Files
hermes-agent/.github/workflows/install-e2e-run.yml
T
ethernet cd667debfa fix(install-e2e): windows transcripts were empty; player gets #zip= hash + one player per run
Windows transcripts were ZERO bytes: ts-prefix.ps1 formatted with
{0:D2}, but Floor() returns a double and the D specifier is
integer-only - it threw per line, and under the driver's relaxed EAP
every line errored into the void. {0:00} fixes it (custom numeric
format works on doubles). Reproduced the exact pipeline locally
(empty file + Format specifier invalid), verified the fix produces
prefixed merged stdout+stderr with exit code intact. That is also
why the log timeline never auto-synced: there was nothing in the
files to sync.

The GitHub artifact URL 307s to /suites/... server-side and strips
the ?zip= query param. The player now reads the zip URL from a
#zip= HASH param (client-side, survives the redirect) with ?zip=
as fallback; the hash path was verified in a real browser against a
real leg zip (auto-fetch + boot).

Per ethie's design, one player artifact for the whole run: new
leg-player job uploads playback.html (archive:false) before the
matrix legs, the report job needs it, and each ran cell gets TWO
links - 📼 to the player with #zip=<that leg's logs zip> and ⬇️ to
the raw zip. Per-leg player uploads removed from all three run
workflows.
2026-08-13 22:51:44 -04:00

160 lines
7.0 KiB
YAML

name: Install & Update E2E (reusable)
# Runs ONE {install-method, update-method} combination against ONE starting
# commit, with a real install (uv, a managed Python, Node, the venv) behind
# it.
#
# Reusable so callers can fan out over the combinations that matter --
# update from the tip vs. from an older release, `hermes update` vs.
# re-running the installer -- without duplicating the runner setup. Each leg
# is independent: its own isolated HOME, its own install, nothing rewound
# or shared.
#
# No sandbox: tests/install/installer-script-e2e.sh points every git
# process at a local bare clone (url.<file://serve.git>.insteadOf in a
# driver-owned GIT_CONFIG_GLOBAL) and isolates HOME, so the installer and
# updater run byte-for-byte against their real URLs on the bare runner --
# which is disposable, and therefore IS the sandbox. That also makes this
# workflow OS-agnostic: the same driver runs on ubuntu and macos runners.
#
# Method ids come from scripts/sandbox/generate-e2e-matrix.mjs. Supported
# today: install via installer-script, update via hermes-update or
# installer-script (re-run the one-liner).
# Anything else NATIVELY SKIPS (grey check, no runner): capability
# knowledge lives here, next to the driver, so the caller can dispatch
# every declared combination without knowing which ones work.
#
# Call it:
#
# jobs:
# tip:
# uses: ./.github/workflows/install-e2e-run.yml
# with:
# install-method: installer-script
# update-method: hermes-update
# install-ref: refs/heads/main
on:
workflow_call:
inputs:
install-method:
description: 'How the starting version gets installed. Supported: installer-script (the real curl | install.sh one-liner) and installer-script+desktop (the same one-liner with --include-desktop).'
required: true
type: string
update-method:
description: 'How the install updates to HEAD. Supported: hermes-update (the updater), installer-script (re-run the one-liner), installer-script+desktop (re-run with --include-desktop), hermes-desktop-app-update (launch via hermes desktop under Playwright, click Update now). Declared-but-TODO methods (open-app-update) skip.'
required: true
type: string
install-ref:
description: 'What to install before updating: a branch, a tag (v2026.7.7), or a SHA reachable from main.'
required: false
type: string
default: refs/heads/main
leg-id:
description: 'Artifact-safe matrix leg id (from generate-e2e-matrix.mjs legId). Names this leg''s logs + player artifacts so the report job can link a row to its zip.'
required: true
type: string
tag-has-desktop:
description: "Whether install-ref ships the desktop app (apps/desktop). The caller annotates this from the tag's own tree; desktop-method legs from pre-desktop releases natively skip."
required: false
type: boolean
default: true
runner:
description: 'Runner label.'
required: false
type: string
default: ubuntu-latest
timeout-minutes:
description: 'Job timeout. A cold run installs real toolchains twice, and app-update legs add a full Electron build + launch.'
required: false
type: number
default: 75
permissions:
contents: read
jobs:
e2e:
name: install & update
# The pairs the driver can run today; anything else is a declared TODO
# and natively skips. Desktop-surface methods (+desktop installs,
# hermes-desktop-app-update) also need the starting tag to ship
# apps/desktop (their flags shipped with it).
if: >-
(inputs.install-method == 'installer-script'
|| (inputs.install-method == 'installer-script+desktop' && inputs.tag-has-desktop))
&& (contains(fromJSON('["hermes-update", "installer-script"]'), inputs.update-method)
|| (contains(fromJSON('["installer-script+desktop", "hermes-desktop-app-update"]'), inputs.update-method) && inputs.tag-has-desktop))
runs-on: ${{ inputs.runner }}
timeout-minutes: ${{ inputs.timeout-minutes }}
steps:
# Full history: the driver bare-clones this checkout as the repo the
# installer/updater talk to, and both OLD and HEAD must be reachable
# in that clone. A shallow clone cannot serve either need.
- uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
with:
fetch-depth: 0
# One recording mechanism on every OS (Xvfb gives headless linux a
# display; the same display serves any app the driver launches).
- name: Start screen recording
uses: ./.github/actions/e2e-screen-record
with:
mode: start
output: ${{ runner.temp }}/e2e-logs/recording.mkv
- name: Run install + update E2E
run: |
set -euo pipefail
tests/install/installer-script-e2e.sh \
--install-method '${{ inputs.install-method }}' \
--update-method '${{ inputs.update-method }}' \
--install-ref '${{ inputs.install-ref }}'
env:
# Outside the workspace on purpose: logs written into the repo
# would trip the driver's own dirty-tree guard.
HERMES_E2E_LOG_DIR: ${{ runner.temp }}/e2e-logs
- name: Stop screen recording
if: always()
uses: ./.github/actions/e2e-screen-record
with:
mode: stop
output: ${{ runner.temp }}/e2e-logs/recording.mkv
# Browsers cannot play Matroska: remux (copy codec, no re-encode) so
# the artifact zip feeds the static playback.html leg player directly.
- name: Remux recording for browser playback
if: always()
run: |
set -euo pipefail
if [ -f "${{ runner.temp }}/e2e-logs/recording.mkv" ]; then
ffmpeg -y -hide_banner -loglevel error -i "${{ runner.temp }}/e2e-logs/recording.mkv" \
-c copy "${{ runner.temp }}/e2e-logs/recording.mp4"
fi
# The leg player: ONE static HTML per run, uploaded up front by the
# leg-player job in install-e2e.yml (archive: false, so GitHub names
# the artifact after the file: playback.html). The report job links
# every ran leg to it with the leg's zip as a #zip= hash param.
- name: Build artifact name
if: always()
id: artifact
run: |
set -euo pipefail
echo "name=install-e2e-logs-${{ inputs.leg-id }}" >> "$GITHUB_OUTPUT"
# The installer's own transcripts say far more than the assertion that
# tripped when a real install breaks.
- name: Upload installer logs
if: always()
uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
with:
# Unique per leg: a matrix over releases runs this workflow several
# times per route, and same-named artifacts collide.
name: ${{ steps.artifact.outputs.name }}-${{ github.sha }}
path: ${{ runner.temp }}/e2e-logs
retention-days: 14
if-no-files-found: ignore