Depends on https://github.com/nix-community/nix-eval-jobs/pull/349 & #1421.
Almost equivalent to #1425, but with a small change: when having e.g. an
aggregate job with a glob that matches nothing, the jobset evaluation is
failed now. This was the intended behavior before (hydra-eval-jobset
fails hard if an aggregate is broken), the code-path was never reached
however since the aggregate was never marked as broken in this case
before.
There were some hangs caused by this. Need to fix them, ideally
reproducing the issue in a test, before trying this again.
This reverts commit 4a4a0f901c70676ee47f830d2ff6a72789ba1baf.
My main motivation here is to get metrics with brackets to work in order
to support "pytest" test names:
- test_foo.py::test_bar[1]
- test_foo.py::test_bar[2]
I couldn't find an "HTML escape"-style function that would generate
valid html `id` attribute names from random strings, so I went with a
hash digest instead.
Needed one more thing before trying out using `LegacySSHStore` directly.
Flake lock file updates:
• Updated input 'nix':
'github:NixOS/nix/674a87462cb93f605d4fbeef607d3453e7e5a7d8?narHash=sha256-TBoHqnIdVWhsBcL05vO2B1hSl9m//5Mz2NU%2BPMk3h3Y%3D' (2025-02-16)
→ 'github:NixOS/nix/e310c19a1aeb1ce1ed4d41d5ab2d02db596e0918?narHash=sha256-q/RgA4bB7zWai4oPySq9mch7qH14IEeom2P64SXdqHs%3D' (2025-02-18)
This avoids some duplicated code, leveraging the same `StoreReference`
type that also undergirds the machine file dedup we just did prior.
By using `LegacySSHStoreConfig`, we're also taking a baby step towards
using the store interface rather than messing around with the protocol
internals.
incrementally ingest eval results
nix-eval-jobs streams output, unlike hydra-eval-jobs. Now that we've
migrated, we can use this to:
1. Use less RAM by avoiding buffering a whole eval's worth of metadata
into a Perl string and an array of JSON objects.
2. Make evals latency a bit lower by allowing the queue runner to start
ingesting builds faster.
Also use the newly-restored constituents support in `nix-eval-jobs`
Note, we pass --workers and --max-memory-size to n-e-j
Lost in the h-e-j -> n-e-j migration, causing evaluation to always be
single threaded and limited to 4GiB RAM. Follow the config settings like
h-e-j used to do (via C++ code).
`nix-eval-jobs` should check `hydraJobs` and then `checks` with flakes
(cherry picked from commit 6d4ccff43c41adaf6e4b2b9bced7243bc2f6e97b)
(cherry picked from commit b0e9b4b2f99f9d8f5c4e780e89f955c394b5ced4)
(cherry picked from commit cdfc5c81e8037d3e4818a3e459d0804b2c157ea9)
(cherry picked from commit 4b107e6ff36bd89958fba36e0fe0340903e7cd13)
Co-Authored-By: Maximilian Bosch <maximilian@mbosch.me>
It seemed there was no self-contained end-to-end test actually doing
this?!
Among other things, this will help ensure that the switch-over to
`nix-eval-jobs` is correct.
Just ordering yourself after network-online.target will not guarantee
that it will be loaded. You'll have to either want or require it. Hence
the following trace on recent nixpkgs versions:
evaluation warning: hydra-queue-runner.service is ordered after 'network-online.target' but doesn't depend on it
* Update version in example
* Update docs to fix invalid indentifier when using 'hello'
* fix build issue for hello example
---------
Co-authored-by: Aaron Honeycutt <aaronhoneycutt@proton.me>
Original commit message:
> There are some known regressions regarding local testing setups - since
> everything was kinda half written with the expectation that build dir =
> source dir (which should not be true anymore). But everything builds and
> the test suite runs fine, after several hours spent debugging random
> crashes in libpqxx with MALLOC_PERTURB_...
I have not experienced regressions with local testing.
(cherry picked from commit 4b886d9c45cd2d7fe9b0a8dbc05c7318d46f615d)
