* Add Arthurtilly to the CODEOWNERS file
* Update CODEOWNERS file to list each global owner on the same line
This should (allegedly) fix the issues of only the last entry being treated as an owner for reviews.
* Remove thecozies from CODEOWNERS file :(
---------
Co-authored-by: Gregory Heskett <gheskett@gmail.com>
Allows fast64 to know a repo's ideal settings as soon as decomp path is picked.
Settings are the default except for:
Matstack is on
Microcode is set to EX2
Force Extended RAM is off because hackersm64 already has it enabled by default
Refresh is set to Refresh 13
* Add more audio banks and add some occasional mappings for audio macros
* Remove config define for extra sound banks
---------
Co-authored-by: gheskett <gheskett@gmail.com>
If a spawned coin lands on the same frame it hits the peak of its arc, it will never have vertical speed < 0, so it will never be collectible.
Making this also set the coin to tangible if it has 0 speed fixes this issue.
Blue coins can be collected more than once if collected on the final frame of the switch timer when BLUE_COIN_SWITCH_RETRY is enabled. Fixing this on its own isn't enough though, as the blue coin switch also doesn't check for whether that coin was collected when it resets.
This allows hack collaborators to toggle debug features such as TEST_LEVEL without the need of pushing it up and negatively impacting other hack contributors. This is also useful for general HackerSM54 development, since it may be desirable to have PUPPYPRINT_DEBUG enabled by default or to not have to worry about pushing config defines up to a development branch that risk getting merged later on.
To use, simply create a new config_local.h file and place it in the config folder. Add #pragma once to the top of the file and add all the defines you wish! If you wish for something to be undefined when it's enabled by default, simply use #undef as opposed to a commented out #define. Use a combination of #undef and #define if you wish to change an existing value of a define. This local configuration will be completely ignored if DISABLE_ALL is being used in config_debug.h.
Regarding HackerSM64 Development: All feature developers are still responsible for making sure their features work in all scenarios regardless of whatever they have toggled in this file by default. This config file may be convenient but it is not an excuse to not properly test your changes!
rnc1 has noticeably slower load times in vanilla than yay0 does. gzip and mio0 are pretty useless and can be yeeted from the repo, but that's not part of this PR's scope.
• Changed the generic object death function to now also check that an object has landed and is over lava before killing it. Previously it would die instantly if it was above lava which resulted in objects randomly dying above lava
• Also changed the same function to kill a non-buoyant and killable object (e.g. goombas) right when it touches water. Previously it would slowly sink to the bottom and then instantly die upon touching the floor underwater which is strange.
• Changed where lava checks are done so they are done in the same place as other object flag checks. This avoids bugs where e.g. objects die when touching the edge of a platform that is over lava, because the ground flags are updated, then the object's floor is updated to be over the lava, and then the lava move flags are updated causing the object to think it is over lava and touching ground at the same time (making it die)
• Fixed the coin flicker define for regular non-bully coins because apparently it just never worked and no-one noticed? Only the coins that come from bullies worked with this define and I guess people only ever tested coins in lava with bullies. I rewrote the coin behavior to be a little closer to the bully one. I think in the future the two types of moving coins could easily be merged into one object, they don't act visibly different at all.
Tested the PR with a lot of goombas being dropped into lava/water and also this stuff is in Mario Builder so I've been testing it there too. Feel free to test more.
