SDL2 on Kandelo - GLSL shader playground demo by mho22 · Pull Request #709 · Automattic/kandelo

mho22 · 2026-06-16T14:33:57Z

Why

Kandelo gained a DRI/KMS graphics surface in #678, but nothing in user space drove it through a real graphics toolkit. This PR ports SDL2 2.30 to the wasm32 sysroot — KMSDRM video on /dev/dri/card0, /dev/dsp audio, evdev input — and ships a Shadertoy-style GLSL shader playground demo on top of it. The playground exercises the whole stack end to end in the browser: a GLES2 fragment-shader render pane, PCM synthesis through a "sound shader", and keyboard-driven editing.

Second layer of the six-PR DRI stack (#698 → #709 → #948 → #976 → #1318 → #1241). It stacks on explore-dri-evdev-and-alsa.

What changed

libdrm and libgbm

Upstream libdrm 2.4.120 replaces the hand-written stub. Everything above this point in the stack — SDL2's KMSDRM backend, and later wlcompositor — talks to the real library rather than a local approximation.
A surface ring in libgbm for SDL2 to scan out through (gbm_surface_*).

SDL2 port

KMSDRM video and evdev input backends, with connector mode-info handling.
libEGL / libGLESv2 wired statically (-DSDL_VIDEO_STATIC_ANGLE=1).
A polling-audio patch (SDL_OpenAudioDevice + SDL_PumpAudioDevices) so audio works under the single-threaded SDL_THREADS_DISABLED model, playing through main's OSS /dev/dsp.
configure overrides for the cross build, plus the dynapi patch.

GLSL shader playground (programs/sdl2/)

A split-pane app on a 1920×1080 /dev/dri/card0 surface: a 2/3-width gap-buffer code editor beside a 1/3-width live GLES2 render pane, auto-recompiling 250 ms after the last keystroke, with an error strip on failure, last-good-shader retention, F5 reload and Ctrl+S persist.
Editor layer: syntax highlighting, selection (Shift+nav), clipboard (Ctrl+C/X/V), select-all, coalesced undo/redo, auto-replace-on-type, and vertical column memory.
Sound-shader mode (F2): a Shadertoy-style vec2 mainSound(float) rendered to an FBO, read back in ≤64 KB row-bands and decoded RGBA8 → S16 as the playback source. A 4-channel chip-synth and a 1024-point radix-2 FFT feed an iAudio spectrum texture back to the image shader.
A preset browser (Ctrl+L cycle, Ctrl+Shift+L list) over /usr/share/shaders, with bundled image and sound presets.
host/src/input/browser-input-source.ts gains a wheel option, so { pointer: false, wheel: true } lets the editor wheel-scroll while the Modeset pane keeps absolute-pointer ownership.

Demo and tests

The sdl2 kandelo preset replaces the evdev and espeak presets from Evdev on Kandelo - /dev/input keyboard and pointer, plus an espeak-ng speech demo #698.
sdl2_demo example plus an end-to-end vitest; KMSDRM / evdev backend smokes; a native editor unit test (programs/sdl2/test/editor_test.c, driven by host/test/sdl2-editor-unit.test.ts) covering selection, undo/redo, auto-replace and movement including the column-memory case.
packages: libdrm and sdl2-demo classified for the local source build.

Notes

No ABI_VERSION bump. This layer is user-space C plus host TypeScript; abi/snapshot.json regenerates additive-compatible. Toolkit on Kandelo — the C/C++ library stack behind foot, Waybar and mako #1318 up-stack makes the 43 → 44 bump.
Dual-host parity: the host TypeScript lands for both hosts. Sound-shader audio playback is browser-only by design — the headless Node GL stub has no glReadPixels readback — and the chip-synth fallback covers the Node path, so the demo runs and tests green there.
Asyncify-free, per the repo fork-instrumentation policy.
A latent bug was fixed in the cleanup pass: move_vertical set the desired column before move_to_offset, whose horizontal primitives reset it, so vertical column memory never worked. The assignment now runs after the move.

