Skip to content

Commit 96b8bbb

Browse files
fredespifredespi
committed
ci(e2e): support manual dispatch and bootstrap the Windows runner
Add a workflow_dispatch trigger with platform and tier inputs so the E2E jobs can be run on demand against this ref (dispatch skips build-and-test for a fast loop; each E2E job builds its own binary via `cargo xtask e2e`, so the skip is safe). Every E2E job's guard is extended to honor the inputs on dispatch while preserving the exact push/PR behavior otherwise, and the consolidated report runs on dispatch too. Also bootstrap Rust on the Strix Halo Windows runner with a PowerShell step: setup-rust-toolchain runs an internal bash script the runner lacks, so install rustup via win.rustup.rs (--default-toolchain none; rust-toolchain.toml pins the version). Idempotent, so it no-ops once the runner is provisioned. Signed-off-by: fredespi <fredrik.espinoza@gmail.com>
1 parent f2ee84e commit 96b8bbb

1 file changed

Lines changed: 151 additions & 11 deletions

File tree

‎.github/workflows/ci.yml‎

Lines changed: 151 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -10,6 +10,24 @@ on:
1010
merge_group:
1111
branches: [main]
1212
types: [checks_requested]
13+
# Manual trigger for on-demand E2E runs on the self-hosted GPU runners, without
14+
# waiting on a push/PR or the full gate. Dispatch against any ref
15+
# (`gh workflow run ci.yml --ref <branch> -f platform=... -f tier=...`) — GitHub
16+
# runs THAT ref's copy of this file, so the E2E jobs it selects live on the ref.
17+
# A dispatch skips build-and-test (see its `if:`) for a fast loop. This trigger
18+
# must exist on the default branch for the workflow to be dispatchable at all.
19+
workflow_dispatch:
20+
inputs:
21+
platform:
22+
description: Which runner(s) to target
23+
type: choice
24+
default: all
25+
options: [all, mock, app-dev-gpu, strix-ubuntu, strix-windows]
26+
tier:
27+
description: Which scenario tier(s) to run
28+
type: choice
29+
default: both
30+
options: [both, expect-pass, known-bugs]
1331

1432
concurrency:
1533
group: ${{ github.workflow }}-${{ github.ref }}
@@ -109,6 +127,9 @@ jobs:
109127
runs-on: ubuntu-latest
110128
# Don't spend per-OS build/test cycles unless the cheap lint gate is green.
111129
needs: [changes, clippy, prek]
130+
# A manual E2E dispatch skips this heavy job for a fast loop; the E2E jobs
131+
# tolerate a skipped build-and-test in their own `if:` guards.
132+
if: github.event_name != 'workflow_dispatch'
112133
steps:
113134
- uses: actions/checkout@v6
114135

@@ -541,7 +562,20 @@ jobs:
541562
name: E2E tests
542563
runs-on: ubuntu-latest
543564
needs: [changes, build-and-test]
544-
if: needs.changes.outputs.heavy == 'true'
565+
# On push/PR/merge_group: unchanged — run when build-and-test succeeded and
566+
# the change is heavy. On workflow_dispatch: build-and-test is skipped, so
567+
# tolerate that and gate on the platform/tier inputs instead (mock tier here).
568+
if: >-
569+
always()
570+
&& needs.changes.result == 'success'
571+
&& (
572+
(github.event_name != 'workflow_dispatch'
573+
&& needs.build-and-test.result == 'success'
574+
&& needs.changes.outputs.heavy == 'true')
575+
|| (github.event_name == 'workflow_dispatch'
576+
&& (inputs.platform == 'all' || inputs.platform == 'mock')
577+
&& (inputs.tier == 'both' || inputs.tier == 'expect-pass'))
578+
)
545579
steps:
546580
- uses: actions/checkout@v6
547581

@@ -576,7 +610,18 @@ jobs:
576610
name: E2E tests (known bugs)
577611
runs-on: ubuntu-latest
578612
needs: [changes, build-and-test]
579-
if: needs.changes.outputs.heavy == 'true'
613+
# See `e2e`: dispatch tolerates skipped build-and-test; mock tier, known-bugs.
614+
if: >-
615+
always()
616+
&& needs.changes.result == 'success'
617+
&& (
618+
(github.event_name != 'workflow_dispatch'
619+
&& needs.build-and-test.result == 'success'
620+
&& needs.changes.outputs.heavy == 'true')
621+
|| (github.event_name == 'workflow_dispatch'
622+
&& (inputs.platform == 'all' || inputs.platform == 'mock')
623+
&& (inputs.tier == 'both' || inputs.tier == 'known-bugs'))
624+
)
580625
steps:
581626
- uses: actions/checkout@v6
582627

