{"entries":[{"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-05-15T08:15:17.000000+01:00"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-05-15T07:44:34.000000Z"},"hash":"f438e311cb5fe49b2d87b7c48d5751215377f862","message":"mod/modcache: retry .partial file writes on Windows\n\nWriting the .partial marker via os.WriteFile occasionally fails on\nWindows with ERROR_ACCESS_DENIED, breaking testscript runs that fetch\nmodules into a fresh per-test cache. The error has been observed three\ntimes in the past year and again this week; see\nhttps://cuelang.org/issue/3413 for prior investigation.\n\nUpstream Go's modfetch uses a plain os.WriteFile here too, but it does\nnot hit the failure as often because most invocations reuse a long-lived\nGOMODCACHE. In CUE each testscript test starts with a cold cache, often\non a drive that is not excluded from antivirus scanning, which makes the\ntransient ERROR_ACCESS_DENIED visible.\n\nAdd a WriteFile wrapper to internal/robustio that retries ephemeral\nerrors using the existing retry/isEphemeralError machinery, and use it\nfor the .partial file write. Other platforms delegate straight to\nos.WriteFile.\n\nFixes #3413, hopefully.\n\nSigned-off-by: Daniel Martí \nChange-Id: Iefa79ef5a1311c91435d46cb4a7560bb64deb4a3\nReviewed-on: https://cue.gerrithub.io/c/cue-lang/cue/+/1237192\nReviewed-by: Roger Peppe \nTryBot-Result: CUEcueckoo \nUnity-Result: CUE porcuepine \n","tree":"eb7dcf507604bcb7bdde1421e53bd5ef4e0d3a5f"},"mode":"0040000","name":"modcache","size":0},{"lastCommit":{"author":{"email":"mpvl@gmail.com","name":"Marcel van Lohuizen","when":"2026-04-10T13:12:56.000000+02:00"},"committer":{"email":"mpvl@gmail.com","name":"Marcel van Lohuizen","when":"2026-04-10T13:08:19.000000Z"},"hash":"30718b728943768b9ea6db59213266e5cd2f395f","message":"mod/modconfig: set DOCKER_CONFIG in tests to prevent hangs\n\nTests that call NewRegistry without setting\nDOCKER_CONFIG fall through to the system\n~/.docker/config.json, which may have a\ncredsStore credential helper that hangs\nwhen Docker Desktop is not running.\n\nWrite an empty config.json to a temp dir\nand point DOCKER_CONFIG at it so that\nociauth skips the credential helper.\n\nSigned-off-by: Marcel van Lohuizen \nChange-Id: Ica338f39e09eada33db16a110ad83cffacf5e8d5\nReviewed-on: https://cue.gerrithub.io/c/cue-lang/cue/+/1235298\nReviewed-by: Daniel Martí \nTryBot-Result: CUEcueckoo \nUnity-Result: CUE porcuepine \nReviewed-by: Roger Peppe \n","tree":"dca1c7562ef5e823522c2a7a8d0e9c742037324a"},"mode":"0040000","name":"modconfig","size":0},{"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-03-16T15:45:18.000000Z"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-03-17T11:18:25.000000Z"},"hash":"6a7dca8096c3826d82e31ad0639d0fc813a65504","message":"mod/modfile,internal/mod: clean up now that #2733 is fixed\n\nUsing a cue.Value concurrently is now supported in the API.\n\nAlso remove the TODO about mixing values from different contexts;\nthat restriction has also been removed from the API docs.\n\nSigned-off-by: Daniel Martí \nChange-Id: I10e321284a44852d9a4866460fa917b793a43a12\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/1233762\nTryBot-Result: CUEcueckoo \nUnity-Result: CUE porcuepine \nReviewed-by: Roger Peppe \n","tree":"e6999fd13e1d4bcdb0d18fa7642525c23772e5ad"},"mode":"0040000","name":"modfile","size":0},{"lastCommit":{"author":{"email":"rogpeppe@gmail.com","name":"Roger Peppe","when":"2025-08-13T12:53:33.000000+01:00"},"committer":{"email":"rogpeppe@gmail.com","name":"Roger Peppe","when":"2025-10-22T08:27:51.000000Z"},"hash":"f0d24a95d5af71abc03682aef8b06f9370d2d033","message":"mod/modregistry: mirror referrers in Client.Mirror\n\nThis change was reverted earlier (twice!) because the Central Registry\ndid not support the Referrers API call. It now does (verified locally), we we'll\ntry again.\n\nExtend Client.Mirror to copy OCI referrers (manifests with subject\nfield) along with the main module manifests and blobs. This ensures that\nartifacts like signatures and attestations that reference a module are\npreserved when mirroring between registries.