summaryrefslogtreecommitdiff
path: root/.gitlab-ci.yml
Commit message (Collapse)AuthorAgeFilesLines
...
* Merge branch 'msrv-1-86-0' into 'main'David Goulet2025-10-211-1/+1
|\ | | | | | | | | Bump MSRV from 1.85.1 to 1.86 See merge request tpo/core/arti!3382
| * Bump MSRV from 1.85.1 to 1.86Clara Engler2025-10-211-1/+1
| |
* | CI: Specify versions for shellcheck and editorconfig-checker.Wesley Aptekar-Cassels2025-10-161-2/+2
|/ | | | Fixes: #2219
* ci: Parallelize minimal-features-testWesley Aptekar-Cassels2025-10-161-1/+7
| | | | | | | 3 seems like a good number to start with, we can tune it later if needed. Fixes: #2207
* CI: remove 2h timeouts from now-smaller jobsJim Newsome2025-10-091-12/+0
| | | | | Hopefully these are now sufficiently small to fit into the default 1h timeout.
* CI rust-dev-latest: add comment noting that it could be split furtherJim Newsome2025-10-091-0/+5
|
* CI rust-recent-dev-test: split out testJim Newsome2025-10-091-5/+22
| | | | | | | | | | | | This job is still expensive and a bottleneck. "cargo test" is the slowest part, and experimentally doesn't seem to generate artifacts used in the other steps nor reuse artifacts genrated in other steps. Splitting it out to its own job exposes more parallelism. "cargo test" is also more prone to flakiness than the other commands, which just build code without running it. Isolating the flakiness is helpful when we need to retry. (Though of course ultimately we want to eliminate such flakinesss)
* CI rust-recent-dev-test: inline rust-dev-test-script yaml anchorJim Newsome2025-10-091-14/+5
|
* CI: drop rust-latest-integrationJim Newsome2025-10-091-37/+19
| | | | | | | There's probably not much benefit to building these on rust-latest in addition to rust-recent; incompatibilities with rust-latest should be found in rust-latest-dev-test. Let's save the template complexity and CI cost.
* CI: split rust-recent script and jobsJim Newsome2025-10-091-8/+41
| | | | | | | | | | | | rust-recent (and rust-latest, for runs on the main branch) is one of our longer jobs, particularly in the 'build' CI stage, making it a bottleneck for starting the 'test' stage. The job does quite a bit of work, some of it with very little overlap due to use of different cargo build profiles. This commit splits the job roughly by build-profile, so that the work done with each build profile can proceed in parallel.
* CI: build arti-bench with --lockedJim Newsome2025-10-091-1/+1
| | | | | This is mostly for consistency with the following "cargo build" invocation to build arti, but is unlikely to make any real difference.
* CI: build arti-bench with quicktest instead of releaseJim Newsome2025-10-091-2/+2
| | | | | | | | | | | Right now building arti-bench take a fair while in the rust-recent and rust-latest jobs. Using the quicktest profile instead of release makes any actual benchmark results less meaningful, but as far as I can tell this build artifact is mostly just to test that it works at all rather than to collect meaningful performance data. e.g. we're currently not controlling at all for runner capacity or load.
* CI: increase timeout to 2h on some jobsJim Newsome2025-10-081-0/+16
| | | | | | | | This is to mitigate https://gitlab.torproject.org/tpo/tpa/team/-/issues/42317 I added the 2h timeout (overriding the default 1h) for all jobs that were cited as examples on that issue, and a couple others that looked likely to be time-consuming.
* CI: dump /proc/loadavg and /proc/pressure/* for debugging purposesJim Newsome2025-10-081-0/+6
|
* Revert restricting heavy cargo jobs to TPA runnersJim Newsome2025-10-081-12/+0
| | | | | * This reverts commit 8da606966e5213cada270fb64f896f69bab041dd. * This reverts commit a3ac2080956b9295000bd7011c34111a6d5ce5a7.