Rebased and rebuilt

This branch sits on explore-dri-evdev-and-alsa at 6c28ee0ec, itself on main at 888e628d5. The history stayed at 6 commits.

Against the previous tip, the only files that differ outside those main itself moved are 16, and every change in them is a comment or a documentation line: plan labels (B1–B5, Phase C, and the playground's Phase 3–Phase 7) were replaced by the fact each label stood for. No code changed.

The earlier description of this PR listed ALSA as the audio backend. That was accurate for a revision of #698 that no longer exists — the stack converged on /dev/dsp, and the rebuilt history never adds an ALSA surface.

Verification

Run at the pre-rebase stack tip ab7ab2b63, which contained all 6 commits here:

cargo test -p kandelo --target aarch64-apple-darwin --lib → 1665 passed, 0 failed
bash scripts/check-abi-version.sh → exit 0
scripts/run-libc-tests.sh → 302 passed, 20 XFAIL, 0 build failures, and 1 unexpected failure: regression/raise-race. That test forks 100 children inside a signal handler; 5 waits return ECHILD, reproducibly. scripts/run-libc-tests.sh:163 skips it whenever CI=true, so it does not gate this PR, and whether it regressed or predates the stack is not yet established.
scripts/run-posix-tests.sh → 174 passed, 0 FAIL, 3 XFAIL, 2 SKIP

Both suites need WASM_POSIX_RESOLUTION_POLICY=source-only-v1 and WASM_POSIX_SOURCE_ONLY_BINARY_ROOT=<repo>/local-binaries/source-only-v1. binaries-abi-v44/index.toml does not exist until the toolkit tier merges, so the binaries/ provenance tier cannot be materialized locally. The full host vitest suite needs both provenance tiers resolvable at once, which only CI's prepared workspace provides — it is the gate for that suite and for the browser demos.

The stack has since been rebased onto main at 888e628d5. Two gates were re-run at the new stack tip ccc7c6214 (explore-dri-omarchy): bash scripts/check-abi-version.sh → exit 0, snapshot in sync and the 43 → 44 bump consistent; scripts/ci-run-test-suite.sh vitest exact-abi-source → 3049 passed, 0 failed, with two files failing to collect on paths #1321 moved to crates/runtime-core/. That collection failure is a main defect and #1337 fixes it.

github-actions · 2026-06-16T17:11:45Z

Phase B-1 matrix build status — `pr-709-staging`

ABI v16. 70 built, 0 failed, 70 total.