* Move audio TTLs to improve dcache
Saves over 100us even when audio is idle
* Improve additional audio runtime by reducing some unnecessary float math
* Pause channel/layer processing on sfx channels when idle
* Fix possible floating point UB/crash with gPitchBendFrequencyScale
* Preserve vanilla reverb downsampling behavior whenever BETTER_REVERB isn't in use (#727)
* Disable unrelated buttons controlling S&Q menu and pausing (#728)
* Update Dockerfile to Ubuntu 22.04 as 18.04 is EOL (#731)
* Update Dockerfile to Ubuntu 22.04 as 18.04 is EOL
* Dockerfile: Use modern syntax for mounting
* Add gcc-mips-linux-gnu to dockerfile
* Add DEBUG_ASSERTIONS define and enable it by default (#730)
* Add relevant model ID asserts to level script load commands (#722)
Co-authored-by: someone2639 <someone2639@gmail.com>
* Define changes for transitions, two bug fixes (#739)
* Fixed one or two bugs, cleaned up a few things, put every non vanilla styled change into POLISHED_TRANSITIONS
And renamed set_and_reset_transition_fade_timer
* Undid the renaming of set_and_reset_transition_fade_timer as suggested
* Fix silhouette occlusion layers rendering twice (#748)
* Fix Boo's Accelerating Upwards During Textbox (Vanilla Bug) (#742)
* Add assert for beginning geo layout command (#738)
* geo layout assert
* new idea
* Change assert message, add crash screen newline support, and change debug safeguards
* Fix weird text on Project64 GLideN64 (#715)
* Update version to 2.1.3 (#754)
* Fix rumble Issues (#756)
* Missed important use of thread6 stack define
* Fix build issues with rumble
---------
Co-authored-by: Gregory Heskett <gheskett@gmail.com>
Co-authored-by: Denis Kopyrin <aglab3@gmail.com>
Co-authored-by: MisterSheeple <27652712+MisterSheeple@users.noreply.github.com>
Co-authored-by: someone2639 <someone2639@users.noreply.github.com>
Co-authored-by: someone2639 <someone2639@gmail.com>
Co-authored-by: Lilaa3 <87947656+Lilaa3@users.noreply.github.com>
Co-authored-by: Arceveti <73617174+Arceveti@users.noreply.github.com>
Co-authored-by: 1ted59 <24933689+1ted59@users.noreply.github.com>
Co-authored-by: arthurtilly <32559225+arthurtilly@users.noreply.github.com>
Co-authored-by: iProgramInCpp <iprogramincpp@gmail.com>
* implemented a form of slope fix
* fixed the slope fix to not allow for hyper cheese
* implemented gheskett's suggestions
* Added space between comment and define
Co-authored-by: Gregory Heskett <gheskett@gmail.com>
---------
Co-authored-by: dnvic <business@dnvic.com>
Co-authored-by: Gregory Heskett <gheskett@gmail.com>
Included in the PR will be a list of comparisons.
This does two things:
1. Restores the `dl_ia_text_tex_settings` from the original Super Mario 64. Not sure why this was changed.
2. Changes the S coordinate of two vertices from `vertex_ia8_char` from 480 to 512. This is an error from the original game.
If a `Go On A Ghost Hunt` Boo is killed while it has a gravity of -1 (at the bottom of its oscillation), and does not hit a wall, it will keep the same negative gravity and continue moving until the textbox is closed. Boo Y speed isn't capped in this scenario, so the constant -1 gravity leads to the boo moving upwards at an increasing rate.
This leads to issues in `boo_act_4`. The distance eventually increases to a point where the boo object is too far to be returned in `cur_obj_nearest_object_with_behavior`.
This can have 2 buggy effects:
- the wrong dialog box being brought up if the textbox is delayed (easiest to do by hanging on a ledge).
- the puzzle solved jingle being played incorrectly if you wait in the textbox long enough
This doesn't actually cause issues with King Boo, it just brings up the wrong text box. King Boo spawning is unaffected by this bug.
The fix is to stop the boo after the death animation has finished playing, stopping the boo from accelerating upwards. I don't believe this to cause any other issues (the boo is already dead after all!)
Thanks to Kyman for finding this bug.
* Fixed one or two bugs, cleaned up a few things, put every non vanilla styled change into POLISHED_TRANSITIONS
And renamed set_and_reset_transition_fade_timer
* Undid the renaming of set_and_reset_transition_fade_timer as suggested
* add displaylist macros
* implement displaylist macro in memory.c
* API update
* implement displaylist macro on all 'hot' code paths
* Remove the macro entirely
* rename tmpDL
* remove the other two macros
* catch the last 2 missing statements
---------
Co-authored-by: someone2639 <someone2639@gmail.com>
* math util cleanup
* clean up min/max 3 and also distance and angle functions
* improve pipelining and make macros safe
* macros safer and cleaned approaches
* fix build
* made sure puppycam builds (and fixed unused var warnings)
* added 32 bit mins and maxs
* removed extraneous mat stack index decrement in geo_process_camera (#708)
* Fixed typo in FCR enums (#707)
* v2.1.1
---------
Co-authored-by: Matt Pharoah <mtpharoah@gmail.com>
* Cannon render mode fix
Slipped past me.
* Removed sharp transitions
I like the idea but it really isn´t vanilla and causes issues.
Might be worth revisiting but for now lets just get rid of it.