* CI: limit cargo parallelism based on memory and ncpuJim Newsome2025-10-081-0/+4
| | | | See https://gitlab.torproject.org/tpo/tpa/team/-/issues/42283
* Merge branch 'ci-remove-check-stage' into 'main'wesleyac2025-10-071-5/+4
|\ | | | | | | | | ci: Move "check" jobs into "build" stage. See merge request tpo/core/arti!3318
| * ci: Move "check" jobs into "build" stage.Wesley Aptekar-Cassels2025-10-021-5/+4
| | | | | | | | | | | | | | | | | | | | | | This moves everything that was in the "check" stage into the "build" stage, so that beginning to build the artifacts isn't blocked by running the checks. This should result in CI taking less time overall. There is some concern that making CI more efficient by increasing parallelism could increase load on runners and cause more problems than it solves, so that's something we should look out for in evaluating this change.
* | ci: build fuzz testsSteven Engler2025-10-061-0/+16
| |
* | Merge branch 'ci-move-maint-check-changelog' into 'main'wesleyac2025-10-061-6/+1
|\ \ | | | | | | | | | | | | ci: Move maint-check-changelog into maint-checks. See merge request tpo/core/arti!3316
| * | ci: Move maint-check-changelog into maint-checks.Wesley Aptekar-Cassels2025-10-021-6/+1
| |/ | | | | | | | | | | | | | | | | | | I hope that this will reduce the load on our CI and speed up these checks. This should also cause this check to run earlier in the process, which is nice for finding issues earlier. It's unclear to me if there was a specific reason this was a separate job. As far as I can tell from the commit message and discussion in !2126, there does not seem to be a specific reason.
* | ci: Move minimal-features-test and every-crate to TPA runners.Wesley Aptekar-Cassels2025-10-021-0/+8
| | | | | | | | | | | | | | We are currently having some issues with these jobs on the OSUOSL runners, using TPA runners should allow us to debug this more effectively. See: tpo/tpa/team#42283
* | ci: Only run rust-recent-test-all-features on TPA runners.Wesley Aptekar-Cassels2025-10-021-0/+4
|/ | | | | | | We are currently having some issues with this job on the OSUOSL runners, using TPA runners should allow us to debug this more effectively. See: https://gitlab.torproject.org/tpo/tpa/team/-/issues/42283
* Remove check_doc_features.Nick Mathewson2025-09-291-10/+0
| | | | | | | | Now that rust nightly has merged doc_auto_cfg into doc_cfg, and completed https://github.com/rust-lang/rust/pull/138907, it is no longer necessary. Closes #1514.
* maint: Add script to check OpenTelemetry macro invocations.Wesley Aptekar-Cassels2025-09-241-0/+1
|
* python3-checks: quote variable definitions and referencesJim Newsome2025-09-161-6/+6
| | | | | No functional differences as far as I can tell, but keeps with gitlab's recommendations for consistency and to avoid potential footguns.
* python3-lints: bump mypy to 1.18.1Jim Newsome2025-09-151-1/+2
|
* python3-lints: bump flake8 to 7.3.0Jim Newsome2025-09-151-1/+2
|
* python3-lints: bump black to 25.1.0Jim Newsome2025-09-151-1/+2
|
* python-checks: lift pinned versions to job vars for visibilityJim Newsome2025-09-151-3/+7
|
* python3-checks: break into separate shell commands and reformatJim Newsome2025-09-151-5/+13
| | | | | This is mostly to allow putting the pip3 dependencies each on their own line without escaping newlines, but also more idiomatic in general.
* python3-checks: don't allow failureJim Newsome2025-09-151-1/+0
| | | | | I think now that the linter versions are pinned (!2683), we don't want to allow these to fail anymore.