Package	Arch	Status	Sha
alsa-lib	wasm32	built	`1caa0b95`
libcurl	wasm32	built	`f458bc9d`
libcxx	wasm32	built	`ff70de04`
libcxx	wasm64	built	`81c2ec20`
libdrm	wasm32	built	`86e06240`
libinput-lite	wasm32	built	`d652e083`
libpng	wasm32	built	`60f4d66b`
libxml2	wasm32	built	`90f0f99a`
libxml2	wasm64	built	`2af7935f`
openssl	wasm32	built	`a24620d0`
openssl	wasm64	built	`cfa4e28e`
sdl2	wasm32	built	`94a0ce9f`
sqlite	wasm32	built	`0e5738b2`
sqlite	wasm64	built	`0e34b4a2`
zlib	wasm32	built	`fe36cb3f`
zlib	wasm64	built	`3f5d181e`
bc	wasm32	built	`a52ea0bc`
bzip2	wasm32	built	`57b54f64`
coreutils	wasm32	built	`d71e7695`
curl	wasm32	built	`0818e0ab`
dash	wasm32	built	`7c056df3`
diffutils	wasm32	built	`5cbf6377`
dinit	wasm32	built	`630dcc01`
fbdoom	wasm32	built	`e5e1c70b`
file	wasm32	built	`dd036c27`
findutils	wasm32	built	`eb035db6`
gawk	wasm32	built	`5487de54`
git	wasm32	built	`5fdb5e16`
grep	wasm32	built	`e32dadf9`
gzip	wasm32	built	`df2c567c`
kandelo-sdk	wasm32	built	`6e2d73a9`
kernel	wasm32	built	`830c3ef8`
less	wasm32	built	`9a60cd57`
lsof	wasm32	built	`85fe5cf2`
m4	wasm32	built	`b714beef`
make	wasm32	built	`e8b2e0c5`
mariadb	wasm32	built	`0c32739e`
mariadb	wasm64	built	`c67da976`
msmtpd	wasm32	built	`bde52af6`
nano	wasm32	built	`328824ad`
ncurses	wasm32	built	`eccd0780`
netcat	wasm32	built	`bb4af825`
nginx	wasm32	built	`1cc7bb16`
php	wasm32	built	`96236a5a`
posix-utils-lite	wasm32	built	`1066f489`
sed	wasm32	built	`4c0c1e20`
spidermonkey	wasm32	built	`94b390a7`
tar	wasm32	built	`a14c8877`
tcl	wasm32	built	`96f124a3`
unzip	wasm32	built	`94f1d123`
userspace	wasm32	built	`dd56ddb4`
vim	wasm32	built	`004fb47b`
wget	wasm32	built	`bfe5d484`
xz	wasm32	built	`e3ff82eb`
zip	wasm32	built	`39a7aee0`
zstd	wasm32	built	`ed37b4a4`
bash	wasm32	built	`0e8d90ce`
mariadb-test	wasm32	built	`67e5f804`
mariadb-vfs	wasm32	built	`72e579b6`
mariadb-vfs	wasm64	built	`38b04794`
nethack	wasm32	built	`98adb3f8`
node	wasm32	built	`54c6bc2d`
spidermonkey-node	wasm32	built	`1b68454d`
vim-browser-bundle	wasm32	built	`c92f8b4f`
nethack-browser-bundle	wasm32	built	`b3833c25`
rootfs	wasm32	built	`e3804f83`
shell	wasm32	built	`010d14a1`
lamp	wasm32	built	`ea14b4c1`
node-vfs	wasm32	built	`2f8273a7`
wordpress	wasm32	built	`e6f3bdb8`

_{Auto-generated; replaced on each push. Raw data in the publish-status workflow artifact.}