\n\nSigned-off-by: Roger Peppe \nChange-Id: I93f2ba9ab172e7bd7e84f8a3743fafe70af4247e\nReviewed-on: https://cue.gerrithub.io/c/cue-lang/cue/+/1224156\nTryBot-Result: CUEcueckoo \nReviewed-by: Daniel Martí \nUnity-Result: CUE porcuepine \n","tree":"63c1baabb9f14ed7e6136e12e8c8f6ed60d24471"},"mode":"0040000","name":"modregistry","size":0},{"lastCommit":{"author":{"email":"rogpeppe@gmail.com","name":"Roger Peppe","when":"2026-04-15T16:43:52.000000+01:00"},"committer":{"email":"rogpeppe@gmail.com","name":"Roger Peppe","when":"2026-04-16T09:06:52.000000Z"},"hash":"97b4cd09b52d5e6edb8730dac503f703e093dc7b","message":"mod/modregistrytest: add metadata capability\n\nWe want the ability to predicate some behavior on the metadata uploaded\nwith a module, so add the capability to specify module metadata inside\nmod/modregistrytest.\n\nSigned-off-by: Roger Peppe \nChange-Id: I934b4aaf004b0ac9e2ca6e03c37b3fc0d2cc581c\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/1235499\nUnity-Result: CUE porcuepine \nReviewed-by: Daniel Martí \nTryBot-Result: CUEcueckoo \n","tree":"112b7f3c76fe195d808b0e8a271d4789e8df8707"},"mode":"0040000","name":"modregistrytest","size":0},{"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-03-17T16:24:07.000000Z"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-03-17T17:05:58.000000Z"},"hash":"1de56c5730b9b7c9bbab04eb3d321312777f4814","message":"mod/module,cue/load,internal/filetypes: reject version in import qualifier\n\nGive a clear error when a version appears inside the package qualifier\nof an import path, such as \"foo.com/bar:baz@v0\" or \"foo.com/bar@v0:baz@v1\",\nrather than confusing messages like \"no files in package directory with\npackage name other@v0\" or \"cannot combine scope with file\".\n\nThe validation is in CheckImportPath, with an early check in the loader\nfor CLI arguments that may not reach CheckImportPath (e.g. no module).\nIsPackage uses the presence of \"/\" in the path to distinguish package\npaths from filetype scopes when the qualifier isn't a valid identifier.\n\nFixes #2859.\n\nSigned-off-by: Daniel Martí \nChange-Id: I92e4e9e50e6f56dc35bf51c97f073ca83b1248c5\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/1233779\nUnity-Result: CUE porcuepine \nReviewed-by: Roger Peppe \nTryBot-Result: CUEcueckoo \n","tree":"00ac74365eb6fb9bcf88f4d3af134333ccd4b04b"},"mode":"0040000","name":"module","size":0},{"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-02-02T13:20:41.000000Z"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-02-09T10:48:47.000000Z"},"hash":"3d3f74bf6502ea4939ab334e7820126bb7dd90c8","message":"all: go fix -any=false ./...\n\nUsing Go at master as of this week.\nThis applies all modernizer fix except the rewrite of `interace{}`\nwith `any`, which is quite noisy, so that is left for a separate step.\n\nSigned-off-by: Daniel Martí \nChange-Id: Ib486b69063203d1774bb3b63fed1d7c716f5f436\nReviewed-on: https://cue.gerrithub.io/c/cue-lang/cue/+/1230267\nTryBot-Result: CUEcueckoo \nUnity-Result: CUE porcuepine \nReviewed-by: Matthew Sackman \n","tree":"026ba3df1010fe7e388db86b93fe8b9e8dfb7b57"},"mode":"0040000","name":"modzip","size":0}],"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-05-15T08:15:17.000000+01:00"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-05-15T07:44:34.000000Z"},"hash":"f438e311cb5fe49b2d87b7c48d5751215377f862","message":"mod/modcache: retry .partial file writes on Windows\n\nWriting the .partial marker via os.WriteFile occasionally fails on\nWindows with ERROR_ACCESS_DENIED, breaking testscript runs that fetch\nmodules into a fresh per-test cache. The error has been observed three\ntimes in the past year and again this week; see\nhttps://cuelang.org/issue/3413 for prior investigation.\n\nUpstream Go's modfetch uses a plain os.WriteFile here too, but it does\nnot hit the failure as often because most invocations reuse a long-lived\nGOMODCACHE. In CUE each testscript test starts with a cold cache, often\non a drive that is not excluded from antivirus scanning, which makes the\ntransient ERROR_ACCESS_DENIED visible.\n\nAdd a WriteFile wrapper to internal/robustio that retries ephemeral\nerrors using the existing retry/isEphemeralError machinery, and use it\nfor the .partial file write. Other platforms delegate straight to\nos.WriteFile.\n\nFixes #3413, hopefully.\n\nSigned-off-by: Daniel Martí \nChange-Id: Iefa79ef5a1311c91435d46cb4a7560bb64deb4a3\nReviewed-on: https://cue.gerrithub.io/c/cue-lang/cue/+/1237192\nReviewed-by: Roger Peppe \nTryBot-Result: CUEcueckoo \nUnity-Result: CUE porcuepine \n","tree":"eb7dcf507604bcb7bdde1421e53bd5ef4e0d3a5f"}}