Design 47 H2, lead answer H-A-488 and its addendum (hq branch design/47-H1-answers-59, commits c16a5376 and ecddaf35):
H-A-488 A test level's bound on a build node is the default times the number of steps the build seat runs at once (its capacity in hand at the step's start), set by the Go build module in the step's environment (MESH_TEST_LEVEL_BOUND), so a set-up that takes 35 s on a node running six steps is not a failure; the step's record says the bound it ran with. P-evidence, P-failure-loud.
H-A-488 (addendum) Every build module that runs test levels (the Node build module too) scales its level bounds the same way; the rule is the kind's, not the Go module's (fix-a-bug-met-twice). P-evidence.
Cause
Three integration steps failed today on "setup took 34-37 s over its setup bound of 30 s" on loaded build nodes (the node-engine twice, the assigner once). The fixed bounds judged the build node's load, not the tests.
Change
mesh-build-go (af817365): a level's bounds, and the MESH_TEST_LEVEL_BOUND its tests get, are their defaults times the steps the seat has in hand at the step's start (at least 1). The step's result keeps naming the bounds it was judged by (bound, setup_bound).
mesh-build-node (5326fb61): the same rule for the Node build module's levels and the MESH_TEST_LEVEL_BOUND its scripts get; the result names them (bound, setup_bound).
Quick fix (raise the 30 s default) against thorough fix (bounds scale with the load, in every build module that runs levels): thorough chosen, since this is the third failure of the kind and it recurs in both build modules.
Tests
mesh-build-go: TestSetupAndRunningTimeAreJudgedApart extended (bounds scaled by steps in hand), the unit-level and integration-level tests adjusted to the scaled bounds (test_test.go, integration_level_test.go, test_real_integration_test.go).
mesh-build-node: TestTheIntegrationLevelRunsItsScriptAndReportsSetupAndRunningApart extended with the scaled bound.
checks/*
Run locally: every check that needs no checker placement passes (agenthome, controllerdeps, deps, environment, failurekit, gitignored, gofmt, imagebuild, imageplace, locks, mailstore, nobinaries, noplacement, noscripts, particulars, removedpackages, requiredmodules, sshdir, toolchains, trunk, words). The rest need MESH_CHECK_BESIDE or the builder's module resolution (zero-version mesh packages, no go.sum entry locally) and run in this pull request's check.
Reviewed-by: mesh-code-reviewer@shanks at 5326fb6142
Pushed-by: lead-47@shanks
## Design line
Design 47 H2, lead answer H-A-488 and its addendum (hq branch design/47-H1-answers-59, commits c16a5376 and ecddaf35):
- **H-A-488** A test level's bound on a build node is the default times the number of steps the build seat runs at once (its capacity in hand at the step's start), set by the Go build module in the step's environment (MESH_TEST_LEVEL_BOUND), so a set-up that takes 35 s on a node running six steps is not a failure; the step's record says the bound it ran with. P-evidence, P-failure-loud.
- **H-A-488 (addendum)** Every build module that runs test levels (the Node build module too) scales its level bounds the same way; the rule is the kind's, not the Go module's (fix-a-bug-met-twice). P-evidence.
## Cause
Three integration steps failed today on "setup took 34-37 s over its setup bound of 30 s" on loaded build nodes (the node-engine twice, the assigner once). The fixed bounds judged the build node's load, not the tests.
## Change
- **mesh-build-go** (af817365): a level's bounds, and the MESH_TEST_LEVEL_BOUND its tests get, are their defaults times the steps the seat has in hand at the step's start (at least 1). The step's result keeps naming the bounds it was judged by (bound, setup_bound).
- **mesh-build-node** (5326fb61): the same rule for the Node build module's levels and the MESH_TEST_LEVEL_BOUND its scripts get; the result names them (bound, setup_bound).