…lper + tests, polling-audio cap error, gbm-ring smoke, docs, vitest exnref flag, erlang subprocess flag, llvm-bin fallback Devil's-advocate audit follow-up for PR #709 + the cluster of test-infrastructure fixes needed to get vitest from 151 fail -> 1 fail. Audit fixes (see docs/plans/2026-06-16-dri-kandelo-port-handoff-58-audit.md): - programs/sdl2_demo.c + host/test/sdl2-demo.test.ts: trim WHAT-comments, rename test, drop redundant SDL_PumpAudioDevices extern. - crates/kernel/src/audio/pcm_ioctl.rs: 3 regression tests for cf61010. - host/src/kernel-worker.ts + host/test/ioctl-encoded-marshalling.test.ts: extract computeIoctlEncodedSize helper + 6 boundary tests. - programs/gbm_surface_smoke.c + host/test/gbm-surface-ring.test.ts: lock/release/has_free_buffers smoke + vitest. - packages/registry/sdl2/patches/0002-polling-audio-eagain.patch (rev3 -> rev4): surface SDL_SetError on >8 concurrent opens. - docs/{architecture,posix-status}.md: document ALSA direct path, IoctlEncoded, polling-audio, KMS surface, 2-BO gbm ring, GL cmdbuf. Test-infra fixes (this session): - host/vitest.config.ts: poolOptions.forks.execArgv += [--experimental-wasm-exnref]. Lets cached php/erlang/spidermonkey/ wordpress/mariadb (built with -fwasm-exceptions) compile. - packages/registry/erlang/test/erlang.test.ts: spawn node with the flag rather than 'npx tsx', since execArgv doesn't inherit across process spawn. - host/test/fork-dlopen-replay-e2e.test.ts: discoverLlvmBin() falls back to 'command -v clang' when /opt/homebrew is absent; skipIf when no clang is found at all. Tests: cargo 1072/0 (handoff-58, no kernel changes since), vitest 882/1/19 (was 710/151/41; 1 remaining fail = installs cowsay with npm, pre-existing at tip e6cc2f5 - see docs/plans/2026-06-16-dri-kandelo-port-handoff-59.md section 'The 1 remaining failure'), libc-test/posix-tests GREEN per handoff-58. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…tion, test pacing Three root causes kept PR #709 red; all fixes verified green in CI (run 28977279667). CI fetch gate (staging-build.yml, prepare-merge.yml, run.sh) The WIP espeak-ng package pins a placeholder all-zero source sha and lists gitignored vendored sources (pcaudiolib-src/) as cache-key inputs, so the resolver cannot hash it on a clean checkout and no archive exists in any binaries-abi-v* index. Disable it in the matrix + fetch-gate skip lists and in run.sh's BROWSER_FETCH_SKIP_PKGS (the single choke point every prepare-browser caller shares). Remove the entries once the port lands a real source pin + published binary. Host runtime (host/src/kernel-worker.ts) — the real platform bug The vblank pump broad-woke every blocked syscall 60x/s in every kernel worker, and each wake re-dispatched select/pselect6 from the entry's original args — re-reading the guest time-struct (which the kernel never decrements) and re-arming the FULL timeout. A select-based sleep woken every 16 ms never expires: mariadbd bootstrap parked in exactly such a sleep after "InnoDB started" and test/exec-brk-base.test.ts timed out 3/3. Fixes, both in the Node/browser-shared worker: - POSIX correctness: handleSelect/handlePselect6 thread a deadlineOverride through every re-dispatch path so the timeout stays measured from the original call; expired re-dispatches complete with 0. - Wake scoping: tickVblank gates the broad wake on real page-flip activity (kernel_kms_commit_count changed since the previous tick, one-tick carry for the submit/drain race). No DRM client → no wakes; a flipping client keeps its 60 fps cadence. The dri-kms-stats-sab tests' throwing sentinel moves from kernel_kms_commit_count (now read every tick by the gate) to kernel_kms_last_frame_us, which only the stats branch reads — the guarded invariant (never touch slots 5/6 of a small SAB) is unchanged. Tests (host/test/sdl2.test.ts, apps/browser-demos/test/kandelo-sdl2.spec.ts) Both paced assertions by wall clock instead of app-observable events, so slow CI runners failed them while fast machines passed: - vitest: wait for the "sdl2: editor recompile" breadcrumb after typing instead of sleeping 300 ms — on a slow runner all queued keys drained in one SDL_PollEvent burst and F2 correctly cancelled the still-pending debounce. - playwright: wait for the app's first presented frame (the Modeset pane unmounts its "waiting for PAGE_FLIP" placeholder on the first flip) before sampling animation, and sample until PNG byte-size variance appears (30 s deadline) instead of a fixed 6x500 ms schedule — under CI's software GL, init outlasted the old sampling window so every screenshot captured the static placeholder. Validated: exec-brk-base 3/3 on a fresh datadir (was 3/3 25 s timeouts; bisected to the Phase 0 vblank-wake commit and confirmed by kernel/host cross-swaps against main), sdl2 vitest + playwright pass locally, host typecheck clean, full host vitest green except pre-existing local-only wasm64 fixture gaps, and the full staging CI run green on this tree. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