* fixes
Missing render mode set for transitions, reduced display list alloc for cannon
tested more this time
* Added functions to access and manipulate the fcr31 register
* Improved WiiVC emulator check to avoid false positives
* Updated comments
* Moved float.h to include directory
* Track whether the RCP hang timer has started to prevent double setting
* removed single letter struct field
* remove debug lines
---------
Co-authored-by: someone2639 <someone2639@gmail.com>
* Added config option for previewable blue coins
* Clean up previewable blue coin implementation to only be a couple lines
---------
Co-authored-by: Colin Tong <tongcolin@live.ca>
* Culling fix
Precison of half fov horizontal being equal to or above 2 leads to many issues. using 1.5 would break rt64 and other emulators with ultra widescreen.
Vertical culling will be disabled by default until 3.0
* Improved emu checks and naming
* Comment changes
* Small typo fix
* Start work on envmap rotation stuff
* n64graphic envmap rotation working
* Add comment to rotation function
* Fix DLs of non 32x32 env maps, fix envmap rotation
* fix metal flying vanish cap
* fix metal mario's medium poly butt
* new asset version for flipped env maps
* added missing env textures to extract script
* restore asset_needs_update
* Skip asset_needs_update calls if local version matches new version
* removed the goddard textures from envmap rotation
---------
Co-authored-by: mineqwerty <mineqwerty25@gmail.com>
Co-authored-by: thecozies <79979276+thecozies@users.noreply.github.com>
* Add functional changes to inertia to support only specific directions specified in the config (needs cleanup)
* Reduce number of inertia defines
* Update Platform Displacement 2 define description
* comment adjustments
* Look for shared baserom directory using XDG_DATA_DIR instead of polluting the home directory
* move global baserom folder detection to a new file that the makefile can use
* ignore __pycache__ in tools
* edit make patch to use global baserom detector
* Extract jp/sh/eu from global folder too
* genericize asset extraction for all versions
* force at least US to extract
---------
Co-authored-by: Matt Pharoah <mtpharoah@gmail.com>
Co-authored-by: someone2639 <someone2639@gmail.com>
* remove gen_asset_list (frozen)
* move format.sh
* remove rasm2armips (decompilation helper)
* remove patch_libmalloc (IDO exclusive)
* remove unnecessary sha1's and makefile definitions
* remove libultra math script (IDO exclusive)
* remove patch_elf_32bit
* actually remove bitwidth tool
* remove the need for all the sha1 files
* extract_assets can now find a baserom named anything
* added error handling and some more graceful error handling
* formatting the error message
* more relevant error info
* add a check for a roms folder
* fix gitignore to prevent patch_elf from being re-committed
* naming clarity; change to baseroms/; feedback addressed
* skip rom verification if assets have been extracted
* changes addressed
---------
Co-authored-by: someone2639 <someone2639@gmail.com>
* Improved transitions
Code quality improvements.
Shape transitions now all scale down to 0 instead of 16 (mario and bowser).
Ease in was added which leads to far smoother transitions.
HD_TRANSITIONS define added, makes use of the tex edge render mode to deliver smooth transition textures.
lerpf was added to math utils.
* Cleaner textures, rounding only during vertex creation
* Another attempt
* cozies naming suggestion
* Improvements and suggestions
SHARP_TRANSITION_TEXTURES is now off by default (I still think the textures for it could be improved).
lerpf, to_smoothstop, smoothstart, smoothstop and smoothstep have all been added.
make_vertex has been inlined.
make_simple_vertex has been added (colourless vertex).
Transition colors (and canon colors) are now defined using primative colour instead of setting individual vertices color´s.
Transition´s now use smoothstep (ease-in-out) instead of ease out only.
Naming has been improved.
* Proper function inling and removal of GBI_FLOATS variant of make_vertex and make_simple_vertex
* More requested changes
* Accidental final new line removal fixed
* Comment about pop in fix.
* Comment fix
The issue with writing comments for stuff is when you want to change code later on.
* Added emulator detection
* Fixed missing assignments
* Disable interrupts around the count factor test
* Added more comments to emutest
* Changed PJ64 enum values so the versions work as a bitfield
* Updated README
* Fixed comments
* Updated enum values so you can safely AND and OR all emulator version flags together
* Remove redundant gIsConsole, gIsVC, and gCacheEmulated global variables
* Changed console check to be more future proof against Ares
* Use assembly file instead of casting to a function pointer
* Moved round_double_to_float back to its own compilation unit
* Moved the ParallelN64 check before the Ares check for better future-proofing
* Align pj64_get_count_factor_asm with cache lines... not that it actually matters for pj64
* Adjusted some comments
* Made emulator check thread-safe on emulators that emulate PI timings
* Added gIsConsole macro for backwards compatibility
* Added cen64 to INSTANT_INPUT_BLACKLIST
* Updated comment
* Dynamic collision fixes + actual OR_LONG
Dynamic collision will now check Mario´s vertical distance seperately, this is done to implement a check for when mario is above an object, in which case the collision distance is expanded by 2000 units to avoid shadows not appearing.
