# i915n display/hotplug tester kit — 2026-10-18

Download directory:
https://dlc.openindiana.org/isos/hipster/test/i915n-20261018/

For Olaf's T470 trial, record the actual Intel PCI ID and external connector/
dock first. Use a disposable boot environment and retain the working BE as
the permanent default. Please return the package plan if it does not resolve.

| Archive | Contents / intended use |
| --- | --- |
| `i915n-display-hotplug-20261018.p5p` | The exact 41 graphics packages used by the verified Live image and passing openQA job61: final i915n/DRM, sysevent bridge, clean Xorg21.1.24 with polling, ABI-matched Xorg drivers, libdrm2.4.134 and coexistence packages |
| `i915n-drivers-20261018.p5p` | Seven matching new DRM packages, including the final hotplug producer; for an existing compatible Xorg stack. Automatic Xorg notification needs the consumer in the full archive |
| `mate-desktop-bbb5f6a.p5p` | Optional separate MATE1.28.2 revision2 trial: removes the old two-second cached RandR update throttle. Native build and virtual-display regression passed; not installed in the X230 trial or the passing Live image |

Use the full archive when testing automatic display notification with the
matching Xorg stack. The driver-only archive is a subset, not an additional
version to install over the full archive. `graphics-fmris.txt` and
`driver-fmris.txt` pin the exact versions/timestamps. All exported graphics
packages use publisher `openindiana.org`, matching the tested image; file
payloads and package timestamps are preserved. These are test archives, not
a change to the public package publisher.

`SHA256SUMS` covers the files in this directory except the outer kit tarball.
The tarball has its own `.sha256sum` sidecar. Verify downloaded hashes before
installation, e.g. `digest -a sha256 <archive>` on OpenIndiana.

## Enablement and evidence

The i915n package enables the full implemented display paths by default for
the PCH Gen6–Gen9.5 hardware admitted by this port. It ships 183 Intel aliases
(`i915n.aliases`) and SKL1.27/KBL1.04 DMC firmware with the upstream Intel license.
It is not an Ivy-Bridge-only display pilot. Unsupported SoCs, Gen10+ and discrete
GPUs remain outside this port's implemented admission; this kit does not claim
support for them. No further feature-disable defaults were added for testing.

Evidence:

* Native driver/Xorg builds, clean Xorg module rebuild, packaging and exact
  image payload/version verification passed.
* The Live image at SHA256
  `ce865e9d9ff927e900d4290f82c3ffdcbc4a4dd11c07369b42a8ea58ad6e2123`
  passed openQA job61 (`boot_live`, `package_kernel`), including MATE and final
  module/firmware hashes. This is guest/package coverage.
* X230 Ivy Bridge: internal panel, VT recovery and final-module operation;
  Acer/VGA at kernel boot; repeated unplug/replug; operator reports the terminal
  moved to the internal panel on unplug and back to the Acer on reconnect.
* That later physical action was outside the bounded trace. Exact event count,
  per-path attribution and framebuffer shrink dimensions were not captured.
* Skylake/T470, mini-DP, dock/MST and other generations still need hardware
  testing. Suspend/resume is not implemented in current i915n DDI glue.

The requested presentation chooser and remembered EDID layout UI are not
implemented in this kit. Existing MATE handles the observed window movement.
The optional MATE library package fixes a separate confirmed notification-loss
bug; it is not credited for the successful X230 trial.

## Installation and recovery

Follow `TESTING.md`. `plan-install.sh` only prints/runs a package dry run in an
already mounted inactive test BE. Do not bypass incorporations or replace core
OS libraries to force a plan. No private host logs or credentials are included.

## Source patches

`patches/README.md` records exact bases, tips and application commands. The
three mbox series were reapplied to their stated bases and their resulting
trees matched the source tips. **Use `git am --keep-cr`** so the firmware
license retains its original CRLF bytes. Native build inputs predate later
test/documentation-only commits; see `SOURCE-REVISIONS.txt`.

Please collect PCI ID, connector/cable/dock, `drmadm status`, loaded modules,
Xorg.0.log, messages around cable actions, `xrandr --current --verbose` before
any forced query, and the visual result. A working fallback desktop alone does
not identify which driver or renderer was used.