QFG4: Castle barrel room door, hero returns when exiting

[darksoul42]

On this version of ScummVM:
ScummVM 2.1.0git (Sep 9 2018 17:36:39)
Features compiled in: Vorbis FLAC MP3 ALSA SEQ TiMidity RGB zLib MPEG2 FluidSynth Theora AAC FreeType2 JPEG PNG

Stumbled upon what looks like a pathfinding/position bug in Castle Borgov, when coming back from the bedroom with the wraith, and trying to get back to the main hall. You can pick the lock (playing as a Thief) on the door and leave, but for some reason the Hero will return to the previous room immediately.

I suspect it's a problem of position evaluation for leaving the room made before the Hero has had the time to move away from the edge of the screen.

Strange thing is, it did not happen with my Paladin run.

[darksoul42]

Pick lock on the door on the right, but the Hero goes back in the previous room right away

[darksoul42]

I actually retried with my machine being under high CPU load, and managed to leave the room... It might be a timing issue. (Come to think of it, I recall some other bugs like the ones with the wraiths not triggering, or not releasing the Hero after dying also trigger somewhat randomly)

[Vhati]

The CD edition's original interpreter also does this.

Occurs in the floppy edition under ScummVM, and its original interpreter, too.

After emerging back into the barrel room, the door sound plays an extra time. This happened reliably when I had my speed slider set high. Lowering the speed slider let me reach the Great Hall.


ScummVM 2.1.0git3797-ge7d23d2cd9 (Oct 25 2018 04:17:12)
Windows 7 64bit
QFG4 Floppy 1.1a + note patch (English)

File - 5kb MD5 - Full MD5
RESOURCE.000 - f64fd6aa3977939a86ff30783dd677e1 - ff42260a665995a85aeb277ad80aac8a
RESOURCE.MAP - d10a4cc177d2091d744e2ad8c049b0ae - 3695b1b0a1d15f3d324ea9f0cc325245
RESOURCE.SFX - 3cf95e09dab8b11d675e0537e18b499a - 7c858d7253f86dab4cc6066013c5ecec

[Vhati]

SavedGame (Floppy) - Castle barrel room

[Vhati]

There's a doorMat that triggers sLeaveWest immediately upon entering the great hall.
doorMats (script 49) stay dormant when the room (global2) has its script property set.

One would expect sEnterTheRoom to preempt the doorMat until it finishes all its states and disposes itself, by which point the hero will have moved past the doorMat's region.

But rm630::init() calls setScript(sEnterTheRoom), not on itself, but on hero.

Now I'm wondering what mechanism stops this bug happening all the time.

[Vhati]

​Pull Request: SCI32: Fix QFG4 great hall entry from barrel room

[sluicebox]

This save game was made using NRS unofficial patches which are unsupported and may even be the cause of the bug on ScummVM. Even if it's not, the save game can't be loaded on ScummVM using the normal game files, it crashes.

Can this bug be reproduced in ScummVM against clean game files? If so, can a save be uploaded so I can test?

[Vhati]

@sluicebox:

NRS unofficial patches which are unsupported

Comment: bluegr on report 10711

solution is to fix the issues in the NRS patch in ScummVM, like with the other games where we fixed them in a similar fashion.

[Vhati]

SavedGame (CD) - Castle barrel room

[bluegr]

@Vhati: to be more explicit, we had issues with some of the NRS patches in the past, because the issues they solve are patched with hacks, which have caused other bugs to appear. Also, people have applied the patches in different (e.g. non-English) versions, which are incompatible.
Also, the way that scripts are patched changes their offsets, making the saved games created with them incompatible with the original.

Hence, we are trying to patch the issues fixed in the NRS patches ourselves, where posdible, and blacklist them.

[Vhati]

@bluegr:
I see. Thank you.

[sluicebox]

@Vhati Thanks for the upload, I can reproduce now

[Vhati]

Polygon - Room 630 doorMat

[Vhati]

Polygon (Fixed) - Room 630 doorMat

[bluegr]

Thanks for your work! The pull request has been merged, so this can be closed now