Forth by luginf · Pull Request #2947 · nesbox/TIC-80

luginf · 2026-05-31T14:26:07Z

adding Forth langages to TIC-80 (using pforth)

luginf · 2026-05-31T14:31:51Z

(made with the help of Claude code)

pforth cell_t = uintptr_t, so the pre-compiled dictionary is cell-size-dependent (64-bit on x86-64, 32-bit on WASM). Loading a 64-bit dictionary in a 32-bit runtime corrupts all addresses and freezes the HTML player. - cmake/pfdicdat_32.h: new 32-bit dictionary (generated with gcc -m32) - cmake/forth.cmake: select pfdicdat_32.h when CMAKE_SIZEOF_VOID_P == 4

ddelemeny · 2026-06-01T15:46:22Z

Note: OP had the courtesy to open a discussion (#2938) regarding the project's AI contribution policy before vibe-coding this payload.

Resolving the policy matter should probably be considered a prerequisite to reviewing this PR.

MineRobber9000

I don't know enough about pforth to critique the actual implementation of the API, but there are definitely some spots in here that I am concerned about.

MineRobber9000 · 2026-06-01T15:38:01Z

+            message(STATUS "Forth: copying bundled ${_PFORTH_DICDAT_BUNDLED} to ${PFORTH_DICDAT}")
+            configure_file(${_PFORTH_DICDAT_BUNDLED} ${PFORTH_DICDAT} COPYONLY)
+        else()
+            message(STATUS "Forth: pfdicdat.h not found — bootstrapping pforth to generate it...")


If we can use pforth to bootstrap the pfdicdat.h generation, why are we bundling a version of this data anyways? What happens if pforth updates in a way that adds more stuff to the dictionary, and now our bundled data is out of date?

@MineRobber9000 that's a good question. I've initially let it generate the pfdicdat.h and pfdicdat_32.h from pforth but there was a problem with some targets (nintendo3D and rpi) do I've included those headers but it's not ideal.

It should be fixed now. I've tested the nintendo 3DS build and it's working fine.

MineRobber9000 · 2026-06-01T15:40:54Z

+\ license: MIT License
+\ script: forth
+\
+\ -- TIC-80 Forth API --------------------------------------------------


You don't need to leave this API reference in the code.

MineRobber9000 · 2026-06-01T15:42:52Z

+;
+
+\ -- Demo 2 : sprite + arrow keys -------------------------------------
+: demo02


This is the only one you actually need to have. The others are all just clutter that a user would have to delete before starting their own game.

sure. I've removed all that, including the API references and additional examples. Do you think it should be better to include such informations in an additional file in the demos/ folder under a new name, or leave this to the wiki, or even somewhere else? I see in the demos folder some lua samples like fire.lua or tetris.lua so I don't know

MineRobber9000 · 2026-06-01T15:48:13Z

+
+: BOOT
+  96 px !  24 py !
+  S" Forth BOOT" 12 TRACE


Remove this, no other language does this. (EDIT: to clarify, since GH shows a seemingly random collection of lines and not the one you explicitly select in the interface, this is referring to tracing "Forth BOOT" in the BOOT callback.)

ok. I've just kept the BOOT procedure for initialising the variables, and kept the "demo01" word outside of the TIC loop, so people can just comment the demo and start creating new words.

MineRobber9000 · 2026-06-01T15:53:41Z

@@ -0,0 +1,129 @@
+// MIT License


This should also go in forth.c. In general, all languages should be one C file apiece; the only exception are the Lua languages (Lua, MoonScript, Yue, and Fennel), since they all share the same plumbing underneath and only differ in how they're initialized.

it has been merged into the forth.c

removing cmake/pfdicdat.h and cmake/pfdicdat_32.h from tree updating cmake/forth.cmake for bootstraping pfdicdat

…3ds, rpi)

MineRobber9000 · 2026-06-02T21:37:19Z

How did you manage to chmod +x the whole repo?

luginf · 2026-06-02T22:07:53Z

How did you manage to chmod +x the whole repo?

oops, long story short I needed to edit some files from another (local) account (on my computer). I can try to revert the 755 to 644 if you wish, but that will trigger a new build for nothing.

MineRobber9000 · 2026-06-03T07:44:56Z

You definitely should undo it. I can't review the PR in this state since I'd have to scroll through basically every file in the repo just to see what you actually changed.

luginf · 2026-06-03T08:11:36Z

You definitely should undo it. I can't review the PR in this state since I'd have to scroll through basically every file in the repo just to see what you actually changed.

