COMP: Build against ITK main; sync ExternalData config; fix docs env#456
COMP: Build against ITK main; sync ExternalData config; fix docs env#456hjmjohnson wants to merge 6 commits into
Conversation
github-actions
Bot
added
area:Filtering
|
Finding from #455's re-run: the sha512→cid conversion here is required, not cosmetic — the top-level |
|
Folded the visual-cid fix from #455 (closed) into this PR as f759c48 — with ITKTestingData#63/#64/#65 all merged, every content link this PR's CI needs is now published. Remaining known blocker: the ITKReview/zlib link regression on ITK main hits the superbuild jobs (fix validated locally; ITK PR pending). |
github-actions
Bot
added
area:Core
hjmjohnson
force-pushed
the
fix/ci-itk-pin-and-deps
branch
from
f759c48 to
d2521d3
Compare
Superbuild ITK_TAG v5.4.2 -> main, matching the C++ CI jobs and picking up post-5.4.2 fixes behind the Python baseline mismatches. pip itk pins -> >=5.4.6; zarr>=3.0.0 satisfies ngff-zarr OME-Zarr 0.5 (nbsphinx notebook execution failed without it). ExternalData URL templates and module now mirror ITK main: adds the Girder hashsum mirror (sole host of some legacy sha512 objects, e.g. ComputeFFTInOneDimension), filebase and pinata public gateways, and the WASI symlink guard; drops dead estuary.tech/w3s.link and the released-root-CID gateways. The project-specific gin.g-node.org mirror and ExternalData_LINK_CONTENT=CID are retained.
Last remaining hash-based links in the repository; blobs published under CID in ITKTestingData#64.
The 20 PNGs referenced since InsightSoftwareConsortium#440 were never published to ITKTestingData and are unrecoverable from any public source; every build of Module_SphinxExamples fails 20 ExternalData fetches. Point the .cid files at recreated diagrams (ITKTestingData#63) and fix the Operator.png/Operation.png reference mismatch.
hjmjohnson
force-pushed
the
fix/ci-itk-pin-and-deps
branch
from
d2521d3 to
cc33939
Compare
|
Fixed the 277-test failure from the previous run: the ExternalData config sync imported ITK's |
The pip itk 5.4.6 wheels and ITK main round differently at quantization boundaries, leaving pixels one gray-level apart (plus sub-integer FFT float noise) between Python and C++ example outputs compared against shared baselines.
github-actions
Bot
added
area:Segmentation
ITK main deprecates ImageConstIterator::GetIndex in favor of the WithIndex iterator variants, and MersenneTwister Initialize() in favor of SetSeed(). The zero-warning dashboard gate fails on these once the superbuild targets ITK main.
Every example includes ${ITK_USE_FILE}, which ITK main deprecates;
migrating ~350 example CMakeLists to imported targets is follow-up
work.
hjmjohnson
force-pushed
the
fix/ci-itk-pin-and-deps
branch
from
7b00030 to
48c37d1
Compare
1 participant
Build against ITK main, sync the ExternalData configuration with ITK upstream, fix the docs notebook env (
zarr>=3), and finish the repo's sha512→cid conversion. With ITKTestingData#63/#64 merged, this addresses the remaining CI-infrastructure failures on every open PR. Blocked on ITKTestingData#65 (CID copies of the FFT blobs).What each change fixes
ITK_TAGv5.4.2 →main(matches the C++ jobs'itk-git-tag)itk>=5.4.2/>=5.4rc3→>=5.4.6zarr>=3.0.0in the superbuild venvbuild_htmlhard failure: nbsphinx executingConvertNumPyArrayToitkImage.ipynbhitsValueError: zarr-python >= 3.0.0b2 required for OME-Zarr >= 0.5ITKSphinxExamplesExternalData.cmake+ExternalData.cmakesynced to ITK mainExternalData_LINK_CONTENT=CIDretainedComputeFFTInOneDimension.sha512→.cid(last 6 hash links repo-wide)MouseLiverRF.mha does not exist); blobs published in ITKTestingData#64 (SHA512) and #65 (CID)