jschoubben 979f2cd718 buildstep: keep a check ask's cache layer out of every step's reach, move through an os.Root, and layer the build cache alone ...

Review of 62a50eb: the layer lay in the workspace, which every step binds read-write,
so a change's command could swap a kept directory for a symlink and the move, run
outside the sandbox, would write into the shared cache (ADR 0541 §2). The layer now
lies beside the workspace (Workspaces.CacheLayer, removed by Release and Sweep),
Command refuses a layer within reach of the workspace, a writable directory or a
cache, and every move goes through an os.Root on the layer and keeps no symlink. Only
LayerCaches (the build cache) are layered; a module cache keeps its throwaway layer
(issue 1921).

2026-10-29 01:04:20 +01:00

..

conformance

Delete a gone process's ask key instead of ending it, so its redelivery still answers the asker (issue 1890)

2026-10-27 01:29:48 +01:00

testdata

Prove a run step reaches no key, login, name or socket and leaves the caches unchanged (issue 1309, S19, ADR 0448 check)

2026-10-22 17:46:10 +02:00

askrun_test.go

A check ask may ask the merge alone, so the checker can name every group member that does not merge in one verdict

2026-10-29 00:07:18 +01:00

askrun.go

A check ask may ask the merge alone, so the checker can name every group member that does not merge in one verdict

2026-10-29 00:07:18 +01:00

blob_test.go

Push a done build's bundle or archive to the artifact store from the build module, in one shared client (issue 1587, ADR 0498 §1-§2)

2026-10-22 17:59:14 +02:00

blob.go

Push a done build's bundle or archive to the artifact store from the build module, in one shared client (issue 1587, ADR 0498 §1-§2)

2026-10-22 17:59:14 +02:00

buildhours.go

Read build capacity as an integer and build hours as one window, default 0: no repository names a node (ADR 0567 §12, issue 1126)

2026-10-28 18:13:18 +01:00

capacity_cpu.go

Read build capacity as an integer and build hours as one window, default 0: no repository names a node (ADR 0567 §12, issue 1126)

2026-10-28 18:13:18 +01:00

capacity_test.go

Read build capacity as an integer and build hours as one window, default 0: no repository names a node (ADR 0567 §12, issue 1126)

2026-10-28 18:13:18 +01:00

capacity.go

Build puller: wake on a settings file write, retry a failed config follow at the next event, say a capacity of 0 and why

2026-10-28 20:01:46 +01:00

cpu.go

Without the composer's share a step has no thread limit; with it, the share over the module's own capacity (ADR 0553 round 2)

2026-10-26 14:39:39 +01:00

digestshape_test.go

Keep a level's digest as a list and say an end the state refuses, so a check of a repository with a package named token ends (issue 1930)

2026-10-23 23:32:00 +02:00

failure.go

Take a crashed file's cause only from what the run wrote after the last test, give a subtest's panic to that subtest, and keep the SDK's test count (issue 1901, review round 1)

2026-10-26 20:17:42 +01:00

fetch_integration_test.go

Test the fetch of one commit over SSH (issue 1474), part 6 of 7

2026-10-20 20:06:12 +00:00

fetch_test.go

Test the fetch of one commit over SSH (issue 1474), part 6 of 7

2026-10-20 20:06:12 +00:00

fetch.go

Keep git's own words in a fetch's reason: a bare "error:" line no longer stands for the forge's refusal

2026-10-26 15:26:29 +01:00

gotest.go

Take a crashed file's cause only from what the run wrote after the last test, give a subtest's panic to that subtest, and keep the SDK's test count (issue 1901, review round 1)

2026-10-26 20:17:42 +01:00

guard_test.go

Test the step runner and the toolchain guard (issues 1471, 1473), part 5 of 7

2026-10-20 20:05:49 +00:00

guard.go

Lay out a build module's long steps, guard its toolchain and fetch one commit over SSH (issues 1471, 1473, 1474), part 4 of 7

2026-10-20 20:05:21 +00:00

isolate_caches_test.go

buildstep: keep a check ask's cache layer out of every step's reach, move through an os.Root, and layer the build cache alone

