looply-cli 0.5.51 → 0.5.53
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/account-connect.d.ts +2 -1
- package/dist/account-connect.d.ts.map +1 -1
- package/dist/account-connect.js.map +1 -1
- package/dist/adaptive-budget.d.ts +6 -0
- package/dist/adaptive-budget.d.ts.map +1 -1
- package/dist/adaptive-budget.js +9 -0
- package/dist/adaptive-budget.js.map +1 -1
- package/dist/autonomy-controller.d.ts.map +1 -1
- package/dist/autonomy-controller.js +19 -0
- package/dist/autonomy-controller.js.map +1 -1
- package/dist/base-check-turns.d.ts +38 -0
- package/dist/base-check-turns.d.ts.map +1 -0
- package/dist/base-check-turns.js +118 -0
- package/dist/base-check-turns.js.map +1 -0
- package/dist/bin.d.ts.map +1 -1
- package/dist/bin.js +32 -2
- package/dist/bin.js.map +1 -1
- package/dist/branch-delivery.d.ts +48 -1
- package/dist/branch-delivery.d.ts.map +1 -1
- package/dist/branch-delivery.js +203 -4
- package/dist/branch-delivery.js.map +1 -1
- package/dist/bug-repro.d.ts +89 -9
- package/dist/bug-repro.d.ts.map +1 -1
- package/dist/bug-repro.js +242 -14
- package/dist/bug-repro.js.map +1 -1
- package/dist/checkout-carry.d.ts +31 -0
- package/dist/checkout-carry.d.ts.map +1 -1
- package/dist/checkout-carry.js +98 -3
- package/dist/checkout-carry.js.map +1 -1
- package/dist/checkout-deps.d.ts +2 -0
- package/dist/checkout-deps.d.ts.map +1 -1
- package/dist/checkout-deps.js +2 -0
- package/dist/checkout-deps.js.map +1 -1
- package/dist/daemon-cmd.d.ts +1 -1
- package/dist/daemon-cmd.d.ts.map +1 -1
- package/dist/daemon-cmd.js +10 -3
- package/dist/daemon-cmd.js.map +1 -1
- package/dist/fresh-gates.d.ts.map +1 -1
- package/dist/fresh-gates.js +13 -0
- package/dist/fresh-gates.js.map +1 -1
- package/dist/gate-health.d.ts +2 -0
- package/dist/gate-health.d.ts.map +1 -1
- package/dist/gate-health.js +41 -4
- package/dist/gate-health.js.map +1 -1
- package/dist/goal-adaptive.d.ts +45 -3
- package/dist/goal-adaptive.d.ts.map +1 -1
- package/dist/goal-adaptive.js +104 -16
- package/dist/goal-adaptive.js.map +1 -1
- package/dist/goal-already-met.d.ts +62 -0
- package/dist/goal-already-met.d.ts.map +1 -0
- package/dist/goal-already-met.js +95 -0
- package/dist/goal-already-met.js.map +1 -0
- package/dist/goal-execution-evidence.d.ts +7 -1
- package/dist/goal-execution-evidence.d.ts.map +1 -1
- package/dist/goal-execution-evidence.js +12 -1
- package/dist/goal-execution-evidence.js.map +1 -1
- package/dist/goal-gap.d.ts +48 -5
- package/dist/goal-gap.d.ts.map +1 -1
- package/dist/goal-gap.js +103 -13
- package/dist/goal-gap.js.map +1 -1
- package/dist/integration-deps.d.ts.map +1 -1
- package/dist/integration-deps.js +18 -0
- package/dist/integration-deps.js.map +1 -1
- package/dist/machine-settings.d.ts +9 -0
- package/dist/machine-settings.d.ts.map +1 -1
- package/dist/machine-settings.js +35 -2
- package/dist/machine-settings.js.map +1 -1
- package/dist/output-gist.d.ts +40 -0
- package/dist/output-gist.d.ts.map +1 -0
- package/dist/output-gist.js +107 -0
- package/dist/output-gist.js.map +1 -0
- package/dist/plan-recovery.d.ts +14 -3
- package/dist/plan-recovery.d.ts.map +1 -1
- package/dist/plan-recovery.js +24 -6
- package/dist/plan-recovery.js.map +1 -1
- package/dist/provider-accounts.d.ts +0 -6
- package/dist/provider-accounts.d.ts.map +1 -1
- package/dist/provider-accounts.js +0 -6
- package/dist/provider-accounts.js.map +1 -1
- package/dist/pursuit-ending.d.ts.map +1 -1
- package/dist/pursuit-ending.js +11 -0
- package/dist/pursuit-ending.js.map +1 -1
- package/dist/quarantine-report.d.ts +32 -0
- package/dist/quarantine-report.d.ts.map +1 -0
- package/dist/quarantine-report.js +75 -0
- package/dist/quarantine-report.js.map +1 -0
- package/dist/round-runner.d.ts +13 -2
- package/dist/round-runner.d.ts.map +1 -1
- package/dist/round-runner.js +5 -1
- package/dist/round-runner.js.map +1 -1
- package/dist/run-cmd.d.ts.map +1 -1
- package/dist/run-cmd.js +84 -11
- package/dist/run-cmd.js.map +1 -1
- package/dist/self-update.d.ts.map +1 -1
- package/dist/self-update.js +5 -15
- package/dist/self-update.js.map +1 -1
- package/dist/serve-cmd.d.ts +9 -0
- package/dist/serve-cmd.d.ts.map +1 -1
- package/dist/serve-cmd.js +45 -1
- package/dist/serve-cmd.js.map +1 -1
- package/dist/serve-log.d.ts +52 -0
- package/dist/serve-log.d.ts.map +1 -0
- package/dist/serve-log.js +146 -0
- package/dist/serve-log.js.map +1 -0
- package/dist/serve.d.ts +7 -7
- package/dist/serve.d.ts.map +1 -1
- package/dist/serve.js +11 -15
- package/dist/serve.js.map +1 -1
- package/dist/server-client.d.ts +14 -2
- package/dist/server-client.d.ts.map +1 -1
- package/dist/server-client.js +24 -2
- package/dist/server-client.js.map +1 -1
- package/dist/setup-check-wiring.d.ts +8 -0
- package/dist/setup-check-wiring.d.ts.map +1 -1
- package/dist/setup-check-wiring.js +78 -22
- package/dist/setup-check-wiring.js.map +1 -1
- package/dist/setup-check.d.ts +4 -0
- package/dist/setup-check.d.ts.map +1 -1
- package/dist/setup-check.js +70 -3
- package/dist/setup-check.js.map +1 -1
- package/dist/steady-log.d.ts +48 -0
- package/dist/steady-log.d.ts.map +1 -0
- package/dist/steady-log.js +100 -0
- package/dist/steady-log.js.map +1 -0
- package/dist/verify-cmd.d.ts +0 -2
- package/dist/verify-cmd.d.ts.map +1 -1
- package/dist/verify-cmd.js +0 -5
- package/dist/verify-cmd.js.map +1 -1
- package/dist/workspace-builds.d.ts +153 -0
- package/dist/workspace-builds.d.ts.map +1 -0
- package/dist/workspace-builds.js +347 -0
- package/dist/workspace-builds.js.map +1 -0
- package/dist/workspace-quarantine.d.ts +36 -2
- package/dist/workspace-quarantine.d.ts.map +1 -1
- package/dist/workspace-quarantine.js +53 -0
- package/dist/workspace-quarantine.js.map +1 -1
- package/dist/workspace.d.ts.map +1 -1
- package/dist/workspace.js +11 -3
- package/dist/workspace.js.map +1 -1
- package/package.json +2 -2
- package/win/contained-node.cjs +110 -1
- package/dist/.src-hash +0 -1
package/dist/bug-repro.d.ts
CHANGED
|
@@ -14,11 +14,27 @@ import type { GoalGapOutcome } from './goal-gap';
|
|
|
14
14
|
* stays in the suite as the regression guard. The pull request and the ticket
|
|
15
15
|
* say "reproduced: <test>, failed before, passes now".
