Revert "update flake inputs"

This reverts commit acf8f97b6b. Reason for revert: ci broke? idfk Change-Id: I2f8f02f3a5eb8562f8ff94341b47d82c29ed74ba
update flake inputs
2026-07-06 11:37:09 +00:00 · 2026-07-04 16:25:50 +02:00 · 2026-01-20 11:11:08 +01:00 · 2026-01-20 07:54:20 +01:00 · 2026-01-19 17:04:05 +01:00 · 2025-10-10 15:30:48 +02:00
1 changed files with 6 additions and 1 deletions
@@ -116,7 +116,12 @@ try:
            'group': group,
            'steps': [
                {
-                    'command': f'nix build --no-link -L ".#hydraJobs.{job.name}"',
+                    'command': textwrap.dedent(f'''
+                        set -ueo pipefail
+                        resultPath=$(nix build --print-out-paths --no-link -L ".#hydraJobs.{job.name}^*")
+                        echo "Built paths: $$resultPath"
+                        attic push afnix-ci $$resultPath || true
+                    '''),
                    'label': ".".join(job.path),
                    'key': job.drv_hash(),
                    'depends_on': [ d.drv_hash() for d in deps[job] ],
Author	SHA1	Message	Date
eldritch horrorsandLix Systems Gerrit	9cfd6d9c84	Revert "update flake inputs" This reverts commit `acf8f97b6b`. Reason for revert: ci broke? idfk Change-Id: I2f8f02f3a5eb8562f8ff94341b47d82c29ed74ba	2026-07-06 11:37:09 +00:00
eldritch horrors	acf8f97b6b	update flake inputs it's been a hot second 😰 Change-Id: I9eb3c93dc300ea658b9eec2c5a111d8f446ba61f	2026-07-04 16:25:50 +02:00
Pierre Bourdon	700f43c2cb	ci-config: explicitly build all outputs in `nix build` invocation Without this, even though all outputs will still be built, only the main output (usually $out) store path gets printed by --print-out-paths. In turn, this means that only the main output will get sent to attic for caching. In Lix's case, this was causing the following scenarios: - builder1 builds build.aarch64-darwin.{out,dev,doc} - builder1 only pushes .out to the caches - builder2 attempts to build perlBindings.aarch64-darwin, which depends on -dev from build.aarch64-darwin. - builder2 needs to rebuild all of build.aarch64-darwin to get -dev. Change-Id: I51f7e641f987eebcd65b3cd57c1f6b1bfbfebd32	2026-01-20 11:11:08 +01:00
Pierre Bourdon	e676a6c5b5	ci-config: print the store paths that have been built For easier debugging of what actually gets pushed to attic. Change-Id: I6ac5844045758c3ecc126c4a65fbfa10da5acfdd	2026-01-20 07:54:20 +01:00
Pierre Bourdon	41c63d68f5	ci-config: best-effort push CI intermediate build results to attic If the builder is set up with an attic configuration (as is the case since today on the Lix Buildkite builders), every Buildkite build step attempts to push all their output store paths to attic so they can be reused in later builds and/or in further steps being scheduled on different builders. Change-Id: Id102be66ede70c46ca8c34e1c929977edb4d5f57	2026-01-19 17:04:05 +01:00
eldritch horrors	f706079df3	update nixpkgs, lix pins Change-Id: I9d93a01751682473dabc5ae9def8348a35c2cef9	2025-10-10 15:30:48 +02:00
eldritch horrors	bc626f868f	fail if any non-final build is not passed the buildkite docs seem to be incomplete. `timed_out` is also a valid failure state, but this isn't mentioned anywhere in the rest api docs Change-Id: I1911dba2e87466215fbafb779dfe65c547fb41a4	2025-10-10 15:30:43 +02:00