@@ -603,7 +648,18 @@ jobs:
603648
name: E2E tests (GPU)
604649
runs-on: [self-hosted, linux, amd-gpu]
605650
needs: [changes, build-and-test]
606-
if: needs.changes.outputs.heavy == 'true'
651+
# See `e2e`: dispatch tolerates skipped build-and-test; app-dev-gpu, expect-pass.
652+
if: >-
653+
always()
654+
&& needs.changes.result == 'success'
655+
&& (
656+
(github.event_name != 'workflow_dispatch'
657+
&& needs.build-and-test.result == 'success'
658+
&& needs.changes.outputs.heavy == 'true')
659+
|| (github.event_name == 'workflow_dispatch'
660+
&& (inputs.platform == 'all' || inputs.platform == 'app-dev-gpu')
661+
&& (inputs.tier == 'both' || inputs.tier == 'expect-pass'))
662+
)
607663
continue-on-error: true
608664
steps:
609665
- uses: actions/checkout@v6
@@ -629,7 +685,18 @@ jobs:
629685
name: E2E tests (GPU, known bugs)
630686
runs-on: [self-hosted, linux, amd-gpu]
631687
needs: [changes, build-and-test]
632-
if: needs.changes.outputs.heavy == 'true'
688+
# See `e2e`: dispatch tolerates skipped build-and-test; app-dev-gpu, known-bugs.
689+
if: >-
690+
always()
691+
&& needs.changes.result == 'success'
692+
&& (
693+
(github.event_name != 'workflow_dispatch'
694+
&& needs.build-and-test.result == 'success'
695+
&& needs.changes.outputs.heavy == 'true')
696+
|| (github.event_name == 'workflow_dispatch'
697+
&& (inputs.platform == 'all' || inputs.platform == 'app-dev-gpu')
698+
&& (inputs.tier == 'both' || inputs.tier == 'known-bugs'))
699+
)
633700
continue-on-error: true
634701
steps:
635702
- uses: actions/checkout@v6
@@ -654,7 +721,18 @@ jobs:
654721
name: E2E tests (Strix Halo, Ubuntu)
655722
runs-on: [self-hosted, linux, strix-halo]
656723
needs: [changes, build-and-test]
657-
if: needs.changes.outputs.heavy == 'true'
724+
# See `e2e`: dispatch tolerates skipped build-and-test; strix-ubuntu, expect-pass.
725+
if: >-
726+
always()
727+
&& needs.changes.result == 'success'
728+
&& (
729+
(github.event_name != 'workflow_dispatch'
730+
&& needs.build-and-test.result == 'success'
731+
&& needs.changes.outputs.heavy == 'true')
732+
|| (github.event_name == 'workflow_dispatch'
733+
&& (inputs.platform == 'all' || inputs.platform == 'strix-ubuntu')
734+
&& (inputs.tier == 'both' || inputs.tier == 'expect-pass'))
735+
)
658736
continue-on-error: true
659737
# The runner's root disk is full and non-persistent; keep the toolchain
660738
# download, install, and temp files on the persistent runner directory so
@@ -685,7 +763,18 @@ jobs:
685763
name: E2E tests (Strix Halo, Ubuntu, known bugs)
686764
runs-on: [self-hosted, linux, strix-halo]
687765
needs: [changes, build-and-test]
688-
if: needs.changes.outputs.heavy == 'true'
766+
# See `e2e`: dispatch tolerates skipped build-and-test; strix-ubuntu, known-bugs.
767+
if: >-
768+
always()
769+
&& needs.changes.result == 'success'
770+
&& (
771+
(github.event_name != 'workflow_dispatch'
772+
&& needs.build-and-test.result == 'success'
773+
&& needs.changes.outputs.heavy == 'true')
774+
|| (github.event_name == 'workflow_dispatch'
775+
&& (inputs.platform == 'all' || inputs.platform == 'strix-ubuntu')
776+
&& (inputs.tier == 'both' || inputs.tier == 'known-bugs'))
777+
)
689778
continue-on-error: true
690779
# See e2e-gpu-strix-ubuntu: root disk is full/non-persistent, so keep the
691780
# toolchain and temp files on the persistent runner directory.
@@ -719,12 +808,36 @@ jobs:
719808
name: E2E tests (Strix Halo, Windows)
720809
runs-on: [self-hosted, windows, strix-halo]
721810
needs: [changes, build-and-test]
722-
if: needs.changes.outputs.heavy == 'true'
811+
# See `e2e`: dispatch tolerates skipped build-and-test; strix-windows, expect-pass.
812+
if: >-
813+
always()
814+
&& needs.changes.result == 'success'
815+
&& (
816+
(github.event_name != 'workflow_dispatch'
817+
&& needs.build-and-test.result == 'success'
818+
&& needs.changes.outputs.heavy == 'true')
819+
|| (github.event_name == 'workflow_dispatch'
820+
&& (inputs.platform == 'all' || inputs.platform == 'strix-windows')
821+
&& (inputs.tier == 'both' || inputs.tier == 'expect-pass'))
822+
)
723823
continue-on-error: true
724824
steps:
725825
- uses: actions/checkout@v6
726826