2026-10-29 01:04:20 +01:00

isolate_integration_test.go

buildstep: keep a check ask's cache layer out of every step's reach, move through an os.Root, and layer the build cache alone

2026-10-29 01:04:20 +01:00

isolate_test.go

buildstep: keep a check ask's cache layer out of every step's reach, move through an os.Root, and layer the build cache alone

2026-10-29 01:04:20 +01:00

isolate.go

buildstep: keep a check ask's cache layer out of every step's reach, move through an os.Root, and layer the build cache alone

2026-10-29 01:04:20 +01:00

launchers_test.go

Share the launcher rule of a TypeScript bundle, so builder and controller name it alike (R4e, the lead's request for dev-r4e-controller)

2026-10-22 17:25:21 +02:00

launchers.go

Compose a node's own modules from their builds and read what providers elsewhere serve, so a self-composing node hands moved modules and settles served settings as the controller does (issue 1126)

2026-10-27 21:11:20 +01:00

merge_integration_test.go

MergeInto also names programs a commit of the branch added and a later one removed: the merge commit would keep the blob in history (issue 1577)

2026-10-26 19:47:17 +01:00

merge_test.go

Keep the unit level cheap: the merge's real-git tests start processes, so they are integration tests; the places tests wait 10 ms, not 50 (issue 895)

2026-10-21 19:26:02 +02:00

merge.go

MergeInto also names programs a commit of the branch added and a later one removed: the merge commit would keep the blob in history (issue 1577)

2026-10-26 19:47:17 +01:00

mirror_integration_test.go

Fetch through a bare mirror with git alternates, so a check transfers only new objects (issue 890, ADR 0536 §2)

2026-10-23 14:03:06 +02:00

mirror_test.go

Keep git's own words in a fetch's reason: a bare "error:" line no longer stands for the forge's refusal

2026-10-26 15:26:29 +01:00

narrow_test.go

S1: keep no package.json or go.mod outside the named tree in a narrowed build

2026-10-22 21:07:22 +02:00

narrow.go

S1: keep no package.json or go.mod outside the named tree in a narrowed build

2026-10-22 21:07:22 +02:00

pack_test.go

Compare Pack's tar stream with the build agent's, since gzip's bytes differ between Go versions (issue 1479)

2026-10-20 20:38:47 +00:00

pack_unix_test.go

Pack a build's output like the build agent and keep the steps state's shape (issues 1471, 1479), part 7 of 7

2026-10-20 20:06:48 +00:00

pack.go

Pack a build's output like the build agent and keep the steps state's shape (issues 1471, 1479), part 7 of 7

2026-10-20 20:06:48 +00:00

phase.go

Read build capacity as an integer and build hours as one window, default 0: no repository names a node (ADR 0567 §12, issue 1126)

2026-10-28 18:13:18 +01:00

programs_test.go

MergeInto also names programs a commit of the branch added and a later one removed: the merge commit would keep the blob in history (issue 1577)

2026-10-26 19:47:17 +01:00

programs.go

MergeInto also names programs a commit of the branch added and a later one removed: the merge commit would keep the blob in history (issue 1577)

2026-10-26 19:47:17 +01:00

pull_bus.go

Pull a build seat's asks delivery first, at most the node's capacity, each ended on its own step key, so no part chooses a node or holds a check (issue 1887 A2a, ADR 0525 §4-§6)

2026-10-23 15:15:57 +02:00

pull_capacity_test.go

Read build capacity as an integer and build hours as one window, default 0: no repository names a node (ADR 0567 §12, issue 1126)

2026-10-28 18:13:18 +01:00

pull_fake.go

Run an ask no more once its stream delivered it past AskRuns, failing its step with why, so a run that kills its node is not offered to each build node in turn (issue 1907)

2026-10-23 16:41:12 +02:00

pull_integration_test.go

Keep a long ask alive with in-progress signals and leave a stopped holder's ask to the stream, so a node that dies hands its ask on (issue 1887 A2c, ADR 0525 §6)