…tion, test pacing Three root causes kept PR #709 red; all fixes verified green in CI (run 28977279667). CI fetch gate (staging-build.yml, prepare-merge.yml, run.sh) The WIP espeak-ng package pins a placeholder all-zero source sha and lists gitignored vendored sources (pcaudiolib-src/) as cache-key inputs, so the resolver cannot hash it on a clean checkout and no archive exists in any binaries-abi-v* index. Disable it in the matrix + fetch-gate skip lists and in run.sh's BROWSER_FETCH_SKIP_PKGS (the single choke point every prepare-browser caller shares). Remove the entries once the port lands a real source pin + published binary. Host runtime (host/src/kernel-worker.ts) — the real platform bug The vblank pump broad-woke every blocked syscall 60x/s in every kernel worker, and each wake re-dispatched select/pselect6 from the entry's original args — re-reading the guest time-struct (which the kernel never decrements) and re-arming the FULL timeout. A select-based sleep woken every 16 ms never expires: mariadbd bootstrap parked in exactly such a sleep after "InnoDB started" and test/exec-brk-base.test.ts timed out 3/3. Fixes, both in the Node/browser-shared worker: - POSIX correctness: handleSelect/handlePselect6 thread a deadlineOverride through every re-dispatch path so the timeout stays measured from the original call; expired re-dispatches complete with 0. - Wake scoping: tickVblank gates the broad wake on real page-flip activity (kernel_kms_commit_count changed since the previous tick, one-tick carry for the submit/drain race). No DRM client → no wakes; a flipping client keeps its 60 fps cadence. The dri-kms-stats-sab tests' throwing sentinel moves from kernel_kms_commit_count (now read every tick by the gate) to kernel_kms_last_frame_us, which only the stats branch reads — the guarded invariant (never touch slots 5/6 of a small SAB) is unchanged. Tests (host/test/sdl2.test.ts, apps/browser-demos/test/kandelo-sdl2.spec.ts, apps/browser-demos/test/kandelo-merge-gate.spec.ts) All paced assertions by wall clock instead of app-observable events, so slow CI runners failed them while fast machines passed: - vitest: wait for the "sdl2: editor recompile" breadcrumb after typing instead of sleeping 300 ms — on a slow runner all queued keys drained in one SDL_PollEvent burst and F2 correctly cancelled the still-pending debounce. - playwright: wait for the app's first presented frame (the Modeset pane unmounts its "waiting for PAGE_FLIP" placeholder on the first flip) before sampling animation, and sample until PNG byte-size variance appears (30 s deadline) instead of a fixed 6x500 ms schedule — under CI's software GL, init outlasted the old sampling window so every screenshot captured the static placeholder. - merge-gate shell demo: wait for bash's first kandelo$ prompt before typing, like the spec's other terminal tests already do. The test typed as soon as the page said "Ready", so on a slow runner the command landed before the shell was reading and was discarded by the boot chain's startup typeahead flush (tcflush) — the CI trace shows the echoed command followed by two bare prompts and an unchanged PS1. Input typed before the first prompt is legitimately flushable on any POSIX terminal, so the prompt is the only correct readiness signal. Validated: exec-brk-base 3/3 on a fresh datadir (was 3/3 25 s timeouts; bisected to the Phase 0 vblank-wake commit and confirmed by kernel/host cross-swaps against main), sdl2 vitest + playwright pass locally, host typecheck clean, full host vitest green except pre-existing local-only wasm64 fixture gaps, and the full staging CI run green on this tree. Merge-gate shell demo test passes 3/3 locally against fetched prebuilt assets after the prompt-wait fix. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

The stub was a hand-maintained subset: every ioctl an SDL2 or Wayland client might reach for had to be added to it by hand, and its headers were a vendored copy of the kernel's UAPI that could drift from what the kernel actually answers. libdrm now builds from upstream sources as a package, KMS subset only. That inverts the test: instead of a stub written to satisfy our clients, unmodified upstream code drives the kernel's `DRM_IOCTL_MODE_*` surface, and anything missing fails loudly at the real call site. `build-musl.sh` resolves the package into the sysroot beside the remaining GBM/EGL/GLES shims rather than compiling a stub, so `wasm32posix-pkg-config` answers the same either way and consumers such as `modeset` need no change beyond a rebuild.