Ok, i've fixed it, I'm just working on another problem (wrong/misleading message during a stack underflow) and I'll commit that ASAP

fix "stack underflow" message

MineRobber9000 · 2026-06-03T10:06:43Z

+: demo01
+  0 BTN IF  py @ 1- py ! THEN
+  1 BTN IF  py @ 1+ py ! THEN
+  2 BTN IF  px @ 1- px ! THEN
+  3 BTN IF  px @ 1+ px ! THEN
+  13 CLS
+  1 px @ py @ 14 3 0 0 2 2 SPR
+  S" Hello Forth!" 84 84 15 FALSE 1 FALSE PRINT DROP
+;
+
+: TIC
+demo01
+;


Suggested change

: demo01

0 BTN IF py @ 1- py ! THEN

1 BTN IF py @ 1+ py ! THEN

2 BTN IF px @ 1- px ! THEN

3 BTN IF px @ 1+ px ! THEN

13 CLS

1 px @ py @ 14 3 0 0 2 2 SPR

S" Hello Forth!" 84 84 15 FALSE 1 FALSE PRINT DROP

;

: TIC

demo01

;

: TIC

0 BTN IF py @ 1- py ! THEN

1 BTN IF py @ 1+ py ! THEN

2 BTN IF px @ 1- px ! THEN

3 BTN IF px @ 1+ px ! THEN

13 CLS

1 px @ py @ 14 3 0 0 2 2 SPR

S" Hello Forth!" 84 84 15 FALSE 1 FALSE PRINT DROP

;

If the user wants to write their own code, they'll probably want to just delete the whole demo code, so having it be a separate word is just clutter.

MineRobber9000 · 2026-06-03T10:23:08Z

Also, you forgot to install gcc-multilib in the Android task, which is why that build is failing.

… task.

…ries

fix mac os build by limiting the number of parallel processes

- Remove references to committed cmake/pfdicdat.h and pfdicdat_32.h (deleted in PR nesbox#2947 forth branch) - Document that cmake/forth.cmake now always regenerates pfdicdat.h at configure time via file(REMOVE) + bootstrap - Add critical warning: a stale 32-bit dict in a 64-bit binary causes a silent segfault when selecting "new forth" (root cause documented) - Update platform/cell-size matrix (WASM/3DS/RPI = 32-bit) - Remove now-outdated manual regeneration instructions Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-authored-by: minerobber <milesrobert374@gmail.com>

MineRobber9000

LGTM!

… clash with FORTH standard words

MiniScript arrived first, so the conflicts were adjacency again: the language list in src/script.c, the sources in core.cmake, the includes in CMakeLists, and the submodule list, which now carries both runtimes. Both sides are kept everywhere.

The merge left the two additions nested inside one another, which the compiler read as an unterminated conditional. The file is main's again with the Forth blocks beside MiniScript's, in both the extern list and the script array; a native build of the merge links.

nesbox · 2026-09-18T17:22:08Z

Merged main in (MiniScript landed first). The conflicts were adjacency again: the language list in script.c, the sources in core.cmake, the includes in CMakeLists and the submodule list; both sides kept. My first pass nested the two script.c blocks inside one another and the compiler read that as an unterminated conditional — the file is main's again with the Forth blocks beside MiniScript's in both the extern list and the script array, and a native build of the merge links.\n\nOn the CI side: the release lists are not touched here either, and that is ours to grow — forth and miniscript in one pass, before the next release, as the owner asked. Your two build.yml changes look right and unrelated to languages: gcc-multilib for the nintendo host toolchain, and --parallel 3 on the macOS release job, which is that runner's core count.

nesbox

Read the binding and the wiring. The runtime arrives as a submodule of upstream pforth, cmake/forth.cmake names the kernel sources and takes the no-IO build with TIC-80's own shim, and the script registers as forth with .fth — the release lists are ours to grow, as with MiniScript. Worth saying plainly: the tick has no time bound, so a runaway Forth loop hangs the client exactly as a runaway Lua or JS loop does; that is the house behaviour and MiniScript is the exception, not Forth the deficiency. The two build.yml changes are unrelated to languages and both check out.

Why: the table lists every runtime the build carries, and the two languages #2992 and #2947 added were simply absent from it — nothing else in the client was told what version they are. Neither vendor keeps its version where the older ten keep theirs, which is why this was left out of the stub-list pass: MiniScript's is in the map its own `version` intrinsic returns (vendor/miniscript2/generated/CoreIntrinsics.g.cpp; `hostVersion` beside it is the embedding application's, not the runtime's) and pForth's is PFORTH_VERSION_NAME in vendor/pforth/csrc/pf_guts.h. The name column grows to ten characters for "miniscript", which was the one name wider than the field — without it that row's version lands a column to the right of every other. Verified by running it: `tic80 --cli --cmd="help version"` prints "forth pForth 2.2.0" and "miniscript MiniScript 2.0" among the rest. Co-authored-by: Claude Code <noreply@anthropic.com>

luginf · 2026-09-18T21:19:00Z

@nesbox thank you very much, and also to @MineRobber9000 for the assistance!