Or long works correctly now, replaces the behaviour command NOP_1.
* Micro optimzation
Avoids sqrtf by using the squared lateral distance and squaring the the collision distance for the comparison.
* Requested changes
* minor change
* Re ordering
* How
* Add SET_ECHO level script command + minor BETTER_REVERB refactoring
* Add Puppyprint debug page for trying out different BETTER_REVERB presets and configurations
Also fix a missing Puppyprint enum that was causing issues
* Modify Puppyprint BETTER_REVERB page + other minor improvements
* Swap order of RSP commands to allow reverb downsampling to sound more similar to that of without downsampling
* yeet undefined_syms; IPL3 documentation; theoretical addition of checksum skip and initial DMA length/location change
* another comment about how to set dma len
* comment is now correct in ultratypes
* do the rename
* remove insinuations that ipl3 can be modified
---------
Co-authored-by: someone2639 <someone2639@gmail.com>
* Culling improvements + earlier GRAPH_RENDER_INVISIBLE check
GRAPH_RENDER_INVISIBLE is now checked during geo_process_object before any uncessesary transformation is applied to
the object, translation is still calculated for sound even if the object is invisible.
Half fov is now computed during geo_process_perspective.
Vertical culling has been added (only when bellow the screen to prevent shadow´s being culled).
Emulators have basically infinite culling aspect ratio to prevent early culling with widescreen viewport hacks.
Default culling radius is now a define.
This was written by both me and Kaze Emanuar, he provided the suggestion to use absf and informed me of
integer division being remarkably slow (although it´s only used once during geo_process_perspective).
* Badly placed new line (major fix)
* integer
Co-authored-by: thecozies <79979276+thecozies@users.noreply.github.com>
* parenthesis
doesn´t affect order of operation, just for code quality
Co-authored-by: thecozies <79979276+thecozies@users.noreply.github.com>
* uncessary whitespace
Co-authored-by: thecozies <79979276+thecozies@users.noreply.github.com>
---------
Co-authored-by: thecozies <79979276+thecozies@users.noreply.github.com>
Now its just part of HackerSM64 entirely, and HackerSM64 will be the source of all changes (as long as said changes can reasonably apply to homebrew games)
* Disable stereoheadseteffects and stub out all known code references via define
By effect this also completely removes Headset sound options from the game config, as it serves no purpose whatsoever compared to stereo.
Because it's stubbed out everywhere rather than only the necessary locations to prevent it, this should boost audio runtime slightly for everything, not just the stuff that actively used it.
* Rename DISABLE_HEADSET_STEREO_EFFECTS to ENABLE_HEADSET_STEREO_EFFECTS
* Reorganize and reduce audio Note struct
* Add performance to define description
* Rename ENABLE_HEADSET_STEREO_EFFECTS to ENABLE_STEREO_HEADSET_EFFECTS
Vanilla's ramping table sounds absolutely horrendous, most noticeably when it comes to slow volume fade-ins/outs. By default, precision has been increased by 4x vanilla US, which overall has an insignificant impact on performance. This is still half as precise as EU's volume ramping table, that's about the point where additional memory and performance aspects are impacted more significantly.
This allows sounds to start instantly without any fade-in. Using new soundbanks with delay values of 0 will not compile with base decomp, but old decomp-compliant soundbanks will work with this commit.
Refactored everything to be more memory efficient and execute fewer function calls than prior
Also accidentally discovered a significant bug with the audio modes always being forced to stereo so this is definitely 2.1 scope lmao
* Start samples at full velocity when using envelopes that start with a delay of 1
This can help eliminate sample artifacting that happens during a transition between a sudden fade-in into normal volume. It's not completely eliminated due to some processing weirdness, but it's a massive improvement.
