Add netsurf package#243
Conversation
|
Your test plan mentions not testing on the rMPP, does this mean you tested on a rMPPM or rMPPure? If so, why is this marked as armv7 only? |
|
@Eeems I wrote |
|
It's already marked as armv7 only. |
|
I know. And I think it is proper to be marked as armv7 only at this stage. |
|
I'm confused now, I asked why it was marked as armv7 only if your test plan seems to indicate that you want the aarch64 devices to be tested. You then asked if it should be marked as armv7 only, to which I replied that it already was. You are now saying you know it was and you think it should only be armv7 only at this stage. So are you wanting aarch64 (rMPP, rMPPM, and rMPPure) testing to be done or not? |
|
I want this to be tested on aarch64 devices, but I didn't understand how I should mark it. I think to express the interest, I need to mark it differently and include both armv7 and aarach64. Currently there are gaps in cross-building on |
What gaps are there currently? If it's not ready for building on aarch64, that is a different matter that doesn't need to hold up this PR, and can be resolved in a future PR. |
|
Like you pointed out, the gap was that |
|
It looks like you did a bad merge and now there are 267 different commits, and unresolved merge conflicts in the tree. |
Ports the MaicroNotHard/netsurf-reMarkable framebuffer browser (previously a Toltec package, originally by alex0809, now archived) to a Vellum VELBUILD recipe, running through appload + qtfb-shim like tilem/yaft. Bundles libevdev statically (matching the other VELBUILD dependencies), fetches DejaVu fonts and Noto Sans CJK SC for CJK fallback rendering, and pulls the netsurf-reMarkable source from upstream at build time (no Vellum font package exists yet; keeps binary blobs out of git history). Reuses the OS-provided OpenSSL 3.x/curl rather than cross-building static OpenSSL 1.1.1k/curl 7.75.0, since the reMarkable OS already ships matching libssl.so.3/libcrypto.so.3/libcurl.so.4. QTFB_SHIM_INPUT_MODE is set to NATIVE so the shim passes real evdev devices through unmodified, matching what netsurf's own libevdev-capability-based device scanner expects (and, separately, ensures /sys/devices/soc0/machine reflects the true hardware model rather than a virtualized one, which matters for netsurf's own native rM1/rM2 detection and X-axis touch handling).
…pport) Bumps the pinned netsurf-reMarkable commit to pick up a rebase onto latest upstream NetSurf plus WebP and JPEG-XL image decoder support, and drops the now-unneeded v0.5.0-wip branch reference from CI. All curl fetches in the upstream build scripts now retry on transient network failures (observed once on freedesktop.org's libevdev mirror). Enables aarch64 (RMPP/RMPPM/RMPPure): netsurf-reMarkable's input layer now recognizes the RMPP machine names and treats them like RM2 (no x-axis invert). Verified armv7 and aarch64 both build a working nsfb binary end-to-end; the touch-input mapping itself is unverified on real RMPP hardware - please report back if it's wrong. license is GPL-2.0-only, not GPL-2.0-or-later: upstream's COPYING grants "version 2 of the License" with no or-later clause, and netsurf statically links that GPL-2.0-only core. GPL-2.0-only also matches the repo's own precedent for v2-only packages (puzzles, vellum-bash-completion, netevent).
|
Yeah, I did a bad merge... The latest forch-push should be good and fixing things.
|
|
/publish-testing |
|
Packages published to testing repo:
Install with: |
There was a problem hiding this comment.
- The netsurf icon does not look good on any of the devices (rM1, rM2, rMPP, and rMPPM), it can be hard to read, is there an alternative that can be use?
- When opening on the rMPPM nothing displays until you swipe down from the top
- Pen/touch input do not work on the rMPP and rMPPM appears to have x and y inverted, either set QTFB_SHIM_MODEL to RM1 for all devices, or properly implement input handling for the other devices. The rMPPure should have the same issue.
- The onscreen keyboard fills the empty spaces where there is no key with black instead of white
Supporting changes are merged into the libnsfb-reMarkable, netsurf-base-reMarkable and netsurf-reMarkable; the package pins their main commits. Verified on rM2 and built green for armv7+aarch64 in my fork's CI. |
|
@MaicroNotHard why are both of your netsurf-reMarkable and netsurf-base-reMarkable forks wildly out of sync with the upstream? I would only expect to see it saying you are ahead commits, not behind by almost the same number. |
Review fixes for the Paper Pro (and general polish): - Replace the wide wordmark launcher icon with a square, high-contrast NetSurf globe that stays legible on all devices. - Switch the qtfb-shim to RM1 emulation (QTFB_SHIM_MODEL=RM1) so input and framebuffer go through the same translation layer used by every other appload app. This is viable now that libnsfb no longer hard-requires the RM1-only physical-button device. - Bump netsurf-reMarkable to pull the libnsfb fixes (optional gpio device + full-screen refresh on first frame) and the white on-screen keyboard.
|
Cleaned up the fork hygiene so each one is ahead-only:
The package now pins those clean |
|
/publish-testing |
|
Packages published to testing repo:
Install with: |
2 participants
Summary
(a fork of alex0809's original package, currently archived). The base repo is up-to-date with the official Netsurf upstream.
Test plan
LLM Acknowledgement: Claude helped with this PR.