`gbm_surface_lock_front_buffer` and `gbm_surface_release_buffer` are how a KMSDRM client hands a rendered frame to the display and gets a free one back. The shim had no surface concept at all, so SDL2's KMSDRM backend had nothing to page-flip between. The ring holds the buffers a surface owns and tracks which one the display currently holds, which is what lets a client render into one while the other is on screen. `gbm_surface_smoke` drives lock/release across a full cycle so a regression shows up as a specific failed transition rather than a black window.

SDL2 was built with the dummy video driver, so a program could link it but never present a frame. Enabling KMSDRM points it at `/dev/dri/card0` through libdrm and libgbm, and enabling the evdev input backend points it at `/dev/input/event{0,1}`. SDL scans devices by their advertised capability bits, so the vendored code table grew the entries its probe expects to find. The two smokes pin each half separately: one asserts a mode is set and a frame reaches the display, the other that a synthesised keystroke arrives as an SDL_KEYDOWN — so a break tells you which backend regressed.

The playground is the first program that uses all three backends at once: a gap-buffer editor on the left, the user's fragment shader running on `/dev/dri/card0` on the right, and audio through `/dev/dsp`. Typing recompiles after 250 ms of idle; a failed compile keeps the last good program and shows the offending line, so a typo never blanks the screen. Sound is generated the same way the image is. A second shader renders PCM, and its output feeds the audio device — the chip synth is the fallback for when GL cannot render (the headless harness reads back zero frames). Text comes from a font atlas baked once at startup with stb_truetype, which is why the TTF is vendored and its generated C header is ignored rather than committed: it is reproduced from the TTF on every build.

The playground exercises everything the evdev input-log demo showed — keystrokes and pointer motion through `/dev/input/event{0,1}` — inside a product surface rather than a proof of life, so both earlier presets give way to it. espeak-ng stays a registry package with its published voice data; only its preset and spec go, and `evdev_demo` goes with the demo it existed to back. The two input paths are split on purpose. `BrowserInputSource` gains options that turn its pointer feed off, because the Modeset pane owns the pointer: it injects framebuffer-absolute coordinates through `KmsDisplayHandle.sendPointerAbs`. That call has to emulate an absolute device over relative deltas — SDL sees `REL_X`/`REL_Y` advertised on event1, classifies it as a relative mouse and ignores `EV_ABS` — so each update slams the cursor into a corner with an oversized delta and then walks it to the target.

The local source-build set names every package it builds and asserts the exact roster, so two new packages have to be placed: `libdrm` is platform — a library the sysroot exposes to any client — and `sdl2-demo` is user-software. Dropping `evdev-demo` leaves the same assertion, which is why the counts move in one place.

mho22 · 2026-08-28T20:19:20Z

Folded into #948.

The stack was rebased onto 888e628d5. The rebase collapsed this branch and #698 into explore-dri-wayland, which now sits directly on main and carries all 28 commits. The six commits of this PR are in #948, unchanged in content:

feat(dri): upstream libdrm 2.4.120 replaces the hand-written stub
feat(dri): a surface ring in libgbm for SDL2 to scan out through
feat(sdl2): the KMSDRM video and evdev input backends
demo(sdl2): a GLSL playground on KMSDRM video, dsp audio and evdev input
demo(browser): the sdl2 preset replaces the evdev and espeak demos
fix(packages): classify libdrm and sdl2-demo for the local source build

Review the work in #948. The pre-rebase tip d1949928d is kept locally as refs/backup/2026-08-28-preforcepush/sdl2.

mho22 · 2026-08-28T20:48:44Z

Retracting the comment above. This PR was closed in error and is reopened.

