GeezerVison is showing github.com — ci: restore build checks and CPU summation accuracy by melonakos · Pull Request #3735 · arrayfire/arrayfire
Conversation
melonakos
commented
Oct 24, 2026
•
edited
Loading
Uh oh!
There was an error while loading. Please reload this page.
Retired runners, stale tools, unavailable fixtures, and LAPACK packaging errors prevented CI from completing. Use supported runners, CMake 3.28.6 on Unix, vcpkg 2026.10.16 tools, provided Windows Fortran, and Homebrew runtime paths. Restore verified public SuiteSparse fixtures and cache dependencies while keeping pinned library versions; expose compiler errors and make dashboard uploads optional. All configured CPU checks, documentation, formatting, integrity tests, and Actionlint pass; the existing Windows pinverse exclusion remains. The CPU summation commit from #3687 fixes variance precision, with guards and regressions for infinities, NaNs, and finite overflow. Merge #3736 and #3710 first; #3731's overlapping reduction must retain the nonfinite guard.
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
std::abs<CT>(x) with an explicit template argument is rejected as
ambiguous by newer libc++ (e.g. the macOS 26 SDK), which brings the
integer abs overloads into scope. s1var/validsVar are already CT, so
drop the argument and let std::abs resolve the <cmath> overload.
Normalize the existing source, test, and example formatting with clang-format 15. Preserve the cast table with formatting guards.
Build documentation with Ubuntu packages and align Windows vcpkg with the manifest baseline. Use explicit tool paths, propagate failures, and build before running CPU tests.
Pin the compatible runner and fetch the complete version registry before invoking vcpkg. Update the binary cache action to its supported runtime.
Test the workflow repairs with the formatting baseline. Merge the separate formatting PR first so the workflow PR retains its own changes.
Use a current vcpkg tool checkout for available MSYS2 packages while keeping the manifest baseline. Use CMake 3.28.6 on Linux and macOS for IntelLLVM support.
Reuse complete cached matrices before invoking the downloader and save them after configuration. Stop before repeated fixture requests when SuiteSparse declares an outage, while keeping matrix tests enabled.
melonakos
changed the title
ci: restore supported runners and the vcpkg baseline
ci: restore runners, build tools, and fixture caching
Oct 24, 2026
Bundle the retained SuiteSparse fixtures with their source metadata and checksums. Validate archive contents before restoring the cache so matrix tests run without the offline provider.
Show compiler failures directly and stop before testing missing binaries. Keep dashboard uploads optional so its outage does not override build and test results.
Enable the provided MinGW Fortran toolchain in the custom Windows triplet so LAPACK can configure with current vcpkg tools.
Add Homebrew's library directory to the test runtime search path so dynamically loaded FreeImage is available on Apple Silicon runners.
Save validated vcpkg binaries after configuration so subsequent builds can reuse them even if compilation or tests fail.
Let the BLAS port own its pkg-config files. Save vcpkg's completed ABI-keyed archives on failed configurations and restore compatible binaries across retries.
Reset a NaN correction term so infinities retain normal sum behavior. Cover real and complex infinities and NaNs through all-element, dimension, and keyed reductions; the regression fails without the guard and the reduction and variance suites pass with it.
melonakos
changed the title
ci: restore runners, build tools, and fixture caching
ci: restore build checks and CPU summation accuracy
Oct 24, 2026
Install the wrapper and find module where the LAPACK meta-port consumes them. Give the reference package its own pkg-config metadata so the meta-port can own the public LAPACK files.
Keep the modified overlay as LF text so patch whitespace checks remain clean.
Reset infinite correction terms as well as NaNs so finite overflow retains IEEE sum behavior. Extend the real and complex API regression; the reduction and variance suites pass.
This was referenced Oct 24, 2026
roaffix
added a commit
that referenced
this pull request
Nov 3, 2026
The sparse tests' matrices come from the SuiteSparse collection, whose
storage refuses every download ("403 AccessDenied"). ArrayFire's
arrayfire-data repository holds the same nine files, unchanged, under
matrixmarket/, with their CC BY 4.0 attribution. CI downloads them from
arrayfire-data at commit 2723164, which matrixmarket.json pins, so the
Matrix Market tests run again; an empty commit would turn them off.
The manifest of each matrix's size and SHA-256 and the idea of checking
the data before a build come from ArrayFire's fix/restore-ci branch
(#3735), whose check_matrix_data.py is adapted as check-matrix-data.py.
Before every build, ci.sh runs it. It downloads each matrix from the
pinned commit on raw.githubusercontent.com, checks it against the
manifest and the tested tree's list of matrices, fails naming every
file that differs or doesn't download, restores the matrices into the
build tree and prints ON; without a pinned commit it prints OFF.
ci.sh passes the answer on as AF_CI_MTX_FILES. The tests download each
matrix through af_dep_check_and_populate, whose FetchContent_Populate
takes FETCHCONTENT_SOURCE_DIR_<name> as the source when it is set; the
ci preset and matrixmarket.cmake (for the GPU builds and the CMake 3.16
build) set those to the restored matrices, and AF_TEST_WITH_MTX_FILES
to AF_CI_MTX_FILES. Nothing is downloaded from SuiteSparse, and nothing
under test/ changes. The GPU build archives carry the matrices to the
test jobs.
syurkevi
pushed a commit
that referenced
this pull request
Nov 3, 2026
The sparse tests' matrices come from the SuiteSparse collection, whose
storage refuses every download ("403 AccessDenied"). ArrayFire's
arrayfire-data repository holds the same nine files, unchanged, under
matrixmarket/, with their CC BY 4.0 attribution. CI downloads them from
arrayfire-data at commit 2723164, which matrixmarket.json pins, so the
Matrix Market tests run again; an empty commit would turn them off.
The manifest of each matrix's size and SHA-256 and the idea of checking
the data before a build come from ArrayFire's fix/restore-ci branch
(#3735), whose check_matrix_data.py is adapted as check-matrix-data.py.
Before every build, ci.sh runs it. It downloads each matrix from the
pinned commit on raw.githubusercontent.com, checks it against the
manifest and the tested tree's list of matrices, fails naming every
file that differs or doesn't download, restores the matrices into the
build tree and prints ON; without a pinned commit it prints OFF.
ci.sh passes the answer on as AF_CI_MTX_FILES. The tests download each
matrix through af_dep_check_and_populate, whose FetchContent_Populate
takes FETCHCONTENT_SOURCE_DIR_<name> as the source when it is set; the
ci preset and matrixmarket.cmake (for the GPU builds and the CMake 3.16
build) set those to the restored matrices, and AF_TEST_WITH_MTX_FILES
to AF_CI_MTX_FILES. Nothing is downloaded from SuiteSparse, and nothing
under test/ changes. The GPU build archives carry the matrices to the
test jobs.
This branch has not been deployed
No deployments
Uh oh!
There was an error while loading. Please reload this page.
Assignees
No one assigned
Milestone
No milestone
Development
Successfully merging this pull request may close these issues.
Uh oh!
There was an error while loading. Please reload this page.
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.
|