QVAC-23075 chore: DO NOT MERGE, overlay-validate the 7 fabric consumers against the VisionPsy fabric branch - #3814
Open
yingying0906 wants to merge 3 commits into
Open
QVAC-23075 chore: DO NOT MERGE, overlay-validate the 7 fabric consumers against the VisionPsy fabric branch#3814yingying0906 wants to merge 3 commits into
yingying0906 wants to merge 3 commits into
Conversation
Contributor
Review StatusCurrent Status: ❌ PENDING Pending reviews: Needs 1 Management or Team Lead, and 1 more from Management, Team Lead, or Member. |
Contributor
License compliance — cleanNo new dependency license findings in this PR. Warn-only (shadow) mode — this check does not block merges yet. Updated automatically by the canonical license compliance workflow. NOTICE presence (advisory)Missing NOTICE (advisory, does not block):
|
yingying0906
added
prebuilds
Contributor
🧪 C++ Test Coverage ReportCoverage: 📊 Detailed Coverage |
…inst VisionPsy
Rollout Phase A for the VisionPsy fabric change. Adds the shared qvac-fabric overlay
port and points every consumer at it, so all 7 build against the fabric branch head
before the tag exists and before anything is published to the registry.
The port is copied from qvac-registry-vcpkg origin/main, which is two files now;
android-vulkan-version.cmake is gone from the registry and is deliberately not
reinstated here. REF is a branch SHA rather than v${VERSION} because the tag does not
exist yet, and the whole overlay dies at the registry publish.
Consumer version>= pins stay at 10069.0.0. The overlay bypasses version resolution
entirely, so only the overlay port's own version matters here; bumping the pins is
Phase B. default-registry.baseline is untouched in all 7, which is the one thing a
rollout PR must never move.
Roster re-derived rather than assumed:
git grep -l "qvac-fabric" origin/main -- "packages/*/vcpkg.json"
classification-ggml is absent because it consumes the published @qvac/fabric npm
package rather than the vcpkg port.
DO NOT MERGE. Validation only.
… fabric head The branch pinned 7a6425a3a, which predates the review round. Fabric is now 8be2131e2, ten commits on, including the OpenCL view readback fix and the preprocessing changes the consumers should be validated against. SHA512 recomputed from the archive of the full SHA, with the method checked by reproducing the old pin's hash exactly.
The fabric branch gained three commits from the second review pass: the zero image_size rejection extended to idefics3, the libmtmd rebuild note in mtmd.h, and the prompt-sequence assertions in tests.sh. Phase A has to validate the consumers against what will actually be tagged, so the overlay follows the head. SHA512 is the sha512 of the GitHub archive tarball for that commit. Method checked by reproducing the previous pin's hash from the same command before replacing it.
Contributor
🧪 C++ Test Coverage ReportCoverage: 📊 Detailed Coverage |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
DO NOT MERGE
Validation branch for QVAC-23075 rollout Phase A. It carries a temporary vcpkg overlay that pins
qvac-fabricto a branch SHA,8be2131e27d5589667db9c7f780fc1a9ef20869aonfeat/QVAC-23075-visionpsy, because the tag does not exist yet. Merging this would point every consumer at an unpublished commit.The overlay dies at the registry publish in Phase B, which replaces it with the real tag.
What this branch does
Adds the shared overlay port at
vcpkg-overlays/ports/qvac-fabric/and"overlay-ports": ["../../vcpkg-overlays/ports"]to all 7 fabric consumers: embed-llamacpp, fabric, llm-llamacpp, model-fit, ocr-ggml, translation-nmtcpp, vla-ggml. No consumerversion>=pin is bumped anddefault-registry.baselineis untouched, both of which belong to Phase B.Why it is open as a PR
So the consumer CI can be driven by labels instead of a dispatch per package. The routers fan out from this one PR because each consumer's own
vcpkg-configuration.jsonis edited, and everyon-pr-*workflow watches its own package directory.Heavy stages need labels:
prebuilds,run-cpp-addon-tests,run-desktop-addon-tests,run-mobile-addon-tests,run-coload-tests. Without them only the light checks run.merge-guardis expected to fail here and mobile legs are expected to fail on theqvac-test-addon-mobilePAT, which is a known infra blocker raised with devops.Related
Fabric PR: tetherto/qvac-fabric-llm.cpp#205
Consumer PR: #3725