A project for reproducing the Vertical Angle Glitch (VAG) bug in Portal 1.
VAG is a combination of several bugs that happens for some combinations of portals when an entity gets very close to the portal boundary. One of the causes of the bug is floating point rounding errors that happen when a portal teleports an entity. Reproducing VAGs and predicting when they will happen is a very frustrating part of speedrun routing.
The best tool we previously had for reproducing VAGs a an IPC script originally written in python that was then ported to SPT which would try to teleport the player close to a portal to see if a VAG was possibly for a specific portal combination. This approach is slow and often reports false results.
The main goal of this project is to provide an environment outside of the game that can reproduce the conditions needed to check if a VAG is possible on a specific set of portals. To do this, some math functions had to be replicated in assembly. You provide the portal positions, angles, which portal was placed first, etc. and the program will determine if a VAG is possible. There is currently no plan to make a CLI for this project - I plan on porting it to a more useful format in the future. See the test scripts for examples.
The old VAG testers only tested for VAGs exactly at the portal center. Then when trying to reproduce a VAG by hand you discover that it doesn't seem to be possible in the bottom half of the portal or something. It would be cool if you could figure out where in a portal a VAG was possible. Here's an example of where in a portal is VAG is possible for the following portals:
ent_fire orange newlocation "874.779541 1159.03931 891.764954 37.8783722 129.905914 0"
ent_fire blue newlocation "1189.75232 1036.28369 923.913574 -58.4471817 -44.0751495 0"
| If blue is placed first | If orange is placed first |
|---|---|
 |
 |
- white pixels: VAG not possible (normal teleport)
- red pixels: VAG
- grey pixels: you end up teleporting back to the entry portal
Often, these images look very boring. Sometimes a VAG is not possible no matter where in the portal you enter, sometimes it's possible but only in some sections. And sometimes you get images like those above (especially if one or both portals has a pitch and yaw) - you get these Moiré-like patterns that depend on the resolution of the image.
I hope to reproduce more complex chains of teleports - this has been partially successful. I can find some valid chains that result in a "double VAG" or a "negative VAG".
For an example:
- load any map
- set the name of your portals using
picker&ent_setnameto 'blue' & 'orange' - noclip
- get into blue's portal bubble in front of it (only necessary on some maps)
- run the following commands in order:
ent_fire orange newlocation "283.346893 370.18576 921.384521 0 90 0"
ent_fire blue newlocation "292.190613 31.7588444 923.996887 0 -90 0"
setpos 292.190613 31.7588463 905.996887
This behaves as if the exit portal did 2 teleports.
I'm using CMake with vs2022. When opening the project with visual studio, the CMakeSettings.json file should automatically setup 32-bit configurations.