Repository navigation
fix(bundle): register fails when a computer env it writes has no build a deploy can boot - #758
Merged
earakely-scale merged 3 commits intoNov 18, 2026
Conversation
…d a deploy can boot A registration whose compose of a computer env failed, a builder's disk full say, wrote the env and the tasks that deploy it and exited 0, though no deploy could boot it. With bootable, as a registration materializes, a computer env written or reused whose compose recorded builds but none ready or unverified stops it once every entity is written, before any task or eval, naming each build's status and reason. Registering again composes what it has left. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…a skipped arch builds Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…oblem quotes Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
earakely-scale
deleted the
edgararakelyan/register-fails-without-ready-build
branch
November 18, 2026 04:05
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
agent-env bundle registernow fails when a computer env it writes or reuses has no build a deploy can boot.Before, a registration whose compose failed still wrote the env and the tasks that deploy it, and exited 0. One case: a builder's disk filled while composing a desktop, leaving its container
failed. Every deploy of those tasks then failed.Now, as a registration materializes (
materialize(..., bootable=True)), each computer env is checked after its compose. That covers one written new, and one reused that composes what it has left. The registration stops when the compose recorded builds but none isreadyorunverified, that is, each failed, was skipped, or was recorded by an agent-env that composes differently. It stops once every entity is written, before any task or eval, raisingBundleError:pending), and writes the tasks once one builds.unverifiedcounts. An unverified container is probed by the first deploy where its arch runs natively (compose_unstarted). An unverified disk or VM bundle deploys as is.--no-emulatewould skip it again.Unchanged:
agent-env run,prepare_bundle,a2a-agent validateand dry runs don't passbootable. A run's deploy composes what it boots first, as before.Limits, said here rather than handled:
container/arm64ready with the emulated amd64 build failed. That passes, though a hub deploying on amd64-only providers can't boot it.computer_appenv that names a computer and driver composes its own one-app computer. If that fails, it stops the tasks even when they use the app only inside another computer.Why
Found while registering PSI's whole-computer desktops on an agent-environments-admin VM. A desktop's container build failed when the disk filled, and
bundle registerexited 0. The failure showed up only at the first deploy.Proof
register_test.py::test_a_computer_env_whose_compose_leaves_nothing_a_deploy_boots_fails_the_registration_before_its_tasks, through the CLI: exit 1, the message, the env written, no task.computer_env_folder_test.py::test_a_computer_env_no_deploy_could_boot_stops_its_tasks_until_composing_it_again_builds_one: a stalereadydoesn't count. Materializing again reuses the env, composes its four pairs, and writes the task.tst/unit/bundle,tst/unit/envandtst/unit/cli: 4,328 passed, 11 skipped. One fails,test_disk_recipe.py::test_a_windows_install_boots_its_iso_first_..., which fails onmainon this Mac too: it runs the recipe's shell with macOS's tools.ruff check src tstpasses.env buildas an alternative to registering again, thoughenv buildwrites no task, and it didn't say how a skipped arch builds.--no-emulate), with this branch and the stacked one:ubuntu-24.04andreference, copied into a bundle: exit 0,container/amd64: ready (56 s), task written.container/amd64: failed at build, exit 1, the message above, and the task not written.unchanged, the failed pair composed again, failed again, exit 1.failed at probe("its launch failed: FileNotFoundError: ... 'xterm'"), exit 1, no task. A build that fails its probe counts too.container/amd64: ready (27 s), exit 0, and the task written (broken-deploysv1).Found on the way, filed as FD-6225: registering a bundle a second time into a shared store, as the hub does, writes new versions of whatever names an installed alias, though nothing changed.
🤖 Generated with Claude Code


Confidence Score: 5/5
The PR appears safe to merge, with no outstanding findings.Summary
Registration now stops before writing tasks or evals when a computer env has recorded builds but none can boot.
Reviews (2) · Last reviewed commit: "refactor(bundle): name how much of a bui..." · Reviewed by Greptile