Skip to content

Get tests passing / CI green again#1440

Merged
maleadt merged 2 commits into
masterfrom
tb/ci
Jul 9, 2026
Merged

Get tests passing / CI green again#1440
maleadt merged 2 commits into
masterfrom
tb/ci

Conversation

@maleadt

@maleadt maleadt commented Jul 8, 2026

Copy link
Copy Markdown
Member

Locally, I was seeing a bunch of failures that seemingly didn't happen on CI. This PR try to hardens against those failures in order to get CI to pass locally.

On CI there was a hash mismatch, fixed here.

Also bump to 1.12, which is what Yggdrasil uses. The rootfs doesn't have it though, so downloading the release for now.

@maleadt maleadt changed the title Get CI green again Get tests passing / CI green again Jul 8, 2026
@maleadt maleadt requested a review from giordano July 8, 2026 12:33
@maleadt

maleadt commented Jul 8, 2026

Copy link
Copy Markdown
Member Author
/etc/profile.d/0_bb_utils.sh: line 35: set: write error: No space left on device
/etc/profile.d/0_bb_utils.sh: line 41: echo: write error: No space left on device
AnyPlatform: Test Failed at /data/staticfloat/bb_azp/rootfs/agent_3/_work/1/s/test/building.jl:460
  Expression: autobuild(build_path, "libfoo", v"1.0.0", [DirectorySource(build_tests_dir)], libfoo_cmake_script, [AnyPlatform()], libfoo_products, Dependency[])
    Expected: CompositeException
      Thrown: SystemError
      SystemError: close: No space left on device

@giordano

giordano commented Jul 8, 2026

Copy link
Copy Markdown
Member

Can we clear up the depot once and the old installation of Julia v1.7, if it's still there? I see there's a call to Pkg.gc(), but that's probably not doing much right now. This is kinda the problem of the tests in this repo: they're very storage-hungry, and redownloading all artifacts every time would take a long time

@giordano

giordano commented Jul 8, 2026

Copy link
Copy Markdown
Member

For Cargo, we may need to implement JuliaPackaging/Yggdrasil#10590 (comment)

@maleadt maleadt force-pushed the tb/ci branch 2 times, most recently from 3f8df92 to 3464515 Compare July 9, 2026 10:52
The CI agents share a single machine and disk, and recent runs died with
ENOSPC (sometimes taking the agent down mid-job, skipping Cleanup and
leaking mounts/tempdirs into subsequent runs):

- run the matrix legs serially (maxParallel: 1) so their peak disk usage
  doesn't stack;
- add a PreCleanup step that removes leftovers of runs that died before
  their Cleanup step, and immediately Pkg.gc()s unreferenced artifacts
  (mainly old compiler shards) instead of after the 7-day grace period;
- move the full shard test to the squashfs leg, where shards stay
  compressed on disk;
- print df -h in SystemInfo/PreCleanup/Cleanup for diagnosability.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@maleadt maleadt merged commit 2e95c7b into master Jul 9, 2026
10 checks passed
@maleadt maleadt deleted the tb/ci branch July 9, 2026 15:48
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants