Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ ZENODOarrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Software
Data sources: ZENODO
addClaim

datalad/datalad: 1.6.1

Authors: Hanke, Michael; Halchenko, Yaroslav O.; Poldrack, Benjamin; Meyer, Kyle; Solanky, Debanjum Singh; Alteva, Gergana; Gors, Jason; +27 Authors

datalad/datalad: 1.6.1

Abstract

🐛 Bug Fixes get_toppath() now returns None instead of crashing with TypeError when called on a path whose existence can't be verified (e.g. a broken symlink or a nonexistent directory). Discovered while investigating #7882. (by @LiamDGray) BF: terminate child before closing stdout/stderr on process timeout. PR #7866 (by @yarikoptic) Fix get_dataset_root() to reject an ancestor .git/ that is reached only by crossing a symlink boundary. Previously, cwd inside a symlink whose apparent parent contained an unrelated .git/ (e.g. an abandoned git init with no commits) caused datalad status / datalad rerun to bind to that parent, while git and git-annex correctly followed the symlink to the real dataset. The existing "symlink itself is a dataset" fallback now handles that case; the test_bf1886 semantics (symlink → subdataset within outer dataset resolves to outer) is preserved via a realpath-ancestry check. Fixes #7882 via PR #7883 (by @yarikoptic) create-sibling no longer requires git-annex on the target (remote or local) when the source dataset is a plain-git dataset (--no-annex, i.e. marked with a .noannex file). Previously it errored out with No working git-annex installation. It's required on the remote machine to create a sibling even though no annex functionality would be used. Via PR #7888 (by @yarikoptic) 🔩 Dependencies Bump minimum required Git version from 2.25.0 to 2.31.0. The -c key=value config-override propagation to subprocesses added in PR #7858 relies on Git's GIT_CONFIG_COUNT / GIT_CONFIG_KEY_<n> / GIT_CONFIG_VALUE_<n> environment mechanism, which was introduced in Git 2.31.0 (2021-03). On older Git the CLI override was silently ignored. This bump makes the requirement explicit at startup rather than manifesting as a silent feature loss. Git 2.31 is available on all currently-supported LTS distributions. Via PR #7887 (by @yarikoptic) 🏠 Internal Pin asv<0.6.6 and asv-runner<0.2.2 in .github/workflows/benchmarks.yml. Starting with asv-runner 0.2.2 (released 2026-06-27), do_setup_cache() invokes any class-level setup(self) method it can bind with no positional args -- for the save_run benchmarks that means setup() runs before setup_cache() has written bm_heavy.tar, breaking every Benchmarks CI run since #7402. asv itself needs pinning too because asv 0.6.6 requires asv-runner>=0.2.5, and tools/ci/benchmark-travis-pr.sh re-runs pip install asv unconditionally which would otherwise defeat the pin. Reported upstream at airspeed-velocity/asv_runner#52. PR #7886 (by @yarikoptic)

Powered by OpenAIRE graph
Found an issue? Give us feedback
Funded by