Category: Uncategorized

Hi everyone! Following up on Week 1, most of this week was spent resolving PVS-Studio warnings in the DM engine. Many of these warnings were responsible for crashes or pointed to code paths that could lead to unstable behavior.

By the end of the week, I was able to reduce the warning count from 103 to 13, with the remaining ones being false positives. With the static analysis issues largely addressed, I can now focus fully on gameplay-related bugs, fixing the remaining crashes, and implementing the remaining methods.

One such crash was in DisplayMan::drawDoorButton():
doorButtonOrdinal = kDMDerivedBitmapFirstDoorButton + (doorButtonOrdinal * 2) + ((doorButton != kDMDoorButtonD3R) ? 0 : doorButton - 1)
The existing logic worked for most door button values but failed when doorButton = 0, leading to an incorrect offset and eventually a crash. I corrected the expression to ((!doorButton) ? 0 : doorButton - 1), fixing the crash and restoring the intended behavior.

Another crash occurred while picking up a coin placed on a pressure plate. The game would immediately crash when processing the resulting door animation. After tracing the issue, I found that processEventDoorAnimation() unconditionally queried creature attributes before checking whether a creature group actually existed on that square. Reordering the checks fixed the heap buffer overflow and matched the behavior of the reversed source code.

I also fixed several other crashes and memory safety issues in the DM engine, including an incorrect bitmask in ProjExpl::projectileDelete() and an out-of-bounds champion access in drawPanelFoodWaterPoisoned().

Towards the end of the week, I shifted my focus back to the GUI, where I fixed a bug in the launcher that caused the collapsed state of groups to be lost when switching between views.
The fix prevents LauncherSimple::updateListing() from saving the collapsed group state during the initial build, avoiding an unnecessary overwrite of the already saved configuration and correctly preserving the collapsed groups when switching views.

I also added support for stopping fluid scrolling animations with a click, allowing the first click to stop the ongoing animation and the next click to perform the intended selection in lists, grids, and scroll containers.

Additionally, I added configuration support for fluid scrolling, allowing it to be toggled from the options menu, as the feature did not turn out to be a pleasant experience for everyone.

That wraps up this week’s progress. Next week, I’ll continue working on the remaining gameplay issues and implementing the remaining methods. Thanks for reading, and see you in the next update!

Hi everyone! Now that the GSoC coding period has begun, I moved from working primarily on the GUI to investigating crashes and gameplay bugs in the DM engine while becoming more familiar with its internals.

The first issue I looked at was a progression blocker near the start of the game. After assembling my party in the Hall of Champions, I found, as Strangerke had pointed out, that the gate to the dungeon remained shut despite stepping on the floor sensors. The issue turned out to be an incorrect early return in the sensor effect logic, preventing the pressure plate from triggering the door-opening event.

Another crash occurred when highlighting a selection in the game’s save dialog. The cause was a simple parameter ordering mistake in a box constructor used to draw the selection highlight.

Another task involved implementing EventManager::highlightScreenBox(), enabling visual highlighting for UI elements such as the movement controls.

 

Next was a save-loading bug, where the first in-game load attempt would start a new game instead of loading the selected save. Setting the correct game mode when a valid save slot was chosen resolved the issue.

Next was a stairs transition crash when moving from level 0 to level 1. The crash is currently fixed, but Sev explained that I will need to revisit this area and make the implementation aware of endianness and struct padding to achieve a fully portable solution.

 

Next was a crash that occurred whenever certain creatures, such as mummies, entered the player’s field of view.

 

Another bug involved a teleporter that was intended to appear as a normal wall. Incorrect bitfield masks caused it to render instead as a room filled with teleporter, hiding the key that was supposed to be visible on the ground.

Before:

After:

 

Next was a crash in the game’s debugger, where entering an invalid gimme command such as gimme gold key instead of the expected gimme GOLD KEY would result in a crash.

Another set of fixes involved correcting several incorrectly ordered CLIP calls across the codebase, which were responsible for a number of crashes.

I also resolved several warnings reported by PVS-Studio, addressing a number of potential issues in the codebase.

Now that I’m gaining momentum, I hope to tackle many more bugs in the coming weeks while continuing to resolve PVS-Studio warnings and improve the overall stability of the DM engine. That’s all for this week. See you in the next update!