The stack was rebased onto 888e628d5. During that rebase this branch was temporarily concatenated into explore-dri-wayland, and I read the concatenation as a merge of scope. It is not: SDL2 on KMSDRM is its own subsystem, and #948 depends on it rather than containing it.

The branch is force-pushed to 42e3a4abc, the same 6 commits rebased onto 888e628d5, subject-for-subject identical to the pre-rebase tip d1949928d. The base stays explore-dri-evdev-and-alsa. The six-PR stack is intact:

main -> #698 -> #709 -> #948 -> #976 -> #1318 -> #1241

The pre-rebase tip is kept locally as refs/backup/2026-08-28-preforcepush/sdl2.

mho22 changed the title ~~[explore-dri-sdl2] sysroot(sdl2): full Phase A–C — backends + GL stubs + sdl2_demo~~ [explore-dri-sdl2] SDL2 on Kandelo — sysroot port + GLSL shader playground demo Jun 22, 2026

mho22 changed the title ~~[explore-dri-sdl2] SDL2 on Kandelo — sysroot port + GLSL shader playground demo~~ SDL2 on Kandelo - GLSL shader playground demo Jun 22, 2026

mho22 force-pushed the explore-dri-sdl2 branch from 94773e5 to b84c8db Compare July 9, 2026 08:05

mho22 force-pushed the explore-dri-sdl2 branch from b84c8db to d2f79e1 Compare July 9, 2026 11:38

mho22 marked this pull request as ready for review July 9, 2026 15:46

brandonpayton mentioned this pull request Jul 13, 2026

Batch PHP support and bundle ABI updates with their runtime changes #907

Merged

13 tasks

mho22 force-pushed the explore-dri-evdev-and-alsa branch from d1b1156 to e439b86 Compare August 21, 2026 10:02

mho22 force-pushed the explore-dri-sdl2 branch from d2f79e1 to a97a334 Compare August 21, 2026 10:02

mho22 mentioned this pull request Aug 21, 2026

Evdev on Kandelo - /dev/input keyboard and pointer, plus an espeak-ng speech demo #698

Open

mho22 force-pushed the explore-dri-evdev-and-alsa branch from e439b86 to f498fc4 Compare August 24, 2026 19:08

mho22 force-pushed the explore-dri-sdl2 branch from a97a334 to 52ad9c6 Compare August 24, 2026 19:08

mho22 force-pushed the explore-dri-evdev-and-alsa branch from f498fc4 to 71654e9 Compare August 26, 2026 14:36

mho22 force-pushed the explore-dri-sdl2 branch from 52ad9c6 to d194992 Compare August 26, 2026 14:37

mho22 added 6 commits August 28, 2026 11:15

mho22 closed this Aug 28, 2026

mho22 reopened this Aug 28, 2026

mho22 force-pushed the explore-dri-evdev-and-alsa branch from 71654e9 to 6c28ee0 Compare August 28, 2026 20:47

mho22 force-pushed the explore-dri-sdl2 branch from d194992 to 42e3a4a Compare August 28, 2026 20:47

Navigation Menu

Uh oh!

SDL2 on Kandelo - GLSL shader playground demo - #709

SDL2 on Kandelo - GLSL shader playground demo#709
mho22 wants to merge 6 commits into
explore-dri-evdev-and-alsafrom
explore-dri-sdl2

mho22 commented Jun 16, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Jun 16, 2026 •

edited

Loading

Uh oh!

mho22 commented Aug 28, 2026

Uh oh!

mho22 commented Aug 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

mho22 commented Jun 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Why

What changed

Notes

Rebased and rebuilt

Verification

Uh oh!

github-actions Bot commented Jun 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Phase B-1 matrix build status — pr-709-staging

Uh oh!

mho22 commented Aug 28, 2026

Uh oh!

mho22 commented Aug 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

mho22 commented Jun 16, 2026 •

edited

Loading

github-actions Bot commented Jun 16, 2026 •

edited

Loading

Phase B-1 matrix build status — `pr-709-staging`