|
|
16
16
|
*
|
|
17
|
-
*
|
|
18
|
-
*
|
|
19
|
-
*
|
|
20
|
-
*
|
|
21
|
-
*
|
|
17
|
+
* 🔴 WHEN THE BUG CANNOT BE SHOWN, LOOPLY GOES ON — IT DOES NOT STOP TO ASK (2026-11-16).
|
|
18
|
+
*
|
|
19
|
+
* It used to ask ONE question on the ticket before spending any fix attempt, whenever the test writer could not
|
|
20
|
+
* tell exactly where the bug was, or its test passed on the current code. Read back over three weeks, those
|
|
21
|
+
* questions were "What did you expect manual mode to do differently?" (TAB-B-101), "What went wrong with the All
|
|
22
|
+
* Payments refund, or is the request just to remove it?" (TAB-B-105), "Which part of the Risk Chamber screen was
|
|
23
|
+
* unclear?" (RBB-B-251, asked twice) — each on a ticket a colleague would simply have had a go at, and each left
|
|
24
|
+
* the ticket waiting days for its reporter. The owner's bar is that a ticket pasted into a chat gets fixed nine
|
|
25
|
+
* times in ten; a question before the first attempt is the opposite of that.
|
|
26
|
+
*
|
|
27
|
+
* - THE REPORT CAN BE READ MORE THAN ONE WAY: the test writer says which reading it would act on (`assume`), and
|
|
28
|
+
* the work goes on to that reading. It travels with the goal's words to the plan, the rounds and the goal check,
|
|
29
|
+
* and the goal check repeats it on the ticket ("Looply read the request as … — reply if you meant something
|
|
30
|
+
* else"), so the reporter corrects it in one reply instead of answering a question first.
|
|
31
|
+
* - THE TEST PASSES ON THE CURRENT CODE: the bug did not show there. It may already be fixed, or lie where the
|
|
32
|
+
* test did not look. The plan is told exactly that and finds out which; "already fixed" then ends as its own
|
|
33
|
+
* checked answer (`goal-already-met.ts`), not as a question.
|
|
34
|
+
*
|
|
35
|
+
* A question is still asked, once, in the one case nothing else can cover: the test writer could form NO reading —
|
|
36
|
+
* the report breaks off mid-sentence, or names nothing it can find in this repository. A goal that already carries
|
|
37
|
+
* the person's answer never asks again.
|
|
22
38
|
*
|
|
23
39
|
* WHAT IS NEVER A REASON TO ASK. A command that could not run (a missing tool,
|
|
24
40
|
* a sandbox denial, a timeout), an agent that failed, or a repository with no
|
|
@@ -55,9 +71,12 @@ export type ReproAnswer = {
|
|
|
55
71
|
file: string;
|
|
56
72
|
test: string;
|
|
57
73
|
command: string;
|
|
58
|
-
}
|
|
74
|
+
}
|
|
75
|
+
/** `assume`: the reading the test writer would act on. With it, nothing is asked: the work goes on to that reading. */
|
|
76
|
+
| {
|
|
59
77
|
kind: 'question';
|
|
60
78
|
question: string;
|
|
79
|
+
assume?: string;
|
|
61
80
|
} | {
|
|
62
81
|
kind: 'skip';
|
|
63
82
|
because: string;
|
|
@@ -79,19 +98,64 @@ export type RunJudgement = {
|
|
|
79
98
|
} | {
|
|
80
99
|
verdict: 'green';
|
|
81
100
|
}
|
|
82
|
-
/**
|
|
101
|
+
/**
|
|
102
|
+
* It did not run the test: a timeout, a runner that could not start, or a failure that never mentions it.
|
|
103
|
+
* `service` is set when what was missing is a service the test suite needs and the sandbox cannot provide.
|
|
104
|
+
*/
|
|
83
105
|
| {
|
|
84
106
|
verdict: 'not-run';
|
|
85
107
|
because: string;
|
|
108
|
+
service?: string;
|
|
86
109
|
};
|
|
110
|
+
/** A service a test suite needed and Looply's sandbox could not give it. */
|
|
111
|
+
export interface MissingService {
|
|
112
|
+
/** What it is, as a person would say it: "a PostgreSQL database". */
|
|
113
|
+
service: string;
|
|
114
|
+
/** Why it could not be had here, in one plain clause. */
|
|
115
|
+
detail: string;
|
|
116
|
+
}
|
|
117
|
+
/**
|
|
118
|
+
* 🔴 A TEST SUITE THAT NEEDS A SERVICE THIS SANDBOX DOES NOT PROVIDE.
|
|
119
|
+
*
|
|
120
|
+
* Measured 2026-11-16 in a real gate container (`network: 'none'`): a server on 127.0.0.1 and a client in
|
|
121
|
+
* the same sandbox, even in another process, talk normally; an outbound connection is refused (EACCES).
|
|
122
|
+
* So a database the suite starts for itself would be reachable — but PostgreSQL cannot start: on Windows
|
|
123
|
+
* it emulates signals through a named pipe, `\\.\pipe\pgsignal_<pid>`, and an AppContainer may create
|
|
124
|
+
* pipes only under `\\.\pipe\LOCAL\`. `initdb` retries and gets through; the server itself stops with
|
|
125
|
+
* "FATAL: could not create signal listener pipe … error code 5", and Tabsy's jest `globalSetup`
|
|
126
|
+
* (embedded-postgres) then waits for a server that is gone until the step's time runs out.
|
|
127
|
+
*
|
|
128
|
+
* That run is not the bug failing and not "no tests": it is the machine missing something the tests
|
|
129
|
+
* need. It gets its own words, so the plan and the goal check are told the change is judged without
|
|
130
|
+
* running these tests, rather than spending plans on a test that cannot start here. Returns null when
|
|
131
|
+
* the output shows no such thing.
|
|
132
|
+
*/
|
|
133
|
+
export declare function missingService(output: string): MissingService | null;
|
|
134
|
+
/** The plain sentence for a run that needed a missing service: the change is not what failed. */
|
|
135
|
+
export declare function missingServiceSentence(m: MissingService): string;
|
|
136
|
+
/**
|
|
137
|
+
* 🔴 A RUN THAT RAN NO TEST (TAB-B-66, -86, -91, -95, -100; 2026-10-18 → 11-11).
|
|
138
|
+
*
|
|
139
|
+
* Inside the sandbox Tabsy's jest could not see its own test roots and printed
|
|
140
|
+
* "No tests found, exiting with code 1 … Pattern: test\\pay-at-table\\x.spec.ts - 0 matches".
|
|
141
|
+
* That output names the test file, so it was judged RED: "reproduced" on the
|
|
142
|
+
* base tree, and "still fails" at every delivered tree after the fix was in —
|
|
143
|
+
* six goals re-planned to the limit over a test that never ran once. A runner
|
|
144
|
+
* saying it found or ran nothing is the machine's problem, never the bug's:
|
|
145
|
+
* neither red nor green. Returns what the runner said, or null.