* Fix vanilla issue causing misalignment when samples are initialized and played back with low attack values
NOW the artifacting is actually gone for real
* Improve artifacting bugfix
The first solution would've honestly been satisfactory, but this one seems to be slightly more accurate while actually removing code rather than adding it
* Fix rooms and doors
* Update object_helpers.c
* Update object_list_processor.h
* Update object_list_processor.c
* Fix formatting
* Revert a formatting
* Clean up geo_switch_area
* Clean up geo_switch_area (more)
* Move TransitionRoomData struct
* Revert star door collision check optimization
* Move load_object_collision_model and bhv_door_rendering_loop back out of star_door.inc.c
* Combine metal door model IDs
* Revert metal door sound fixes
* Add light configuration settings to BETTER_REVERB
This can reduce runtime demand down to ~75% of the standard demand, at the cost of the configurability of generally more advanced parameters
* Reformat the BETTER_REVERB preset entries to be easier to look at
* :peterVOID:
* haha lol formatting change definitely related to reverb yes this isn't unwarranted at all lmao hehe hoho
* Puppyprint text size
* Update puppyprint.c
* Revert broken render_multi_image changes
* revert multi image tweak
parallel momento
* Update puppyprint debug colors to work with new hexadecimal system
* Puppyprint color reversion to currEnv + some bugfixes
* Remove a pointless color array
* Add forgotten comment
* i can type yes definitely
* Fix some puppyprint formatting edge cases
* Missed one last puppyprint edge case
* Additional strLen check
* Variable length fix + command ignoring strlen
* type changes
* Update puppyprint.c
* Box clamping
* Revamped profiling, fixed right alignment
* Update puppycam2.c
* Deferred printing
* Update puppyprint.c
* Fixed negatives being stupid
* Puppyprint Memory view rework
* Update level_script.c
* tweaks
* Fix new line spacing with different scales
also small bit of optimisation
* Spacing fixes
* Lightweight text option
* puppyprint debug now using light text where applicable
* Update puppyprint.c
* Update puppyprint.c
* level select menu
* merge profilers into one WIP
* Fully ascii compliant
* Finishing touches hopefully
* Update puppyprint.c
* new font
* vanilla font
* o
* Update hud.c
* last kerning fixes
* Requested changes
* requested changes
* fix the j
* requested changes
* change ur clothes
* Update surface_collision.c
* requested changes
* Update puppyprint.h
* Update camera.c
* .
* Update puppyprint.c
* Update puppyprint.c
* Update puppyprint.c
* fix J
* font
* font system rework
* ia4 outline font
* pain
* fix deferred printing
* fix format change
* Add audio profiling to Puppyprint Debug
* Optimize audio profiling substantially
* Minor audio optimizations
* requested changes
* oops accidentally commited 6 lmfao
* Update game_init.c
* Update puppyprint.c
* Update puppyprint.c
* Improve Puppyprint deferred prints
This is safer, more readable, and more runtime efficient all in the same package
* change return type
* fix building with profiler only
* Update profiling.c
* fix upscaled texture rectangles from multi image
---------
Co-authored-by: gheskett <gheskett@gmail.com>
* Update game_init.c
* fuck you nintendo fuck you nintendo fuck you nintendo fuck you nintendo fuck you nintendo fuck you nintendo
* Update segment2.c
* Update model.inc.c
* Small improvements to object loading within rooms
* Only check if Mario is in the same room as the object once
* Implement object render distance within rooms (from thecozies)
* Remove sLevelsWithRooms (also from thecozies)
Honestly surprised this wasn't already in Hacker
* Remove comment for sLevelsWithRooms in config_game
* Make compatible with the painting objects PR
Move room detection to get_room_at_pos (from Arceveti)
* address enum comment
* Replace debug libultra builds with their ido counterparts
This is necessary because GameCube controllers apparently rely on the IDO build, also the release build is IDO anyway
* Remove fallthrough warnings from usb/debug.c
* remove need for debug libultra with unf and use lib define in ld
Co-authored-by: thecozies <collinpferguson@gmail.com>
* Bugfix: BETTER_REVERB starts before buffers are ready
* Add support for preset options to be used with BETTER_REVERB
* Some BETTER_REVERB cleanup
* The funny whoops
* Address suggested level script change
* Bugfix: useReverb crashes when disabled
* Allow window size override for vanilla-esque BETTER_REVERB presets
* Remove unnecessary downsampling limit check
* Fix up a few things + some slight reformatting
* Add light documentation to the BETTER_REVERB parameters
* Adjust BETTER_REVERB preset types to make more sense
* Reset BETTER_REVERB data buffer to be unallocated when not in use
* Properly disable BETTER_REVERB when disabling all reverb processing
* Remove an unnecessary BETTER_REVERB ifdef
* Remove D_80332108
* Revise BETTER_REVERB config description
* various tweaks to tackle redundancy
* Update collision.inc.c
* inline the function
* fix surface denorms
* oops forgot to remove a comment that I just proved false lol
* stack asserts
* Update main.c
* Update main.c
* fix bug plural s
* Update main.c
* comment
Optimize vibrato to even exceed vanilla's implementation, all without restricting usability from the current implementation. This should save up to about 1ms of audio runtime.
