➟ Proceed to June 2008.
The Track.Relay command is now processed, however as with the normal signal, it will currently only show the proceed aspect. Working signals are yet to come. Also, I created an article about the signals and uploaded it to the homepage. I created graphics for the speed limit post, that is three backgrounds (straight, left, right) and the digits, and also the respective object files. However, the Track.Limit command is not implemented yet.
Signals are now partially functional. When passing a signal, it turns red (or the lowest aspect it has).
The Route.Signal command is now available, and signals store the speeds applying for each of their aspects. The train also knows at which speed it is allowed to travel, both incorporating the speed limit (using Track.Limit) and the signal limit (using Track.Sig). Furthermore, the train spring has been revised. It now performs more realistic in curves and on turns, but could still be made better.
The brake indicator, which was previously loaded as a TextureShifting section element with one big texture, is now loaded as a Cyclic section element with split textures. Previously, the one big texture could not be displayed correctly on all graphics cards and was shown as pure white in these cases, while it works perfectly now with individual textures. Also, the brake indicator is fully functional now, except for the hold brake state, as the hold brake is not yet implemented. The brake indicator textures are currently the only textures that are flagged as not to be unloaded by the texture manager due to the costly process of splitting the base texture. The renderer has been reworked to increase performance, and in fact, the frame rates are about 20 to 40 percent higher than before on some routes. Also, the changes seem to have stopped the stutter that appeared on some routes even with high frame rates.
The LED-type pressure indicator is now supported, too. This makes openBVE almost fully compatible with BVE 2 panels, with only the ATC support missing. Also, the digital indicator, which was previously loaded as a TextureShifting section element, is now loaded as a Cyclic section element with split textures.
The Track.Stop command is fully supported, as the stop post can be shown now. Furthermore, textual output has been added which has basic support for Unicode text, thus providing international interoperability. As all major goals for version 0.3 have now been achieved, the new version will be made available within this week.
The hold brake (抑速制御) and the const speed (定速制御) are now available. They might work in a different way compared to BVE, but the end result is similar. Also, error handling in the route parser has been modified to do some late-bound tests for Track.Form and Track.Crack in order to accommodate for rails that are introduced after the respective commands. All parsers (route, object, train, panel) now also employ support for arbitrary encodings (codepages), so legacy BVE files from around the world (Western Windows, Japanese Shift_JIS etc.) can be read on all computers if the correct encoding is provided.
I conducted some further tests, this time around 20 routes. There are only minor problems, so all routes I tested almost completely work. Exceptions are the currently unsupported fog and some violations of mackoy's specifications by route developers. I released version 0.3.0.0 today.
I converted all remaining error generations into late-bound checks when possible, thus Track.Wall, Track.Dike, Track.FreeObj etc. commands will work when the corresponding rail is introduced after the respective command. A warning message is however generated discouraging developers to code this way as it can lead to ambiguity. The error message dialog now also distinguishes between errors and warnings and uses appropriate icons for a visual aid. The changes made so far have been released as version 0.3.0.2.
The Track.Sta and Track.Station commands can now play arrival and departure sounds. The train can now play door, emergency brake application, leave and compressor sounds. I made a small adaption to the route parser which is now able to load routes more efficiently, which will mostly affect very long routes. However, very long routes having very many objects will currently require a lot of memory. The various air and brake sounds, the loop sound and the three klaxons are now supported, too.
I have decided to switch from SDL to OpenAL whose 3D positional audio system is better suited for openBVE's 3D world. Also, doppler effects, pitch changes and speed changes are a little hard to achieve (if at all) in SDL, however seem to be an easy thing to do in OpenAL. I have created a Features: Audio section on the homepage. After many hours of frustration, I could get OpenAL to work the way I want it to. The 3D system was simple to implement and didn't take long, and it is quite nice to listen to the train in surround sound in DD 5.1. The problem was that OpenAL did not load all wave files correctly, while SDL did. Now I managed to load the sounds via SDL and pass them to OpenAL for playback, which works quite fine. Additionally, stereo to mono conversion is possible this way (which is needed for 3D capabilities), and also OGG Vorbis support, possible also MP3, AIFF, VOC, MOD and MIDI. I even tested playing 6 OGG Vorbis music files as loop.ogg sounds, one for each car, and doppler effects provided out-of-the-box while driving high-speed. However, in this extreme test (the files are about 5 minutes long), the frame rate dropped about 15%. For normal operation, expect at most 3-5% compared to having no audio at all, but still having OpenAL loaded. For some unknown reasons, having OpenAL loaded gives a penalty of around 10% even without playing any sound. I need to investigate why. To be fair, the simpler SDL audio didn't seem to have any noticable impact on the frame rate, so it could be used as a fall-back for slower computers, just without the positional effects, and thus quite silly-sounding exterior views. Motor sounds, and all other with changing pitch don't seem to be possible with SDL in a simple way, so OpenAL will be the main audio implementation for now.
Currently, I have both a functional OpenAL sound implementation and a SDL-based implementation, and can switch between the two at will. For the latter one, I also implemented a simple sort of 3D effect by adjusting the volume for each sound and making use of stereo panning. Still, SDL will not be able to handle pitch changes that are necessary for BVE's motor sounds and the like, so it will not be a feature-complete implementation. For the WIN32 platform, I could consider writing a DirectSound implementation in the future. For now, OpenAL and SDL will suffice, and of course both are portable.