|
|
146
|
+
*/
|
|
147
|
+
export declare function ranNoTests(output: string): string | null;
|
|
87
148
|
/**
|
|
88
149
|
* What one run of the test says. RED needs more than a non-zero exit: the
|
|
89
150
|
* output must name the test or its file, or the failure could be a missing
|
|
90
151
|
* tool or a syntax error elsewhere — a red that shows nothing about the bug.
|
|
152
|
+
* A run that ran no test at all is neither red nor green (`ranNoTests`).
|
|
91
153
|
*/
|
|
92
154
|
export declare function judgeRun(result: CommandResult, repro: {
|
|
93
155
|
test: string;
|
|
94
156
|
file: string;
|
|
157
|
+
}, opts?: {
|
|
158
|
+
ranHereBefore?: boolean;
|
|
95
159
|
}): RunJudgement;
|
|
96
160
|
export interface ReproRun extends ReproductionRun {
|
|
97
161
|
/** The end of the run's output, for a person reading the record. */
|
|
@@ -104,13 +168,24 @@ export interface ReproRecord {
|
|
|
104
168
|
at: string;
|
|
105
169
|
/**
|
|
106
170
|
* reproduced — the test ran red on the base tree; the fix must turn it green
|
|
107
|
-
* question — the
|
|
108
|
-
* not-reproduced — the bug
|
|
171
|
+
* question — no reading of the report was possible; a person is asked before any fix (rare)
|
|
172
|
+
* not-reproduced — the bug was not shown, and the work goes on without a reproduction: on a stated reading of
|
|
173
|
+
* the report (`assumed`), because the test passed on the current code (`passedOnBase`), or
|
|
174
|
+
* because the person already answered once
|
|
109
175
|
* skipped — no reproduction could be attempted (the machine's problem, not the person's)
|
|
110
176
|
*/
|
|
111
177
|
outcome: 'reproduced' | 'question' | 'not-reproduced' | 'skipped';
|
|
112
178
|
because?: string;
|
|
179
|
+
/**
|
|
180
|
+
* Set on a `skipped` record when the test could not run because the suite needs a service the sandbox
|
|
181
|
+
* cannot provide ("a PostgreSQL database"). The plan and the goal check are told so (`reproductionBrief`).
|
|
182
|
+
*/
|
|
183
|
+
needs?: string;
|
|
113
184
|
question?: string;
|
|
185
|
+
/** The reading of the report the work goes on to, when it allowed more than one (see the header). */
|
|
186
|
+
assumed?: string;
|
|
187
|
+
/** The test written from the report passed on the base tree: already fixed, or the bug lies where it did not look. */
|
|
188
|
+
passedOnBase?: true;
|
|
114
189
|
test?: string;
|
|
115
190
|
file?: string;
|
|
116
191
|
command?: string;
|
|
@@ -175,6 +250,11 @@ export interface ReproInput {
|
|
|
175
250
|
* never reset — that work is somebody's).
|
|
176
251
|
*/
|
|
177
252
|
export declare function reproduceBug(input: ReproInput, deps: ReproDeps): Promise<ReproRecord>;
|
|
253
|
+
/**
|
|
254
|
+
* What a bug that was NOT shown adds to the goal's words: the reading the work goes on to, or that a test written from
|
|
255
|
+
* the report already passes. '' for anything else (including a record written before these fields existed).
|
|
256
|
+
*/
|
|
257
|
+
export declare function unreproducedBrief(r: ReproRecord | null): string;
|
|
178
258
|
/** The block a reproduced bug adds to the goal's words, or '' for any other outcome. */
|
|
179
259
|
export declare function reproductionBrief(r: ReproRecord | null): string;
|
|
180
260
|
/** The goal's words with the reproduction attached, for the plan, the goal check and the rounds. */
|
package/dist/bug-repro.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"bug-repro.d.ts","sourceRoot":"","sources":["../src/bug-repro.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,kBAAkB,EAAE,eAAe,EAAE,MAAM,cAAc,CAAC;AACxE,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,cAAc,CAAC;AAEhD,OAAO,KAAK,EAAE,aAAa,EAAE,aAAa,EAAE,MAAM,QAAQ,CAAC;AAC3D,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,YAAY,CAAC;AAEjD
|
|
1
|
+
{"version":3,"file":"bug-repro.d.ts","sourceRoot":"","sources":["../src/bug-repro.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,kBAAkB,EAAE,eAAe,EAAE,MAAM,cAAc,CAAC;AACxE,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,cAAc,CAAC;AAEhD,OAAO,KAAK,EAAE,aAAa,EAAE,aAAa,EAAE,MAAM,QAAQ,CAAC;AAC3D,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,YAAY,CAAC;AAEjD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2CG;AAEH,iGAAiG;AACjG,eAAO,MAAM,sBAAsB,QAAgB,CAAC;AACpD,2BAA2B;AAC3B,eAAO,MAAM,oBAAoB,QAAgB,CAAC;AAClD,yEAAyE;AACzE,eAAO,MAAM,qBAAqB,QAAS,CAAC;AAC5C,iGAAiG;AACjG,eAAO,MAAM,gBAAgB,mBAAmB,CAAC;AAOjD,4FAA4F;AAC5F,eAAO,MAAM,sBAAsB,2JACuH,CAAC;AAI3J;;;;GAIG;AACH,wBAAgB,SAAS,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAE/C;AAED,+GAA+G;AAC/G,wBAAgB,iBAAiB,CAAC,IAAI,EAAE,MAAM,GAAG,MAAM,EAAE,CAcxD;AAED,mGAAmG;AACnG,wBAAgB,iBAAiB,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAEvD;AAID,MAAM,MAAM,WAAW,GACnB;IAAE,IAAI,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAA;CAAE;AAC/D,uHAAuH;GACrH;IAAE,IAAI,EAAE,UAAU,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,MAAM,CAAC,EAAE,MAAM,CAAA;CAAE,GACvD;IAAE,IAAI,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAA;CAAE,CAAC;AAEtC,wBAAgB,gBAAgB,CAAC,KAAK,EAAE;IACtC,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,EAAE,MAAM,CAAC;IACjB,KAAK,EAAE,aAAa,CAAC;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;IACpD,8EAA8E;IAC9E,WAAW,EAAE,SAAS,MAAM,EAAE,CAAC;CAChC,GAAG,MAAM,CAqCT;AAED,