* Imminent fixes for bugs found on master branch (#512)
* The vanilla level checks define for Yoshi is inverted, causing him to require 120 stars when it is off and appear at 0 stars when it's on
* The downwarp fix results in Mario levitating in midair when grabbing Bowser midair. While downwarps should still be fixed, the change should be reverted immediately until a better fix is made.
* Some checks of Mario's floor class were using the wrong defines, which can lead to unexpected behavior in the event anyone wants to reorder surface types. By default the SURFACE_CLASS_SLIPPERY and SURFACE_SLIPPERY defines have the same value, which is why this mistake is hard to notice.
* The firsty frames define was implemented poorly, not allowing for vanilla firsty behavior no matter what the values were set to. This has been reverted, while avoiding the UB in the original vanilla code.
* Removed the ledge grab changes that fix QSLGs and change the false ledgegrab define since Arceveti wanted to in his PR
* Update version to 2.0.4 (#515)
* HackerSM64 2.0.5 (#517)
* Bugfix: race condition with accessing cleared audio memory pools
* Bugfix: skyboxes access tiles out of range when looking straight down (#518)
* Vscode fixes (#516)
* Revert s32 angles and remove Angle typedef usage (#529)
* Fix some surface shadowing (#531)
* Delete rtc.h
* Remove `wait_for_audio_frames()`
Co-authored-by: thecozies <79979276+thecozies@users.noreply.github.com>
Co-authored-by: Arceveti <73617174+Arceveti@users.noreply.github.com>
Co-authored-by: JoshDuMan <40190173+JoshDuMan@users.noreply.github.com>
* Update VERSION.txt (#534)
* HackerSM64 2.0.6 (#539)
* revert random_u16 to be a u16 (#538)
* Added safety gcc math flags (#537)
* removed useless compilation flags that could cause performance loss
Co-authored-by: thecozies <79979276+thecozies@users.noreply.github.com>
Co-authored-by: thecozies <collinpferguson@gmail.com>
* Update VERSION.txt
HackerSM64 2.0.6
* Add GCC opt flag to suppress warnings
* v2.0.7
* 2.0.8 (fix more compiler warnings) (#543)
* fix Waddress in mario_step.c
* fix Winfinite-recursion in unused dynlist_proc function
* in fact remove the offending function entirely
* fix warnings in RNC decompress functions
* update version
* fix version in VERSION.txt
premature on my part, will update it to 2.0.10 next time
Co-authored-by: someone2639 <someone2639@gmail.com>
Co-authored-by: arthurtilly <32559225+arthurtilly@users.noreply.github.com>
Co-authored-by: axollyon <20480418+axollyon@users.noreply.github.com>
Co-authored-by: Gregory Heskett <gheskett@gmail.com>
Co-authored-by: thecozies <79979276+thecozies@users.noreply.github.com>
Co-authored-by: JoshDuMan <40190173+JoshDuMan@users.noreply.github.com>
Co-authored-by: thecozies <collinpferguson@gmail.com>
Co-authored-by: Reonu <danileon95@gmail.com>
Co-authored-by: someone2639 <someone2639@users.noreply.github.com>
Co-authored-by: someone2639 <someone2639@gmail.com>
* fix Waddress in mario_step.c
* fix Winfinite-recursion in unused dynlist_proc function
* in fact remove the offending function entirely
* fix warnings in RNC decompress functions
* update version
* fix version in VERSION.txt
premature on my part, will update it to 2.0.10 next time
Co-authored-by: someone2639 <someone2639@gmail.com>
* The vanilla level checks define for Yoshi is inverted, causing him to require 120 stars when it is off and appear at 0 stars when it's on
* The downwarp fix results in Mario levitating in midair when grabbing Bowser midair. While downwarps should still be fixed, the change should be reverted immediately until a better fix is made.