The human driver AI can now look ahead and brake in advance for speed limits, signals and station stops. It can automatically open the doors and then depart again, thus it is a virtual train driver. Some adjustments have been made to the hold brake and the brake sounds. The AI can now also make use of the const speed and the hold brake systems. When trying to stop at a station which has a downward incline, the AI currently misses the stop point in some occasions.
The signal system is now able to handle the pretrain's schedule. Also, the game is fast-forwarded to skip the time the pretrain needs to be in advance, which causes a slight delay upon loading the route if the pretrain is used.
The train manager has been adapted to hide train objects (exterior, panel, etc.) from the renderer when they are far enough away. This prevents a reduction of frame rates when for example the pretrain is used but not currently in visible range. Furthermore, some mathematical simplifications have been made in the whole game.
When the effective ranges of marker images overlapped each other, there was ambiguity over which one to show, while now, multiple marker images can be shown at the same time to fix this problem. The Track.Accuracy command is now available and is not restricted to integers from 0 to 4, but can use any floating point number.
The Track.PreTrain command is now available and creates an invisible pretrain which is not accurately simulated, but simply interpolated along the scripted positions and timings. This makes the signalling system fully operational and enjoyable with simulated traffic in the form of simulated and visible trains (via Route.RunInterval) and scripted, non-visible trains (via Track.PreTrain). Thus, all main goals for version 0.4 have been achieved.
The human driver AI has been revised to employ less artificial timings and delays and also to make better use of the hold brake and const speed systems. It also performs generally better when driving down a hill.
A bug with klaxons and some other sounds has been fixed where sounds are prematurely stopped. Previously, when for example the music klaxon was stopped, it was still internally considered playing, and thus, when the user tried to play it again, it stopped another sound instead until no other sounds where playing. This has been corrected now.
As some sounds, notably 8 bit sound, played with a higher pitch than usual, I have tried to improve the conversion process that is used when loading sounds via SDL and feeding them to OpenAL. However, my attempts were not successful. Instead, I try to load sounds via ALUT (OpenAL) again when possible. Some WAV files cannot be loaded this way and stereo sounds are also not supported, but the pitch is correct. Only when the sounds cannot be loaded via ALUT or are stereo, the sounds are loaded via SDL and converted to OpenAL as done before. This still makes some sounds incompatible because of incorrect pitch, but the number of cases in which this happens should have been vastly reduced now. There was a bug in the panel configuration parser that did not discard comments correctly as was intended, which forced some statements to be incorrectly parsed. This has been corrected now and thus, comments are supported (and completely ignored). Inaccuracy is now not as regular as it was before, and vertical inaccuracy has also been added. Previously, the pretrain was forced to wait at the first signal when this signal was bound to a station's departure schedule, thus allowing the pretrain to leave only when the game started for the player, and thus, the pretrain "came out" of the player's train. This has been corrected now. The bugfixes made so far have been released as version 0.4.0.1.
The flange sounds have been added. Neither of the three sounds implemented so far work in a similar way compared to BVE, but rather different. The rub and run sounds are now basically similar to what BVE produces. The flange sound is currently played more often than BVE plays it.
The previously missing brake sound for automatic air brakes is now played when the handle is set to the service position. A lot of adjustments have been made to the run, flange and rub sounds.
The EB device (deadman's device) has been implemented, along with the Eb and Ats sounds. Also, the ATS lamps can be displayed and have working ATS, ATS RUN and EB states.
I have created textures and objects for the five different types of transponders BVE 2 supports. The objects are automatically shown when using the Track.Transponder (or Track.Tr) command and use reasonable-quality textures for such small objects. The first parts of ATS-P are implemented. The two ATS-P transponders are supported and the immediate stop can be performed. However, the brake pattern is not implemented yet.
Adjustments have been made to the ATS-P system, which now works similar to BVE, but not in the exact same way. Additionally, some missing sounds have been added, namely the ATS-P ding sound and the continuation chime for ATS-S, and the accidental departure transponder (via Track.Transponder or Track.Tr) is now also available and working.
Previously, the route parser sometimes confused indices in parantheses with the BVE 4 style of supplying arguments within parantheses and also with arguments which included parantheses themselves (like in file names). This has hopefully been resolved now.
ATC is partially working now and the speedometer can show the current speed restriction. Switching between ATS and ATC is also possible manually. Previously, when being in the exterior view, the doppler effect was totally different when being on the left side of the train than when being on the right side. This was due to a bug in the sound manager thanks to which the velocity of the camera (a vector) was wrongly assigned. This bug has been fixed. Additionally, the doppler effect was also applied to train sounds, which resulted in pitch changes even when being in the cab. For an open wagon, this is realistic, but for a closed cab, it is not. Therefore, the listener velocity is assumed to be zero when being in the cab, and non-zero when being in the external view. The ATS-SN and ATS-P systems will now always assume a red signal when a transponder is encountered beyond the last signal. Furthermore, the ToAts and ToAtc sounds are now available and the Route.Change command has been implemented.
➟ Proceed to June 2008. |