wGAAwG;AACxG,wBAAgB,kBAAkB,CAAC,UAAU,EAAE,MAAM,GAAG,WAAW,GAAG,IAAI,CAiCzE;AAQD,MAAM,MAAM,YAAY,GACpB;IAAE,OAAO,EAAE,KAAK,CAAA;CAAE,GAClB;IAAE,OAAO,EAAE,OAAO,CAAA;CAAE;AACtB;;;GAGG;GACD;IAAE,OAAO,EAAE,SAAS,CAAC;IAAC,OAAO,EAAE,MAAM,CAAC;IAAC,OAAO,CAAC,EAAE,MAAM,CAAA;CAAE,CAAC;AAE9D,4EAA4E;AAC5E,MAAM,WAAW,cAAc;IAC7B,qEAAqE;IACrE,OAAO,EAAE,MAAM,CAAC;IAChB,yDAAyD;IACzD,MAAM,EAAE,MAAM,CAAC;CAChB;AAaD;;;;;;;;;;;;;;;GAeG;AACH,wBAAgB,cAAc,CAAC,MAAM,EAAE,MAAM,GAAG,cAAc,GAAG,IAAI,CAiCpE;AAED,iGAAiG;AACjG,wBAAgB,sBAAsB,CAAC,CAAC,EAAE,cAAc,GAAG,MAAM,CAEhE;AAED;;;;;;;;;;GAUG;AACH,wBAAgB,UAAU,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,GAAG,IAAI,CAWxD;AAED;;;;;GAKG;AACH,wBAAgB,QAAQ,CAAC,MAAM,EAAE,aAAa,EAAE,KAAK,EAAE;IAAE,IAAI,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,MAAM,CAAA;CAAE,EAAE,IAAI,GAAE;IAAE,aAAa,CAAC,EAAE,OAAO,CAAA;CAAO,GAAG,YAAY,CA+B3I;AASD,MAAM,WAAW,QAAS,SAAQ,eAAe;IAC/C,oEAAoE;IACpE,IAAI,EAAE,MAAM,CAAC;CACd;AAED,MAAM,WAAW,WAAW;IAC1B,MAAM,EAAE,CAAC,CAAC;IACV,MAAM,EAAE,MAAM,CAAC;IACf,WAAW,EAAE,MAAM,CAAC;IACpB,EAAE,EAAE,MAAM,CAAC;IACX;;;;;;;OAOG;IACH,OAAO,EAAE,YAAY,GAAG,UAAU,GAAG,gBAAgB,GAAG,SAAS,CAAC;IAClE,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB;;;OAGG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,qGAAqG;IACrG,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,sHAAsH;IACtH,YAAY,CAAC,EAAE,IAAI,CAAC;IACpB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,sFAAsF;IACtF,KAAK,CAAC,EAAE,KAAK,CAAC;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,OAAO,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;IACjD,YAAY,CAAC,EAAE,QAAQ,CAAC;IACxB,SAAS,CAAC,EAAE,QAAQ,GAAG,IAAI,CAAC;IAC5B,uEAAuE;IACvE,SAAS,CAAC,EAAE,QAAQ,GAAG;QAAE,OAAO,EAAE,KAAK,GAAG,SAAS,GAAG,SAAS,CAAA;KAAE,CAAC;CACnE;AAaD,iHAAiH;AACjH,qBAAa,UAAU;IACT,OAAO,CAAC,QAAQ,CAAC,QAAQ;gBAAR,QAAQ,EAAE,MAAM;IAE7C,OAAO,CAAC,GAAG;IAIX,OAAO,CAAC,IAAI;IAIZ,IAAI,CAAC,MAAM,EAAE,MAAM,EAAE,WAAW,EAAE,MAAM,GAAG,WAAW,GAAG,IAAI;IAS7D,0EAA0E;IAC1E,MAAM,CAAC,MAAM,EAAE,MAAM,GAAG,WAAW,GAAG,IAAI;IAiB1C,KAAK,CAAC,CAAC,EAAE,WAAW,GAAG,IAAI;CAO5B;AAID,MAAM,WAAW,SAAS;IACxB,4EAA4E;IAC5E,MAAM,EAAE,WAAW,CAAC;IACpB,2EAA2E;IAC3E,GAAG,EAAE,aAAa,CAAC;IACnB,GAAG,EAAE;QACH,cAAc,IAAI,OAAO,CAAC,KAAK,CAAC;YAAE,IAAI,EAAE,MAAM,CAAC;YAAC,MAAM,EAAE,MAAM,CAAA;SAAE,CAAC,CAAC,CAAC;QACnE,OAAO,IAAI,OAAO,CAAC,MAAM,CAAC,CAAC;QAC3B,sFAAsF;QACtF,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;KAC1B,CAAC;IACF,kDAAkD;IAClD,QAAQ,CAAC,IAAI,EAAE,MAAM,GAAG,MAAM,GAAG,IAAI,CAAC;IACtC,iHAAiH;IACjH,QAAQ,CAAC,CAAC,GAAG,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,GAAG,IAAI,CAAC,CAAC;IAC9D,GAAG,IAAI,IAAI,CAAC;IACZ,GAAG,CAAC,IAAI,EAAE,MAAM,GAAG,IAAI,CAAC;CACzB;AAED,MAAM,WAAW,UAAU;IACzB,MAAM,EAAE,MAAM,CAAC;IACf,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,EAAE,MAAM,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,KAAK,EAAE,aAAa,CAAC;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;CACrD;AAED;;;;;GAKG;AACH,wBAAsB,YAAY,CAAC,KAAK,EAAE,UAAU,EAAE,IAAI,EAAE,SAAS,GAAG,OAAO,CAAC,WAAW,CAAC,CAsG3F;AAID;;;GAGG;AACH,wBAAgB,iBAAiB,CAAC,CAAC,EAAE,WAAW,GAAG,IAAI,GAAG,MAAM,CAyB/D;AAED,wFAAwF;AACxF,wBAAgB,iBAAiB,CAAC,CAAC,EAAE,WAAW,GAAG,IAAI,GAAG,MAAM,CAyB/D;AAkBD,oGAAoG;AACpG,wBAAgB,gBAAgB,CAAC,QAAQ,EAAE,MAAM,EAAE,CAAC,EAAE,WAAW,GAAG,IAAI,GAAG,MAAM,CAEhF;AAED,mFAAmF;AACnF,wBAAgB,aAAa,CAAC,CAAC,EAAE,WAAW,GAAG,MAAM,CAEpD;AAID,MAAM,MAAM,UAAU,GAClB;IAAE,KAAK,EAAE,OAAO,CAAC;IAAC,GAAG,EAAE,QAAQ,CAAA;CAAE,GACjC;IAAE,KAAK,EAAE,KAAK,CAAC;IAAC,GAAG,EAAE,QAAQ,CAAA;CAAE;AACjC,iDAAiD;GAC/C;IAAE,KAAK,EAAE,SAAS,CAAC;IAAC,OAAO,EAAE,MAAM,CAAA;CAAE,GACrC;IAAE,KAAK,EAAE,SAAS,CAAC;IAAC,OAAO,EAAE,MAAM,CAAC;IAAC,GAAG,EAAE,QAAQ,CAAA;CAAE,CAAC;AAEzD,+DAA+D;AAC/D,wBAAsB,eAAe,CACnC,CAAC,EAAE,WAAW,EACd,IAAI,EAAE,IAAI,CAAC,SAAS,EAAE,KAAK,GAAG,UAAU,GAAG,KAAK,CAAC,GAAG;IAAE,OAAO,IAAI,OAAO,CAAC,MAAM,CAAC,CAAA;CAAE,GACjF,OAAO,CAAC,UAAU,CAAC,CAsBrB;AAED;;;;;;;;;;;;GAYG;AACH,wBAAsB,wBAAwB,CAC5C,OAAO,EAAE,cAAc,EACvB,CAAC,EAAE,WAAW,GAAG,IAAI,EACrB,IAAI,EAAE,IAAI,CAAC,SAAS,EAAE,KAAK,GAAG,UAAU,GAAG,KAAK,GAAG,KAAK,CAAC,GAAG;IAC1D,OAAO,IAAI,OAAO,CAAC,MAAM,CAAC,CAAC;IAC3B,YAAY,IAAI,OAAO,CAAC,SAAS,MAAM,EAAE,CAAC,CAAC;IAC3C,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;IACzB,KAAK,EAAE,IAAI,CAAC,UAAU,EAAE,OAAO,CAAC,CAAC;CAClC,GACA,OAAO,CAAC,cAAc,CAAC,CAmCzB;AAYD,sGAAsG;AACtG,wBAAgB,kBAAkB,CAAC,CAAC,EAAE,WAAW,GAAG,IAAI,GAAG,kBAAkB,GAAG,IAAI,CAOnF;AAID,2DAA2D;AAC3D,wBAAgB,cAAc,CAAC,GAAG,EAAE,MAAM,GAAG,CAAC,IAAI,EAAE,MAAM,KAAK,MAAM,GAAG,IAAI,CAS3E;AAYD;;;;;GAKG;AACH,wBAAgB,wBAAwB,CACtC,GAAG,EAAE,MAAM,EACX,OAAO,GAAE,OAAO,KAAa,GAC5B,CAAC,GAAG,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,KAAK,OAAO,CAAC,MAAM,GAAG,IAAI,CAAC,CAyBxD"}
|
package/dist/bug-repro.js
CHANGED
|
@@ -6,8 +6,12 @@ exports.ticketAttachments = ticketAttachments;
|
|
|
6
6
|
exports.ticketWasAnswered = ticketWasAnswered;
|
|
7
7
|
exports.buildReproPrompt = buildReproPrompt;
|
|
8
8
|
exports.extractReproAnswer = extractReproAnswer;
|
|
9
|
+
exports.missingService = missingService;
|
|
10
|
+
exports.missingServiceSentence = missingServiceSentence;
|
|
11
|
+
exports.ranNoTests = ranNoTests;
|
|
9
12
|
exports.judgeRun = judgeRun;
|
|
10
13
|
exports.reproduceBug = reproduceBug;
|
|
14
|
+
exports.unreproducedBrief = unreproducedBrief;
|
|
11
15
|
exports.reproductionBrief = reproductionBrief;
|
|
12
16
|