* Some checks of Mario's floor class were using the wrong defines, which can lead to unexpected behavior in the event anyone wants to reorder surface types. By default the SURFACE_CLASS_SLIPPERY and SURFACE_SLIPPERY defines have the same value, which is why this mistake is hard to notice.
* The firsty frames define was implemented poorly, not allowing for vanilla firsty behavior no matter what the values were set to. This has been reverted, while avoiding the UB in the original vanilla code.
* Removed the ledge grab changes that fix QSLGs and change the false ledgegrab define since Arceveti wanted to in his PR
2022-10-26 19:12:29 +13:00
288 changed files with 34312 additions and 17658 deletions
- Lighting Engine is NOT finished. It's a WIP. Expect issues and expect it to be somewhat hard to use. Point lights in particular are extremely finnicky.
- In order for lighting engine to work, your materials must have Shade Color in the combiner, **but you have to DISABLE Shade Color in sources**. If you leave Shade Color ticked for even a single material, the lighting will break **for the entire mesh**. Until Fast64 gets some sort of Lighting Engine support, it's recommended you create some custom presets to make this less annoying.
- You can intentionally make a material fullbright by leaving Shade Color out of the combiner for that material. Useful for, say, a lamp.
- Use the function `set_directional_light` to set the directional light for the current area. It takes a `Vec3f` for the direction and three `s32` for the RGB values.
- Use the function `set_ambient_light` to set the ambient light for the current area. It takes three `s32` for the RGB values.
- You can use the function `emit_light` to emit a point light via code. It takes a `Vec3f` for the position, three `s32` for the RGB values and three `u32` for each one of the falloff types. The falloff is basically how strong the light is: the HIGHER the falloff is, the DIMMER the light will be. The three types of falloffs can be stacked, and work like this:
- 2 for constant falloff = the point light is half as bright always.
- 2 for linear falloff = the point light is half as bright when you double the distance
- 2 for quadratic falloff = the point is half as bright when you double the square of the distance
- Important note regarding point lights: They are affected by WORLD_SCALE, which is set based on your extended bounds mode. Esentially, at 2x bounds lights will be twice as bright as in 1x bounds, and with 4x bounds they will be twice as bright as that. This can be dealt with simply by using high falloffs, but it can be annoying if you decide to change the extended bounds mode mid-development.
- You can also make an object emit light by using the `OBJ_FLAG_EMIT_LIGHT` object flag, as well as `SET_LIGHT_COLOR` and `SET_LIGHT_FALLOFF` in the script.
- **Pyro Jay**: Texture improvements, repo banner art, some QoL stuff
- **CrashOveride**: creating the [ultrasm64](https://github.com/CrashOveride95/ultrasm64) repo
- **falcobuster**: Original coordinate overflow fix (world scale), ASM version of extended bounds
- **falcobuster**: Original coordinate overflow fix (world scale), ASM version of extended bounds, emulator detector
- **anonymous_moose**: porting falco's extended bounds to decomp
- **tuxlovesyou**: `LOAD_MIO0_TEXTURE` macro and moral support
- **devwizard**: the PJ64 pre-v3.0 detection part of the emulator detector
Thanks to Frame#5375 and AloXado320 for also helping with silhouette stuff
**Lighting Engine by Wiseguy**
- Lighting Engine is available on a separate branch ([base/lighting-engine](https://github.com/Reonu/HackerSM64/tree/base/lighting-engine)). Instructions on how to use it are in the readme of that branch.
-Alternatively, the main repo has `Puppylights` available, which is a more lightweight, but limited lighting library intended to be used to modify existing light properties. You can look at `puppylights.c` to find out how to use it.
- Lighting Engine is NOT finished. It's a WIP. Expect issues and expect it to be somewhat hard to use. Point lights in particular are extremely finnicky.
-In order for lighting engine to work, your materials must have Shade Color in the combiner, **but you have to DISABLE Shade Color in sources**. If you leave Shade Color ticked for even a single material, the lighting will break **for the entire mesh**. Until Fast64 gets some sort of Lighting Engine support, it's recommended you create some custom presets to make this less annoying.
- You can intentionally make a material fullbright by leaving Shade Color out of the combiner for that material. Useful for, say, a lamp.
- Use the function `set_directional_light` to set the directional light for the current area. It takes a `Vec3f` for the direction and three `s32` for the RGB values.
- Use the function `set_ambient_light` to set the ambient light for the current area. It takes three `s32` for the RGB values.