2026-10-23 15:17:18 +02:00

pull_seat_test.go

Pull a build seat's ask through the bundle's runtime one at a time and settle it by id, so a tools bundle pulls by its capacity and order (issue 1887 A11)

2026-10-23 15:29:05 +02:00

pull_seat.go

Pull a build seat's ask through the bundle's runtime one at a time and settle it by id, so a tools bundle pulls by its capacity and order (issue 1887 A11)

2026-10-23 15:29:05 +02:00

pull_test.go

Settings watch: arm again when the kernel drops it (IN_IGNORED), said once; the test waits on the arm, not a sleep

2026-10-28 20:08:21 +01:00

pull.go

Settings watch: arm again when the kernel drops it (IN_IGNORED), said once; the test waits on the arm, not a sleep

2026-10-28 20:08:21 +01:00

refusal.go

Run a pulled ask's whole job with a build module's own step tools, kept verdicts taken after its digest, so both build modules run asks one way and their tools stay the hand path (issue 1887 A3, ADR 0525 §3, §8)

2026-10-23 15:22:25 +02:00

request_test.go

Hand a build's fetch the named paths in a shared request, so no build module reads them from a manifest (issue 1758, ADR 0499 §4, ADR 0495 §1)

2026-10-22 17:29:58 +02:00

request.go

Hand a build's fetch the named paths in a shared request, so no build module reads them from a manifest (issue 1758, ADR 0499 §4, ADR 0495 §1)

2026-10-22 17:29:58 +02:00

seat_tools_test.go

Serve a build seat's current, pause and resume from the SDK, naming the asks running, so the three build modules hold their seats one way (issue 1887 A3node, A9a, ADR 0525 §10)

2026-10-23 15:38:03 +02:00

seat_tools.go

A build seat's capacity follows its node's build CPU, and a running ask says its phases

2026-10-26 19:52:54 +01:00

stale_test.go

Delete a gone process's ask key instead of ending it, so its redelivery still answers the asker (issue 1890)

2026-10-27 01:29:48 +01:00

stale.go

Read build capacity as an integer and build hours as one window, default 0: no repository names a node (ADR 0567 §12, issue 1126)

2026-10-28 18:13:18 +01:00

step_test.go

End a step left by a previous process at once: every step names its owner (issue 1890)

2026-10-27 01:19:29 +01:00

step.go

End a step left by a previous process at once: every step names its owner (issue 1890)

2026-10-27 01:19:29 +01:00

tap.go

Take a crashed file's cause only from what the run wrote after the last test, give a subtest's panic to that subtest, and keep the SDK's test count (issue 1901, review round 1)

2026-10-26 20:17:42 +01:00

trees_integration_test.go

Prove on real git that a Go-only commit leaves mesh-sdk's package tree equal (issue 1758, S5, ADR 0499 §2 check)

2026-10-22 17:30:25 +02:00

trees_named_test.go

S2: read a manifest for a merge's named tree as the indexer does

2026-10-22 21:07:40 +02:00

trees_test.go

Put the module trees of a merge into go/buildstep for both fetch steps, and let a preview name the conflicts of a head that does not merge (issues 1675, 1676, 1445)

2026-10-21 21:21:48 +02:00

trees.go

Keep git's own words in a fetch's reason: a bare "error:" line no longer stands for the forge's refusal

2026-10-26 15:26:29 +01:00

watch_settings_linux.go

Settings watch: arm again when the kernel drops it (IN_IGNORED), said once; the test waits on the arm, not a sleep

2026-10-28 20:08:21 +01:00

watch_settings_other.go

Build puller: wake on a settings file write, retry a failed config follow at the next event, say a capacity of 0 and why

2026-10-28 20:01:46 +01:00

workspaces_test.go

buildstep: keep a check ask's cache layer out of every step's reach, move through an os.Root, and layer the build cache alone

2026-10-29 01:04:20 +01:00

workspaces.go

buildstep: keep a check ask's cache layer out of every step's reach, move through an os.Root, and layer the build cache alone

2026-10-29 01:04:20 +01:00