exports.withReproduction = withReproduction;
|
|
13
17
|
exports.reproQuestion = reproQuestion;
|
|
@@ -31,11 +35,27 @@ const agent_envelope_1 = require("./agent-envelope");
|
|
|
31
35
|
* stays in the suite as the regression guard. The pull request and the ticket
|
|
32
36
|
* say "reproduced: <test>, failed before, passes now".
|
|
33
37
|
*
|
|
34
|
-
*
|
|
35
|
-
*
|
|
36
|
-
*
|
|
37
|
-
*
|
|
38
|
-
*
|
|
38
|
+
* 🔴 WHEN THE BUG CANNOT BE SHOWN, LOOPLY GOES ON — IT DOES NOT STOP TO ASK (2026-11-16).
|
|
39
|
+
*
|
|
40
|
+
* It used to ask ONE question on the ticket before spending any fix attempt, whenever the test writer could not
|
|
41
|
+
* tell exactly where the bug was, or its test passed on the current code. Read back over three weeks, those
|
|
42
|
+
* questions were "What did you expect manual mode to do differently?" (TAB-B-101), "What went wrong with the All
|
|
43
|
+
* Payments refund, or is the request just to remove it?" (TAB-B-105), "Which part of the Risk Chamber screen was
|
|
44
|
+
* unclear?" (RBB-B-251, asked twice) — each on a ticket a colleague would simply have had a go at, and each left
|
|
45
|
+
* the ticket waiting days for its reporter. The owner's bar is that a ticket pasted into a chat gets fixed nine
|
|
46
|
+
* times in ten; a question before the first attempt is the opposite of that.
|
|
47
|
+
*
|
|
48
|
+
* - THE REPORT CAN BE READ MORE THAN ONE WAY: the test writer says which reading it would act on (`assume`), and
|
|
49
|
+
* the work goes on to that reading. It travels with the goal's words to the plan, the rounds and the goal check,
|
|
50
|
+
* and the goal check repeats it on the ticket ("Looply read the request as … — reply if you meant something
|
|
51
|
+
* else"), so the reporter corrects it in one reply instead of answering a question first.
|
|
52
|
+
* - THE TEST PASSES ON THE CURRENT CODE: the bug did not show there. It may already be fixed, or lie where the
|
|
53
|
+
* test did not look. The plan is told exactly that and finds out which; "already fixed" then ends as its own
|
|
54
|
+
* checked answer (`goal-already-met.ts`), not as a question.
|
|
55
|
+
*
|
|
56
|
+
* A question is still asked, once, in the one case nothing else can cover: the test writer could form NO reading —
|
|
57
|
+
* the report breaks off mid-sentence, or names nothing it can find in this repository. A goal that already carries
|
|
58
|
+
* the person's answer never asks again.
|
|
39
59
|
*
|
|
40
60
|
* WHAT IS NEVER A REASON TO ASK. A command that could not run (a missing tool,
|
|
41
61
|
* a sandbox denial, a timeout), an agent that failed, or a repository with no
|
|
@@ -58,6 +78,8 @@ exports.TICKET_FILES_DIR = '.looply-ticket';
|
|
|
58
78
|
const MAX_ATTACHMENTS = 4;
|
|
59
79
|
const MAX_ATTACHMENT_BYTES = 5 * 1024 * 1024;
|
|
60
80
|
const MAX_QUESTION = 200;
|
|
81
|
+
/** The reading the test writer would act on, when the report allows more than one. One sentence for the ticket. */
|
|
82
|
+
const MAX_ASSUMED = 300;
|
|
61
83
|
/** The question asked when the bug did not show and the agent did not word a better one. */
|
|
62
84
|
exports.DEFAULT_REPRO_QUESTION = "I couldn't make this bug happen. Which page or table is it on, what exactly did you click or enter, and what did you see instead of what you expected?";
|
|
63
85
|
// ── reading the goal's words ──
|
|
@@ -119,8 +141,13 @@ function buildReproPrompt(input) {
|
|
|
119
141
|
'',
|
|
120
142
|
'ANSWER with one JSON object on a single line and nothing after it, one of:',
|
|
121
143
|
'{"test":{"file":"<repository path of the test file>","name":"<the test name as written>","command":"<runs only this test>"}}',
|
|
122
|
-
'{"question":"<one short question
|
|
123
|
-
'
|
|
144
|
+
'{"question":"<what is unclear, as one short question>","assume":"<the reading you would act on>"} — only when',
|
|
145
|
+
' you cannot tell WHERE the bug is or WHAT goes wrong well enough to write a test. ALWAYS give "assume": the most',
|
|
146
|
+
' reasonable reading of the report, as one plain sentence saying where and what (for example "the Refund button',
|
|
147
|
+
' on the All Payments page can be pressed twice for the same payment"). Looply then works to that reading and',
|
|
148
|
+
' tells the reporter, who can correct it in one reply — nobody is made to wait for an answer first. Leave',
|
|
149
|
+
' "assume" out ONLY when no reading is possible: the report breaks off mid-sentence, or names nothing you can',
|
|
150
|
+
' find in this repository at all. Each under 200 characters.',
|
|
124
151
|
'{"skip":"<why>"} — when this bug cannot be shown by a test in this repository (for example it is only visual).',
|
|
125
152
|
].join('\n');
|
|
126
153
|
}
|
|
@@ -146,13 +173,17 @@ function extractReproAnswer(transcript) {
|
|
|
146
173
|
const file = typeof r['file'] === 'string' ? r['file'].trim().replace(/\\/g, '/') : '';
|
|
147
174
|
const name = typeof r['name'] === 'string' ? r['name'].trim() : '';
|
|
148
175
|
const command = typeof r['command'] === 'string' ? r['command'].trim() : '';
|
|
149
|
-
|
|
176
|
+
// A path longer than the delivery report allows (core `validateReproduction`) is no answer: sent on, it would
|
|
177
|
+
// have the server refuse every report of this goal's delivery.