727-
- uses: actions-rust-lang/setup-rust-toolchain@v1
827+
# setup-rust-toolchain runs an internal bash script, which this Windows
828+
# runner lacks (bash: command not found). Bootstrap rustup with the
829+
# PowerShell-native installer instead; idempotent, so it only downloads on
830+
# a runner that doesn't already have the toolchain.
831+
- name: Ensure Rust toolchain (PowerShell)
832+
shell: pwsh
833+
run: |
834+
if (-not (Get-Command cargo -ErrorAction SilentlyContinue)) {
835+
Invoke-WebRequest https://win.rustup.rs/x86_64 -OutFile $env:TEMP\rustup-init.exe
836+
# --default-toolchain none: rust-toolchain.toml pins the exact
837+
# version (1.96.0 + components), auto-installed on first cargo use.
838+
& $env:TEMP\rustup-init.exe -y --default-toolchain none
839+
"$env:USERPROFILE\.cargo\bin" | Out-File -FilePath $env:GITHUB_PATH -Append
840+
}
728841
729842
- name: Run E2E tests on Strix Halo Windows (excluding known bugs)
730843
shell: pwsh
@@ -741,12 +854,34 @@ jobs:
741854
name: E2E tests (Strix Halo, Windows, known bugs)
742855
runs-on: [self-hosted, windows, strix-halo]
743856
needs: [changes, build-and-test]
744-
if: needs.changes.outputs.heavy == 'true'
857+
# See `e2e`: dispatch tolerates skipped build-and-test; strix-windows, known-bugs.
858+
if: >-
859+
always()
860+
&& needs.changes.result == 'success'
861+
&& (
862+
(github.event_name != 'workflow_dispatch'
863+
&& needs.build-and-test.result == 'success'
864+
&& needs.changes.outputs.heavy == 'true')
865+
|| (github.event_name == 'workflow_dispatch'
866+
&& (inputs.platform == 'all' || inputs.platform == 'strix-windows')
867+
&& (inputs.tier == 'both' || inputs.tier == 'known-bugs'))
868+
)
745869
continue-on-error: true
746870
steps:
747871
- uses: actions/checkout@v6
748872

749-
- uses: actions-rust-lang/setup-rust-toolchain@v1
873+
# See e2e-gpu-strix-windows: bootstrap rustup via PowerShell because this
874+
# runner has no bash for setup-rust-toolchain's internal script.
875+
- name: Ensure Rust toolchain (PowerShell)
876+
shell: pwsh
877+
run: |
878+
if (-not (Get-Command cargo -ErrorAction SilentlyContinue)) {
879+
Invoke-WebRequest https://win.rustup.rs/x86_64 -OutFile $env:TEMP\rustup-init.exe
880+
# --default-toolchain none: rust-toolchain.toml pins the exact
881+
# version (1.96.0 + components), auto-installed on first cargo use.
882+
& $env:TEMP\rustup-init.exe -y --default-toolchain none
883+
"$env:USERPROFILE\.cargo\bin" | Out-File -FilePath $env:GITHUB_PATH -Append
884+
}
750885
751886
- name: Run known-bug E2E scenarios on Strix Halo Windows
752887
shell: pwsh
@@ -782,7 +917,12 @@ jobs:
782917
- e2e-gpu-strix-ubuntu-known-bugs
783918
- e2e-gpu-strix-windows
784919
- e2e-gpu-strix-windows-known-bugs
785-
if: always() && needs.changes.outputs.heavy == 'true'
920+
# Consolidate whatever ran. On dispatch `heavy` is unset, so also run when the
921+
# trigger was manual; `always()` still lets it collect partial/failed tiers.
922+
if: >-
923+
always()
924+
&& (needs.changes.outputs.heavy == 'true'
925+
|| github.event_name == 'workflow_dispatch')
786926
steps:
787927
- uses: actions/checkout@v6
788928

0 commit comments

Comments
 (0)