Collapses the fork's 54 commits since the upstream/main merge-base into one. Note: the actual Forth work is already in upstream/main (nesbox#2947, squash-merged as bbb0e012), so this commit's real diff is small: fork-local additions that never went upstream (CLAUDE.md project docs, .claude/skills.md, .gitignore, a vendor/miniscript2 bump, a build/html/index.html tweak). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

luginf added 4 commits May 31, 2026 11:45

adding forth (using pforth)

4165603

forth now working

e800be7

working example

54641a1

fix demo cart

e473ca5

luginf added 6 commits May 31, 2026 18:43

fix missing pfdicdat from pforth

ab80697

add missing asset

92cd9cc

adding pfdicdat.h

d822781

fix rpi baremetal build

f9cc26e

fix pfdicdat for html build (32 bit)

6352221

luginf force-pushed the forth branch from d176dc5 to 6352221 Compare June 1, 2026 08:23

MineRobber9000 suggested changes Jun 1, 2026

View reviewed changes

luginf added 4 commits June 2, 2026 18:37

fusion of api/forth.h and api/forth_io.h

688734d

removing cmake/pfdicdat.h and cmake/pfdicdat_32.h from tree updating cmake/forth.cmake for bootstraping pfdicdat

fix forth.cmake & workflows to bootstrap on crossplatforms (nintendo …

9b16fdc

…3ds, rpi)

rebuild forthdemo card

091f503

fix 3DS build

787f601

luginf requested a review from MineRobber9000 June 2, 2026 19:03

fix wrong permission

a6643bf

fix "stack underflow" message

MineRobber9000 suggested changes Jun 3, 2026

View reviewed changes

luginf added 4 commits June 3, 2026 16:10

updating forthmark. Clean forthdemo. Fix missing multilib for android…

f9c1ebb

… task.

adding 5 minutes timeout in case of problem during generation of bina…

1551c18

…ries

fix windows build.

8762b9b

fix mac os build by limiting the number of parallel processes

fix windows build for real this time?

7d7b473

luginf added 2 commits June 4, 2026 21:40

update forth.cmake for windows build

0e3a08c

compile in release mode on windows, change also timeout

69452cf

luginf requested a review from MineRobber9000 June 4, 2026 20:53

MineRobber9000 suggested changes Jun 5, 2026

View reviewed changes

Comment thread src/api/forth.c Outdated

Comment thread src/api/forth.c Outdated

Comment thread assets.bat

Comment thread assets.bat

luginf and others added 5 commits June 5, 2026 08:25

Update src/api/forth.c

25588aa

Co-authored-by: minerobber <milesrobert374@gmail.com>

Update src/api/forth.c

896a8a8

Co-authored-by: minerobber <milesrobert374@gmail.com>

Update assets.bat

6389821

Co-authored-by: minerobber <milesrobert374@gmail.com>

Update assets.bat

ef3fd39

Co-authored-by: minerobber <milesrobert374@gmail.com>

adding forthmark.tic.dat

6e47191

MineRobber9000 approved these changes Jun 6, 2026

View reviewed changes

luginf and others added 5 commits June 22, 2026 12:32

updating API: EXIT renamed to EXITGAME and KEY to KEYPRESSED to avoid…

4c4201d

… clash with FORTH standard words

add individual forth build in the ci workflow

fb4fc87

Merge branch 'upstream-main' into forth

18113f3

Merge main into the Forth branch

a5e8f4c

MiniScript arrived first, so the conflicts were adjacency again: the language list in src/script.c, the sources in core.cmake, the includes in CMakeLists, and the submodule list, which now carries both runtimes. Both sides are kept everywhere.

nesbox approved these changes Sep 18, 2026

View reviewed changes

nesbox merged commit bbf0e01 into nesbox:main Sep 18, 2026
11 checks passed

This was referenced Sep 18, 2026

release: grow the stub lists with MiniScript and Forth #3009

Merged

help version: name the two runtimes added after 1.2.0 #3011

Merged

nesbox mentioned this pull request Sep 20, 2026

Forth and YueScript share script id 21 #3024

Closed

Conversation

luginf commented May 31, 2026

Uh oh!

luginf commented May 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ddelemeny commented Jun 1, 2026

Uh oh!

MineRobber9000 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MineRobber9000 Jun 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MineRobber9000 commented Jun 2, 2026

Uh oh!

luginf commented Jun 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

MineRobber9000 commented Jun 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

luginf commented Jun 3, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

MineRobber9000 commented Jun 3, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

MineRobber9000 left a comment

Choose a reason for hiding this comment

Uh oh!

nesbox commented Sep 18, 2026

Uh oh!

nesbox left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

luginf commented Sep 18, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

luginf commented May 31, 2026 •

edited

Loading

MineRobber9000 Jun 1, 2026 •

edited

Loading

luginf commented Jun 2, 2026 •

edited

Loading

MineRobber9000 commented Jun 3, 2026 •

edited

Loading