|
|
178
|
+
if (file && file.length <= 300 && name && command && !file.startsWith('/') && !/^[A-Za-z]:/.test(file) && !file.split('/').includes('..')) {
|
|
150
179
|
return { kind: 'test', file, test: name.slice(0, 200), command: command.slice(0, 400) };
|
|
151
180
|
}
|
|
152
181
|
continue;
|
|
153
182
|
}
|
|
154
183
|
if (typeof o['question'] === 'string' && o['question'].trim()) {
|
|
155
|
-
|
|
184
|
+
const question = oneLine(o['question']).slice(0, MAX_QUESTION);
|
|
185
|
+
const assume = typeof o['assume'] === 'string' ? oneLine(o['assume']).slice(0, MAX_ASSUMED) : '';
|
|
186
|
+
return assume === '' ? { kind: 'question', question } : { kind: 'question', question, assume };
|
|
156
187
|
}
|
|
157
188
|
if (typeof o['skip'] === 'string')
|
|
158
189
|
return { kind: 'skip', because: oneLine(o['skip']).slice(0, 300) || 'no reason given' };
|
|
@@ -162,16 +193,131 @@ function extractReproAnswer(transcript) {
|
|
|
162
193
|
function oneLine(s) {
|
|
163
194
|
return s.replace(/\s+/g, ' ').trim();
|
|
164
195
|
}
|
|
196
|
+
/** Well-known ports, for naming the server a refused connection was meant for. */
|
|
197
|
+
const SERVICE_PORTS = {
|
|
198
|
+
'5432': 'a PostgreSQL database',
|
|
199
|
+
'3306': 'a MySQL database',
|
|
200
|
+
'1433': 'a SQL Server database',
|
|
201
|
+
'27017': 'a MongoDB database',
|
|
202
|
+
'6379': 'a Redis server',
|
|
203
|
+
'5672': 'a RabbitMQ server',
|
|
204
|
+
'9200': 'an Elasticsearch server',
|
|
205
|
+
};
|
|
206
|
+
/**
|
|
207
|
+
* 🔴 A TEST SUITE THAT NEEDS A SERVICE THIS SANDBOX DOES NOT PROVIDE.
|
|
208
|
+
*
|
|
209
|
+
* Measured 2026-11-16 in a real gate container (`network: 'none'`): a server on 127.0.0.1 and a client in
|
|
210
|
+
* the same sandbox, even in another process, talk normally; an outbound connection is refused (EACCES).
|
|
211
|
+
* So a database the suite starts for itself would be reachable — but PostgreSQL cannot start: on Windows
|
|
212
|
+
* it emulates signals through a named pipe, `\\.\pipe\pgsignal_<pid>`, and an AppContainer may create
|
|
213
|
+
* pipes only under `\\.\pipe\LOCAL\`. `initdb` retries and gets through; the server itself stops with
|
|
214
|
+
* "FATAL: could not create signal listener pipe … error code 5", and Tabsy's jest `globalSetup`
|
|
215
|
+
* (embedded-postgres) then waits for a server that is gone until the step's time runs out.
|
|
216
|
+
*
|
|
217
|
+
* That run is not the bug failing and not "no tests": it is the machine missing something the tests
|
|
218
|
+
* need. It gets its own words, so the plan and the goal check are told the change is judged without
|
|
219
|
+
* running these tests, rather than spending plans on a test that cannot start here. Returns null when
|
|
220
|
+
* the output shows no such thing.
|
|
221
|
+
*/
|
|
222
|
+
function missingService(output) {
|
|
223
|
+
const out = output.replace(/\x1b\[[0-9;]*m/g, '');
|
|
224
|
+
if (/could not create signal listener pipe/i.test(out)) {
|
|
225
|
+
return {
|
|
226
|
+
service: 'a PostgreSQL database',
|
|
227
|
+
detail: "the tests start their own PostgreSQL, and PostgreSQL cannot run inside Looply's sandbox on Windows",
|
|
228
|
+
};
|
|
229
|
+
}
|
|
230
|
+
if (/Could not find a working container runtime strategy|Cannot connect to the Docker daemon|docker_engine/i.test(out)) {
|
|
231
|
+
return { service: 'Docker', detail: "the tests start containers, and there is no Docker inside Looply's sandbox" };
|
|
232
|
+
}
|
|
233
|
+
// A refused connection to a database's port, on this PC or elsewhere, from a sandbox with no network.
|
|
234
|
+
for (const refused of out.matchAll(/\bconnect (?:ECONNREFUSED|EACCES) (\S+):(\d+)\b/g)) {
|
|
235
|
+
const port = refused[2];
|
|
236
|
+
const named = SERVICE_PORTS[port];
|
|
237
|
+
if (named !== undefined) {
|
|
238
|
+
const local = /^(127\.|::1|\[::1\]|localhost)/i.test(refused[1]);
|
|
239
|
+
return {
|
|
240
|
+
service: named,
|
|
241
|
+
detail: local
|
|
242
|
+
? `nothing inside Looply's sandbox answers on port ${port}, and the sandbox cannot reach one running elsewhere on this PC`
|
|
243
|
+
: "Looply's sandbox has no network, so the tests cannot reach the server they connect to",
|
|
244
|
+
};
|
|
245
|
+
}
|
|
246
|
+
}
|
|
247
|
+
const prisma = /Can't reach database server at [`'"]?[^`'"\s:]+[`'"]?:[`'"]?(\d+)/.exec(out);
|
|
248
|
+
if (prisma !== null) {
|
|
249
|
+
return {
|
|
250
|
+
service: SERVICE_PORTS[prisma[1]] ?? 'a database server',
|
|
251
|
+
detail: "the tests connect to a database server, and Looply's sandbox cannot reach one",
|
|
252
|
+
};
|
|
253
|
+
}
|
|
254
|
+
return null;
|
|
255
|
+
}
|
|
256
|
+
/** The plain sentence for a run that needed a missing service: the change is not what failed. */
|
|
257
|
+
function missingServiceSentence(m) {
|
|
258
|
+
return `this repository's tests need ${m.service}, which Looply cannot start in its sandbox here (${m.detail})`;
|
|
259
|
+
}
|
|
260
|
+
/**
|
|
261
|
+
* 🔴 A RUN THAT RAN NO TEST (TAB-B-66, -86, -91, -95, -100; 2026-10-18 → 11-11).
|
|
262
|
+
*
|
|
263
|
+
* Inside the sandbox Tabsy's jest could not see its own test roots and printed
|
|
264
|
+
* "No tests found, exiting with code 1 … Pattern: test\\pay-at-table\\x.spec.ts - 0 matches".
|
|
265
|
+
* That output names the test file, so it was judged RED: "reproduced" on the
|
|
266
|
+
* base tree, and "still fails" at every delivered tree after the fix was in —
|
|
267
|
+
* six goals re-planned to the limit over a test that never ran once. A runner
|
|
268
|
+
* saying it found or ran nothing is the machine's problem, never the bug's:
|
|
269
|
+
* neither red nor green. Returns what the runner said, or null.