Quick fix (raise the 30 s default) against thorough fix (bounds scale with the load, in every build module that runs levels): thorough chosen, since this is the third failure of the kind and it recurs in both build modules.
## Tests
- mesh-build-go: `TestSetupAndRunningTimeAreJudgedApart` extended (bounds scaled by steps in hand), the unit-level and integration-level tests adjusted to the scaled bounds (`test_test.go`, `integration_level_test.go`, `test_real_integration_test.go`).
- mesh-build-node: `TestTheIntegrationLevelRunsItsScriptAndReportsSetupAndRunningApart` extended with the scaled bound.
## checks/*
Run locally: every check that needs no checker placement passes (agenthome, controllerdeps, deps, environment, failurekit, gitignored, gofmt, imagebuild, imageplace, locks, mailstore, nobinaries, noplacement, noscripts, particulars, removedpackages, requiredmodules, sshdir, toolchains, trunk, words). The rest need MESH_CHECK_BESIDE or the builder's module resolution (zero-version mesh packages, no go.sum entry locally) and run in this pull request's check.
Reviewed-by: mesh-code-reviewer@shanks at 5326fb61426fa60d62a44e7604fbaa9b3c70d1f1
Pushed-by: lead-47@shanks
On a loaded build node the integration level's setup took 34-37 s against
its 30 s bound three times today: the node-engine twice, the assigner once.
The bounds were judging the node's load, not the tests. A level's bounds,
and the MESH_TEST_LEVEL_BOUND its tests get, are now their defaults times
the steps the seat has in hand at the step's start (at least 1). The step's
result keeps naming the bounds it was judged by (bound, setup_bound).
Builds: design 47 H2 H-A-488
The Node build module judged its levels by the same fixed bounds, and its
integration setup took 40 s against 30 s while three nodes' checks ran.
The rule is the kind's, not the Go module's: its bounds, and the
MESH_TEST_LEVEL_BOUND its scripts get, are now their defaults times the
steps the seat has in hand at the step's start (at least 1), and the
result names them (bound, setup_bound).
Builds: design 47 H2 H-A-488, H-A-488 addendum
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Design line
Design 47 H2, lead answer H-A-488 and its addendum (hq branch design/47-H1-answers-59, commits c16a5376 and ecddaf35):
Cause
Three integration steps failed today on "setup took 34-37 s over its setup bound of 30 s" on loaded build nodes (the node-engine twice, the assigner once). The fixed bounds judged the build node's load, not the tests.
Change
af817365): a level's bounds, and the MESH_TEST_LEVEL_BOUND its tests get, are their defaults times the steps the seat has in hand at the step's start (at least 1). The step's result keeps naming the bounds it was judged by (bound, setup_bound).5326fb61): the same rule for the Node build module's levels and the MESH_TEST_LEVEL_BOUND its scripts get; the result names them (bound, setup_bound).Quick fix (raise the 30 s default) against thorough fix (bounds scale with the load, in every build module that runs levels): thorough chosen, since this is the third failure of the kind and it recurs in both build modules.
Tests
TestSetupAndRunningTimeAreJudgedApartextended (bounds scaled by steps in hand), the unit-level and integration-level tests adjusted to the scaled bounds (test_test.go,integration_level_test.go,test_real_integration_test.go).TestTheIntegrationLevelRunsItsScriptAndReportsSetupAndRunningApartextended with the scaled bound.checks/*
Run locally: every check that needs no checker placement passes (agenthome, controllerdeps, deps, environment, failurekit, gitignored, gofmt, imagebuild, imageplace, locks, mailstore, nobinaries, noplacement, noscripts, particulars, removedpackages, requiredmodules, sshdir, toolchains, trunk, words). The rest need MESH_CHECK_BESIDE or the builder's module resolution (zero-version mesh packages, no go.sum entry locally) and run in this pull request's check.
Reviewed-by: mesh-code-reviewer@shanks at
5326fb6142Pushed-by: lead-47@shanks