* Bump MSRV from 1.85 to 1.85.1Nick Mathewson2025-09-091-1/+1
| | | | Closes #2107
* CI: Remove --verbose from cargo commandsNick Mathewson2025-09-041-11/+11
| | | | | | | This is taking up a lot of space in our CI longs, and hasn't actually helped us out in the past. Possible solution for #2155
* Pin CI to cargo-sort 2.0.1Nick Mathewson2025-09-041-1/+1
| | | | | | | There are significant changes in the output from cargo-sort 2.0.2, so let's pause at 2.0.1 until we decide on our next step. Temporary solution for #2156.
* shadow tests: use a randomized simulation seedJim Newsome2025-09-031-2/+10
| | | | | | | | | | | | | | Mitigates #2154 The determinism inside CI is of limited usefulness since a CI seed generally won't perfectly replicate the behavior when debugging locally. Arguably it's a feature that when there's some flakiness it does replicate inside of CI, but it leads to some confusion about blame (e.g. as in #2154 where an unrelated MR probably perturbed the PRNG to trigger some pre-existing flakiness), and means that we can't just rerun the job if we get really unlucky. i.e. even if we set a lower success threshold, a sufficiently unlucky MR can still fail and keep failing as long as it runs with the same seed.
* Merge branch 'ci-case' into 'main'David Goulet2025-09-021-0/+1
|\ | | | | | | | | maint: Add a check for case-insensitive files See merge request tpo/core/arti!3207
| * maint: Add a check for case-insensitive filesClara Engler2025-08-281-0/+1
| | | | | | | | | | | | | | | | | | | | | | Filenames within the tree that exists in multiple versions only differing by their case cause problems on case-insensitive filesystems. This commit introduces a simple check that lists all files indexes by git and searches for duplicate lines while ignoring case during comparison. See !3166, #2144
* | Merge branch 'use-quicktest' into 'main'Jim Newsome2025-08-271-4/+12
|\ \ | | | | | | | | | | | | Use quicktest build for integration testing See merge request tpo/core/arti!3157
| * | .gitlab-ci.yml: add pipeline variables for which cargo profile to use to ↵Jim Newsome2025-08-261-5/+12
| | | | | | | | | | | | build arti
| * | Integration testing: use quicktest buildJim Newsome2025-08-261-4/+5
| |/
* | Merge branch '2064-artifact-expiration-times' into 'main'opara2025-08-271-2/+2
|\ \ | | | | | | | | | | | | | | | | | | Increase artifact expiration times to 1 day minimum Closes #2064 See merge request tpo/core/arti!3098
| * | If artifact expiration time is less than 1 day, increase to 1 day.luciole2025-07-161-2/+2
| | |
* | | ci: update "recent rust image" to 1.89Steven Engler2025-08-261-2/+2
| |/ |/|
* | ci: drop dependency proxy experimentJérôme Charaoui2025-08-211-19/+10
| | | | | | | | | | | | | | | | | | | | | | the dependency proxy didn't work as well as we hoped (tpo/tpa/team#42089) and it also fix all rate-limiting issues from dockerhub since a lot of contributors were bypassing the proxy because of insufficient credentials from now on pulls from "docker.io" will go through a new pull-though cache deployed by tpa, similar to what osuosl folks are using successfully with their own ci runners
* | CI: bump chutney versionJim Newsome2025-08-111-2/+2
| | | | | | | | | | Primarily to get arti hidden services support, added in tpo/core/chutney!77
* | Try updating the miri toolchain tooNick Mathewson2025-08-051-2/+2
| |
* | Bump minimal-versions in CI to new MSRV (1.85)Nick Mathewson2025-08-051-1/+1
| |
* | ci: Pin to cargo-license 0.7.0Nick Mathewson2025-07-301-2/+1
| | | | | | | | This should prevent recurrences of issues like #2083.
* | ci: downgrade cargo-license to 0.6.1 as workaroundSteven Engler2025-07-291-1/+2
|/
generated by cgit v1.2.3 (git 2.46.0) at 2026-08-26 18:52:30 +0000