|
|
270
|
+
*/
|
|
271
|
+
function ranNoTests(output) {
|
|
272
|
+
const out = output.replace(/\x1b\[[0-9;]*m/g, '');
|
|
273
|
+
if (/\bNo tests found\b/i.test(out))
|
|
274
|
+
return 'the test runner found no tests to run';
|
|
275
|
+
if (/\bNo test (?:files|suites?) found\b/i.test(out))
|
|
276
|
+
return 'the test runner found no test files to run';
|
|
277
|
+
// jest / vitest summaries in which every test was skipped (a `-t` name that matched nothing).
|
|
278
|
+
const summary = /^\s*Tests:?\s+(.*\btotal\b.*|.*\bskipped\b.*)$/im.exec(out)?.[1];
|
|
279
|
+
if (summary !== undefined && !/\b(passed|failed)\b/i.test(summary) && /\bskipped\b|\b0 total\b/i.test(summary)) {
|
|
280
|
+
return 'the test runner skipped every test (the test name matched none)';
|
|
281
|
+
}
|
|
282
|
+
if (/^\s*0 passing\b/m.test(out) && !/^\s*\d+ failing\b/m.test(out))
|
|
283
|
+
return 'the test runner ran no tests';
|
|
284
|
+
return null;
|
|
285
|
+
}
|
|
165
286
|
/**
|
|
166
287
|
* What one run of the test says. RED needs more than a non-zero exit: the
|
|
167
288
|
* output must name the test or its file, or the failure could be a missing
|
|
168
289
|
* tool or a syntax error elsewhere — a red that shows nothing about the bug.
|
|
290
|
+
* A run that ran no test at all is neither red nor green (`ranNoTests`).
|
|
169
291
|
*/
|
|
170
|
-
function judgeRun(result, repro) {
|
|
292
|
+
function judgeRun(result, repro, opts = {}) {
|
|
293
|
+
/**
|
|
294
|
+
* 🔴 A TEST THAT ALREADY RAN ON THIS MACHINE IS NOT EXCUSED BY A SERVICE MESSAGE (review of 0.5.53). The check of a
|
|
295
|
+
* delivered tree re-runs a test whose red run on the base was real — the suite started, the test ran. If it fails
|
|
296
|
+
* now and its output mentions a refused connection, that is the FIX's doing (it added a call to something that is
|
|
297
|
+
* not there) or a client's log line beside a real failure: red, a gap the next plan closes — never "could not run,
|
|
298
|
+
* not blocking". Only a first run on the base, where nothing is yet known to start, reads the message as "this
|
|
299
|
+
* suite needs a service the sandbox cannot give".
|
|
300
|
+
*/
|
|
301
|
+
const names = (out) => {
|
|
302
|
+
const f = (0, node_path_1.basename)(repro.file);
|
|
303
|
+
const st = f.replace(/\.[^.]+$/, '');
|
|
304
|
+
return out.includes(repro.test) || out.includes(f) || (st.length >= 4 && out.includes(st));
|
|
305
|
+
};
|
|
306
|
+
const provenRunner = opts.ranHereBefore === true && !result.timedOut && result.unavailable !== true && names(result.output ?? '');
|
|
307
|
+
// Before the timeout: a suite waiting for a database that died at startup runs out of time, and its
|
|
308
|
+
// output says why (`missingService`). A green run is green whatever it logged on the way.
|
|
309
|
+
if (!(result.ok && !result.timedOut) && !provenRunner) {
|
|
310
|
+
const missing = missingService(result.output ?? '');
|
|
311
|
+
if (missing !== null)
|
|
312
|
+
return { verdict: 'not-run', because: missingServiceSentence(missing), service: missing.service };
|
|
313
|
+
}
|
|
171
314
|
if (result.timedOut)
|
|
172
315
|
return { verdict: 'not-run', because: 'the test run timed out' };
|
|
173
316
|
if (result.unavailable === true)
|
|
174
317
|
return { verdict: 'not-run', because: 'the test could not be started on this machine' };
|
|
318
|
+
const none = ranNoTests(result.output ?? '');
|
|
319
|
+
if (none !== null)
|
|
320
|
+
return { verdict: 'not-run', because: `${none} (exit ${result.exitCode ?? 'unknown'})` };
|
|
175
321
|
if (result.ok)
|
|
176
322
|
return { verdict: 'green' };
|
|
177
323
|
const out = result.output ?? '';
|
|
@@ -185,6 +331,17 @@ function tail(s, n = 1500) {
|
|
|
185
331
|
const t = (s ?? '').trim();
|
|
186
332
|
return t.length > n ? `…${t.slice(-n)}` : t;
|
|
187
333
|
}
|
|
334
|
+
/**
|
|
335
|
+
* A record written before `ranNoTests` existed may say "reproduced" about a run that ran no test.
|
|
336
|
+
* Read back, it is what it always was — no reproduction — so no later check gates on it and no
|
|
337
|
+
* pull request claims "failed before". The file itself is left as it was written.
|
|
338
|
+
*/
|
|
339
|
+
function standing(r) {
|
|
340
|
+
const none = r.outcome === 'reproduced' && r.failedBefore !== undefined ? ranNoTests(r.failedBefore.tail ?? '') : null;
|
|
341
|
+
if (none === null)
|
|
342
|
+
return r;
|
|
343
|
+
return { ...r, outcome: 'skipped', because: `the reproduction test never ran on the base tree: ${none}` };
|
|
344
|
+
}
|
|
188
345
|
/** One file per goal version, under the PROJECT's state, so every checkout of the project reads the same one. */
|
|
189
346
|
class ReproStore {
|
|
190
347
|
stateDir;
|
|
@@ -200,7 +357,7 @@ class ReproStore {
|
|
|
200
357
|
read(goalId, goalVersion) {
|
|
201
358
|
try {
|
|
202
359
|
const r = JSON.parse((0, node_fs_1.readFileSync)(this.path(goalId, goalVersion), 'utf8'));
|
|
203
|
-
return r && r.schema === 1 && r.goalId === goalId ? r : null;
|
|
360
|
+
return r && r.schema === 1 && r.goalId === goalId ? standing(r) : null;
|
|
204
361
|
}
|
|
205
362
|
catch {
|
|
206
363
|
return null;
|
|
@@ -292,8 +449,20 @@ async function reproduceBug(input, deps) {
|
|
|
292
449
|
return skipped('the agent gave no reproduction answer');
|
|
293
450
|
if (answer.kind === 'skip')
|
|
294
451
|
return skipped(`no test can show this bug here: ${answer.because}`);
|
|
295
|
-
if (answer.kind === 'question')
|
|
452
|
+
if (answer.kind === 'question') {
|
|
453
|
+
// A reading to act on: go on to it, and say so — never a question before the first attempt.