- You can use the function `emit_light` to emit a point light via code. It takes a `Vec3f` for the position, three `s32` for the RGB values and three `u32` for each one of the falloff types. The falloff is basically how strong the light is: the HIGHER the falloff is, the DIMMER the light will be. The three types of falloffs can be stacked, and work like this:
- 2 for constant falloff = the point light is half as bright always.
- 2 for linear falloff = the point light is half as bright when you double the distance
- 2 for quadratic falloff = the point is half as bright when you double the square of the distance
- Important note regarding point lights: They are affected by WORLD_SCALE, which is set based on your extended bounds mode. Esentially, at 2x bounds lights will be twice as bright as in 1x bounds, and with 4x bounds they will be twice as bright as that. This can be dealt with simply by using high falloffs, but it can be annoying if you decide to change the extended bounds mode mid-development.
- You can also make an object emit light by using the `OBJ_FLAG_EMIT_LIGHT` object flag, as well as `SET_LIGHT_COLOR` and `SET_LIGHT_FALLOFF` in the script.
**Puppycam**
- Puppycam is available on the master branch now, you can toggle it in `config/config_camera.h`. *
@@ -82,7 +93,7 @@ Thanks to Frame#5375 and AloXado320 for also helping with silhouette stuff
- Many general use defines for object struct members, meant for use in custom object behaviors. Check `include/object_fields.h` for more info on this. (By MrComit)
- Included `actors/group0.c` in `behavior_data.c`
- The internal ROM name is now set with a define in `config/config_rom.h` to make it simpler
- There is a `gIsConsole` variable that is 1 when running on console and 0 when running on emulator. This way you can wrap your code in a console check.
- There is a `gEmulator` variable to detect console or specific emulators and emulator versions
- Expanded audio heap allows for a larger concurrent note count and the importing of more m64 sequences and sound banks (By ArcticJaguar725) *
- You can set a test level in `config/config_debug.h` in order to boot straight into it, so you can quickly test the level you're working on. *
- Allow all surfaces in the game to have a `force` parameter. Activating this doesn't REQUIRE you to set `force` for every surface: If you don't set, it will default to 0x0000 rather than crashing. Increases RAM usage of collision. *
@@ -105,7 +116,7 @@ Thanks to Frame#5375 and AloXado320 for also helping with silhouette stuff
**Neat Misc. Changes:**
- Instant Input patch by Wiseguy (Removes all input lag caused by plugins supporting framebuffer)
- This means that you'll have to do your framebuffer effects on buffer 0 for emulator, but NOT for console. You can use the `gIsConsole` variable to check for console when doing your framebuffer effects.
- This means that you'll have to do your framebuffer effects on buffer 0 for emulator, but NOT for console. You can use the `gEmulator` variable to check for console when doing your framebuffer effects.
- Widescreen (16:9) support toggleable by pressing `L` in the pause menu. *
- S2DEX engine by someone2639! To use it, compile with `make TEXT_ENGINE=s2dex_text_engine` or just set `TEXT_ENGINE` to `s2dex_text_engine` in the makefile.
- ia8 (64x64) coins, the vanilla coin texture is upgraded to accomodate. *
@@ -121,7 +132,7 @@ Thanks to Frame#5375 and AloXado320 for also helping with silhouette stuff
- Use `render_multi_image` to draw large texture rectangles consisting of multiple images on the screen.
- More info in `puppyprint.c`
- Wiseguy's Farcall TLB mapping allows to store executable code inside uncompressed segments, that can be loaded and ran as needed, instead of it having to be loaded at all times. See `farcall.h` in the include folder for instructions and details.
- Red Coin Stars now support up to 99 red coins! In addition, multi-area red coin missions can now be created by setting the 2nd behavior paramater of the red coin star to the number of reds required for the star to spawn.
# UltraSM64
- This repo contains a full decompilation of Super Mario 64 (J), (U), (E), and (SH).
@@ -135,7 +146,7 @@ Thanks to Frame#5375 and AloXado320 for also helping with silhouette stuff
- It has been patched with someone2639's shiftable segments patch
- Wiseguy's instant input patch has been added to allow for less input lag on emulation (Does not affect console)
This does mean that any framebuffer effects will have to be done on buffer 0 if targeting emulators
- Automatic console and emulator detection: Use the `gIsConsole` variable to wrap your code in an emulator check.
- Automatic console and emulator detection: Use the `gEmulator` variable to wrap your code in an emulator check.
- Separate defines for emulator and console black border height.
- Getting HVQM FMV support to work with the game is WIP.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
