o6_isActorInBox(). That's easy, but perhaps there's a more correct bugfix...?
comment:4 by , 18 years ago
| Summary: | FT Invalid Actor when "Combining" Two Items → FT: Invalid Actor when "Combining" Two Items |
|---|
comment:5 by , 18 years ago
Wouldn't be the first script bug of that kind. I guess we'll have to disassemble the relevant scripts and stare at the them for some time to figure out what to do :-)
comment:6 by , 18 years ago
Well, I believe I have the correct script descummed already, so I've attached it. But to be honest, I'm a bit uncertain about what some of it means. It does this:
[00C3] (5D) if (getOwner(localvar1) == 1) { [00CE] (43) localvar3 = localvar2 [00D4] (73) } else { [00D7] (43) localvar3 = localvar1 [00DD] (**) } [00DD] (0C) dup[1] = localvar3
Does that dup[1] line at the end mean that localvar3 is being pushed onto the stack? Anyway, then it does
[00E1] (5D) if (dup[1] == 195) { [00E8] (D5) jumpToScript(0,28,[localvar0,localvar1,localvar2]) [00FC] (5D) } else if (dup[1] == 0) { [0107] (D5) jumpToScript(0,28,[localvar0,localvar1,localvar2]) [011B] (5D) } else if (!isActorInBox(21)) {
I know o6_isActorInBox() pops two parameters: The box number, and the actor number. I take it 21 is the box number, and according to the error message, 108 is the actor number. That would correspond to localvar2, I believe, which could be what localvar3 contains.
Either way, except for some special cases it's probably assuming that either 'objectA' or 'objectB' is an actor. At least, that's how I understand it.
comment:7 by , 18 years ago
This is definitely a script bug.
I just ran the original interpreter through a debugger and played this scene. Like in scummvm, the sentence script (104 in my french version) is run with object numbers/ids as parameters. Then o6_isActorInBox is called and silently read data of out of the actor tables...
comment:8 by , 18 years ago
Hmm... Adding a workaround in just o6_isActorInBox() isn't enough because other actor opcodes will be called with the same bad actor. Setting one of the parameters to a valid actor causes odd things to happen...
It looks like the usual sentence script is script 28. Perhaps we could call that instead, it's about to call script 103 with bad parameters. I've attached a possible patch.
comment:9 by , 18 years ago
Well, since it's now confirmed as a script bug (thanks cyx), we shouldn't hesitate to add a "WORKAROUND bug #XY" :-). Erik, you could try to implement your trick and see if it does something sensible?
Alternatively, instead of calling the usual sentence script, we could simply do nothing, no?
comment:10 by , 18 years ago
The advantage of calling the usual sentence script is that then Ben will say "A hose won't help here." just as in any other room. If you think it's a safe workaround, I'd be happy to apply it.
Not calling any script at all will make nothing happen.
comment:12 by , 18 years ago
| Owner: | set to |
|---|
comment:13 by , 18 years ago
| Resolution: | → fixed |
|---|---|
| Status: | new → closed |
comment:14 by , 18 years ago
I've applied the fix to the trunk, but not the 0.8 branch. Perhaps I should, but it is a pretty obscure bug...
comment:15 by , 18 years ago
Is there a way to differentiate FT versions (ie. a versus b) ? It seems your script numbers are all off-by one compared to the ones in my french version.
room-28-2031
[034D] (5E) startScript(0,2024,[]) [0357] (43) VAR_SENTENCE_SCRIPT = 104 [035D] (5E) startScript(1,2030,[])
exit-91
[003D] (9D) actorOps.setShadowMode(1) [0042] (43) VAR_SENTENCE_SCRIPT = 29 [0048] (7C) stopScript(2005)
Adding an extra check is always possible, but that doesn't look like really a 'nice' solution to me....
comment:16 by , 18 years ago
| Status: | closed → new |
|---|
comment:18 by , 18 years ago
| Resolution: | fixed |
|---|
comment:19 by , 18 years ago
I've made some changes in an attempt to deal with the different Full Throttle versions. There's some guesswork involved, so I don't know if it's working correctly.
comment:23 by , 18 years ago
I slightly modified your code in SVN: It now doesn't check the language, but rather checks which value VAR_SENTENCE_SCRIPT is set to the very first time it is set (by script 1).
Plus I confirmed that the scripts are correct for the German Mac version, too.
comment:24 by , 18 years ago
Is that really a safe thing to do? What happens if you use the -x command-line option to load a saved game, thus (I assume) bypassing script 1?
Mind you, I haven't been able to find a case where it breaks yet...
comment:25 by , 18 years ago
Arr, err... gee, saving and loading is for wuzzies only, I'll just disable it and we'll be fine!
Gee, hm, embarassing but I totally forgot about that. So we either need to save those script IDs, or revert, or come up with another method :-/
comment:26 by , 18 years ago
OK here's another possibility: Use the value of res.num[rtScript]. For me it is 456 in the first case, and 469 for my German/Mac FT version.
Can you guys check which values you are seeing there?
comment:27 by , 18 years ago
My English PC version has res.num[rtScript] = 456, but perhaps that's what you meant with the first case.
If this detection method works, we could remove the variables keeping track of the script numbers. That'd be nice.
comment:28 by , 18 years ago
Yes, that's what I meant. I have three versions of FT/English (V1 PC, V2 PC, and Mac), all have 456. And my German Mac FT has 469.
Now, if cyx says that his French FT also has 469, I say we can simply modify the code accordingly. And maybe add a warning() in the startup code if a different number is found (or maybe initially let's use error() so that people notice it and report it *g*).
comment:29 by , 18 years ago
... allocResTypeData(Script/script,PRCS,467,1) ... I get the same debug output for both of my FT Mac/PC french versions.
Just to satisfy my curiosity, can someone with an english version of FT attach here a dump of script-1 ? Thanks.
comment:30 by , 18 years ago
OK, so the count doesn't match, but we still might be able to do something like "if script count > 456" or "if script count > 460" or so. It definitly would be good to find out the script count for more versions of FT, we could ask people in the forums and on IRC for some more testing.
I will attach a dump of script-1 once I am back home (unless somebody is faster :). I already determined which script they inserted (IIRC it was script 12 or 13 or so, need to look it up at home), if that's what you are interested in.
comment:32 by , 18 years ago
Thanks for uploading it ; and yes, that was what I was interested in, looking at which scripts they may have added.
comment:33 by , 18 years ago
It's script 12 that was inserted (for me one of the first things it does is to call "printDebug.msg("start dialog scan")"). Compare this with script 13 (resp. the script 12 in the "old" numbering, i.e. in the english version).
comment:34 by , 18 years ago
| Resolution: | → fixed |
|---|
comment:35 by , 18 years ago
I commited a new fix based on the script count. It would be nice to know whether this works correctly for other FT versions, of course.
BTW, it's really simple to find out the second script: One just has to run scummvm -b28 -u ft to get a dump of the relevant scripts from room 28. In my case that it is room-28-2031 which modifies VAR_SENTENCE_SCRIPT.
comment:37 by , 18 years ago
| Status: | new → closed |
|---|---|
| Summary: | FT: Invalid Actor when "Combining" Two Items → FT: Invalid Actor when |
FT Save Game: Right click, click hose, click can to crash ScummVM