|
|
454
|
+
if (answer.assume !== undefined && !answered) {
|
|
455
|
+
deps.log(`repro : the report can be read more than one way — going on with this reading: ${answer.assume}`);
|
|
456
|
+
return {
|
|
457
|
+
...base,
|
|
458
|
+
outcome: 'not-reproduced',
|
|
459
|
+
because: 'the report did not say exactly where the bug is, so this is fixed without a reproduction, on a stated reading of it',
|
|
460
|
+
question: answer.question,
|
|
461
|
+
assumed: answer.assume,
|
|
462
|
+
};
|
|
463
|
+
}
|
|
296
464
|
return ask(answer.question, 'the test writer could not tell where the bug is');
|
|
465
|
+
}
|
|
297
466
|
const content = deps.readFile(answer.file);
|
|
298
467
|
if (content === null)
|
|
299
468
|
return skipped(`the reproduction test file ${answer.file} was not written`);
|
|
@@ -317,11 +486,19 @@ async function reproduceBug(input, deps) {
|
|
|
317
486
|
return { ...base, outcome: 'reproduced', ...kept, failedBefore: run, passesNow: null };
|
|
318
487
|
}
|
|
319
488
|
if (judged.verdict === 'green') {
|
|
489
|
+
// Not a question either: already fixed, or somewhere this test did not look. The plan finds out which.
|
|
490
|
+
deps.log(`repro : "${answer.test}" passes on the current code — the bug did not show there; going on without a reproduction`);
|
|
320
491
|
return {
|
|
321
|
-
...
|
|
492
|
+
...base,
|
|
493
|
+
outcome: 'not-reproduced',
|
|
494
|
+
because: `the test "${answer.test}" passes on the current code, so the bug did not show there`,
|
|
495
|
+
passedOnBase: true,
|
|
322
496
|
...kept,
|
|
323
497
|
};
|
|
324
498
|
}
|
|
499
|
+
if (judged.service !== undefined) {
|
|
500
|
+
return { ...skipped(`the reproduction test was written but cannot run here: ${judged.because}`), needs: judged.service, ...kept };
|
|
501
|
+
}
|
|
325
502
|
return { ...skipped(`the reproduction test could not be run: ${judged.because}`), ...kept };
|
|
326
503
|
}
|
|
327
504
|
catch (e) {
|
|
@@ -335,8 +512,43 @@ async function reproduceBug(input, deps) {
|
|
|
335
512
|
}
|
|
336
513
|
}
|
|
337
514
|
// ── what the plan and every round are told ──
|
|
515
|
+
/**
|
|
516
|
+
* What a bug that was NOT shown adds to the goal's words: the reading the work goes on to, or that a test written from
|
|
517
|
+
* the report already passes. '' for anything else (including a record written before these fields existed).
|
|
518
|
+
*/
|
|
519
|
+
function unreproducedBrief(r) {
|
|
520
|
+
if (r === null || r.outcome !== 'not-reproduced')
|
|
521
|
+
return '';
|
|
522
|
+
if (r.assumed !== undefined && r.assumed !== '') {
|
|
523
|
+
return [
|
|
524
|
+
'',
|
|
525
|
+
'',
|
|
526
|
+
'THIS REPORT CAN BE READ MORE THAN ONE WAY, AND NOBODY WAS ASKED FIRST. Looply is working to this reading of it:',
|
|
527
|
+
`- ${r.assumed}`,
|
|
528
|
+
...(r.question === undefined || r.question === '' ? [] : [`- (what was unclear: ${r.question})`]),
|
|
529
|
+
'Work to that reading. When the work is checked, the answer must say it was assumed ("assumed"), in these words,',
|
|
530
|
+
'so the person who reported it can correct it in one reply.',
|
|
531
|
+
].join('\n');
|
|
532
|
+
}
|
|
533
|
+
if (r.passedOnBase === true && r.test) {
|
|
534
|
+
return [
|
|
535
|
+
'',
|
|
536
|
+
'',
|
|
537
|
+
'A TEST WRITTEN FROM THIS REPORT PASSES ON THE CURRENT CODE, so the bug did not show there:',
|
|
538
|
+
`- test "${r.test}"${r.file ? ` in ${r.file}` : ''} (not kept; the tree was reset)`,
|
|
539
|
+
'Either the bug is already fixed, or it lies somewhere that test did not look. Find out which BEFORE changing',
|
|
540
|
+
'anything: read how the report\'s case is handled now. If it is already fixed, change nothing — say where it is',
|
|
541
|
+
'handled. If it is not, fix it where it really is.',
|
|
542
|
+
].join('\n');
|
|
543
|
+
}
|
|
544
|
+
return '';
|
|
545
|
+
}
|
|
338
546
|
/** The block a reproduced bug adds to the goal's words, or '' for any other outcome. */
|
|
339
547
|
function reproductionBrief(r) {
|
|
548
|
+
if (r !== null && r.outcome === 'not-reproduced')
|
|
549
|
+
return unreproducedBrief(r);
|
|
550
|
+
if (r !== null && r.outcome === 'skipped' && r.needs !== undefined)
|
|
551
|
+
return missingServiceBrief(r);
|
|
340
552
|
if (r === null || r.outcome !== 'reproduced' || !r.test || !r.file || !r.command)
|
|
341
553
|
return '';
|
|
342
554
|
const head = [
|
|
@@ -361,6 +573,21 @@ function reproductionBrief(r) {
|
|
|
361
573
|
}
|
|
362
574
|
return `${head}${body}`;
|
|
363
575
|
}
|
|
576
|
+
/**
|
|
577
|
+
* 🔴 THE TESTS CANNOT RUN HERE, SAID TO THE PLAN, THE ROUNDS AND THE GOAL CHECK. Without it, Tabsy's goals
|
|
578
|
+
* were judged against a test that could never start in the sandbox, and re-planned until the plan limit.
|
|
579
|
+
* With it, the work is judged on the change itself, and nobody tries to bring a database into the sandbox.
|
|
580
|
+
*/
|
|
581
|
+
function missingServiceBrief(r) {
|
|
582
|
+
return [
|
|
583
|
+
'',
|
|
584
|
+
'',
|
|
585
|
+
`THIS REPOSITORY'S TESTS CANNOT RUN ON THIS MACHINE: they need ${r.needs}, which Looply cannot start in its sandbox.`,
|
|
586
|
+
...(r.test && r.file ? [`- Looply wrote the test "${r.test}" in ${r.file} to show the bug; it could not be run here.`] : []),
|
|
587
|
+
`- Make the change. Do not try to install, start or configure ${r.needs}, and do not change the test setup to get round it.`,
|
|
588
|
+
'- Judge the work by the change itself and the checks that can run; a test that cannot start here is not a gap in the work.',
|
|
589
|
+
].join('\n');
|
|
590
|
+
}
|
|
364
591
|
/** The goal's words with the reproduction attached, for the plan, the goal check and the rounds. */
|
|
365
592
|
function withReproduction(goalText, r) {
|
|
366
593
|
return `${goalText}${reproductionBrief(r)}`;
|
|
@@ -388,7 +615,8 @@ async function checkReproGreen(r, deps) {
|
|
|
388
615
|
try {
|
|
389
616
|
const result = await deps.run(r.command, exports.REPRO_RUN_TIMEOUT_MS);
|
|
390
617
|
const run = { sha, exitCode: result.exitCode ?? null, at: deps.now().toISOString(), tail: tail(result.output) };
|
|
391
|
-
|
|
618
|
+
// Only a record that says the test ran red here is checked (`evaluateWithReproduction`): its runner is proven.
|
|
619
|
+
const j = judgeRun(result, { test: r.test, file: r.file }, { ranHereBefore: r.outcome === 'reproduced' });
|
|
392
620
|
if (j.verdict === 'green')
|
|
393
621
|
return { state: 'green', run };
|
|
394
622
|
if (j.verdict === 'red')
|