Repository navigation
learn: retrospective learnings - #1637
peco-engineer-bot[bot] wants to merge 35 commits into
Conversation
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 2 Low
Looks good: this PR only adds documentation, and I checked the new 2026-10-23 entries against ArrowStreamResult. There are two low-severity issues: a leftover "No learnings have been recorded yet" line, and a new rule that claims more than the PR #1734 fix supports.
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The seed text still says "No learnings have been recorded yet", but this PR adds about 15 dated sections right below it. The author phase reads this file as author.knowledge_log, so a stale "empty log" line contradicts everything after it. Delete or reword that sentence when the first entries go in. Also, ## Entries starts directly after the paragraph with no blank line before it, while every other section has one.
(Anchored to the nearest changed line — see the description for the exact location.)
| - **Context:** In author run #39813458379, the implement phase re-read the same files the plan/diagnosis phase had already examined (DatabricksResultSet constructors, ColumnMetadata, DatabricksTypeUtil, MetadataParser, the e2e test file), repeating most of the exploration across the phase boundary. | ||
| **Rule:** Diagnosis-phase findings (exact file:line anchors, constructor signatures, helper locations) should be carried into the implement-phase prompt so the implement agent doesn't re-grep/re-read the same files from scratch. | ||
|
|
||
| ### 2026-10-23: learnings since 2026-10-22T17:30:18Z |
There was a problem hiding this comment.
🔵 Low — The new 2026-10-23 rule is broader than the fix it comes from. It says to pick the SEA chunk provider "from the shape of the server response … not from client config flags". But the merged code still needs a client flag to choose SeaInlineArrowChunkProvider: isInlineArrow && connectionContext.isBoundedSeaApiEnabled() (line 76). PR #1734 only removed the CloudFetch gate. Read literally, the rule would push a future author to drop the bounded-SEA gate as well, and that would send non-bounded inline results to the wrong provider. Narrow the rule to CloudFetch, e.g. "don't let the CloudFetch setting decide between inline and remote providers when attachment != null".
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The preamble right above the new ## Entries header still says "No learnings have been recorded yet." That's no longer true once this PR merges, and the author phase reads this file as author.knowledge_log, so it's a contradiction for both people and the bot. There's also no blank line between that paragraph and ## Entries. Consider replacing the sentence with something like "Dated sections are appended below by the retrospective flow." and adding a blank line before the header. The retrospective flow's template should change too, so the stale line doesn't come back on the next rolling PR.
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The preamble right above the new ## Entries heading still says "No learnings have been recorded yet." This PR adds about 20 dated sections, so that sentence is now wrong. The author phase reads this file as author.knowledge_log, so it sees a contradiction at the top of its knowledge source. The sentence should be removed or reworded (e.g. "Dated sections are appended below by the retrospective flow."). Also add a blank line before ## Entries. Right now the heading sits directly against the paragraph above it, which some Markdown renderers and linters handle badly.
| **Rule:** `UserAgent.sanitize()` only applies a fixed regex replacement and cannot throw `IllegalArgumentException` — the exception the sibling call sites catch comes from `withOtherInfo`'s key validation, so a bare `sanitize` call (e.g. inside the request-path transport wrapper) needs no try/catch. | ||
|
|
||
| ### 2026-10-05: learnings since 2026-10-04T18:11:23Z | ||
| - **Context:** PR #1707 unified OAuth `Auth_Scope` parsing but a reviewer flagged that the sibling auth flows had drifted — client-secret M2M and U2M split the value on whitespace into multiple scopes, while JWT-assertion M2M still wrapped the raw string in `Collections.singletonList(...)`, so `Auth_Scope=sql jobs` became one malformed scope `"sql jobs"` on that path. |
There was a problem hiding this comment.
🔵 Low — The 2026-09-12 learning names UseQueryForMetadata=0 as an active user configuration. The 2026-10-27 entry later in this same PR records that PR #1747 removed that DatabricksJdbcUrlParams key. The author phase reads this log as guidance, so the earlier entry could lead it to reference or test a param that no longer exists. Consider adding a short note to the older entry, e.g. "(param later removed in #1747)". The rule itself, logging at WARN when a fallback overrides explicit config, is still valid.
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — ## Entries is added right under the preamble line that says "No learnings have been recorded yet." That line is now wrong. The author phase reads this file, so it gets a contradictory preamble. There is also no blank line before the new heading. Change the sentence to something like "Dated sections are appended below by the retrospective flow." and add a blank line before ## Entries.
| - **Context:** PR #1739. A bad caller argument (`index < 1`) first threw `DatabricksDriverException`, which is a RuntimeException, using the conversion error code `COMPLEX_DATA_TYPE_ARRAY_CONVERSION_ERROR`. The reviewer pointed out that the JDBC contract requires a `SQLException`. The fix throws `DatabricksSQLException` with `DatabricksDriverErrorCode.INPUT_VALIDATION_ERROR`, and a test asserts both the SQLState and the vendor error code. | ||
| **Rule:** In a JDBC method that declares `throws SQLException`, an invalid caller argument must throw `DatabricksSQLException` with `INPUT_VALIDATION_ERROR`, never an unchecked `DatabricksDriverException` or an unrelated conversion code. Add a test that asserts `getSQLState()` and `getErrorCode()` on that path. | ||
| - **Context:** PR #1739 built an in-memory `DatabricksResultSet` whose column types came from complex-type element text. Raw type text such as `DECIMAL(10,2)` or `ARRAY<INT>` does not match the cases in `DatabricksTypeUtil.getColumnInfoType`, so the column type silently became `USER_DEFINED_TYPE`/`Types.OTHER`. Also, `DatabricksResultSetMetaData` crashed on `ColumnInfoTypeName.valueOf("TIMESTAMP_NTZ")` because that enum has no TIMESTAMP_NTZ constant. | ||
| **Rule:** Before mapping Databricks type text with `getColumnInfoType`/`ColumnInfoTypeName.valueOf`, strip the type parameters (`split("[<(]")`), take DECIMAL precision and scale from the declared type, and map TIMESTAMP_NTZ to TIMESTAMP. Add a metadata test for at least one parameterized element type and one non-STRING element type. |
There was a problem hiding this comment.
🔵 Low — The rule says "map TIMESTAMP_NTZ to TIMESTAMP" with no qualifier. A reader could take that to mean the reported column type name should be normalized too. By default the driver keeps reporting TIMESTAMP_NTZ as the type name (DatabricksResultSetMetaData.java:491-493, gated by EnableTimestampNtzTypeName). Only the internal ColumnInfoTypeName/Types.TIMESTAMP lookup should be mapped. Please scope the rule, e.g. "map TIMESTAMP_NTZ to TIMESTAMP for the ColumnInfoTypeName/SQL-type lookup only, and keep the reported type name as controlled by EnableTimestampNtzTypeName". Otherwise a future author run could undo the #1495 behavior.
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 2 Low
This PR only adds learning-log entries and looks good, with two small wording issues. I checked the factual claim in the 2026-10-30 entry against DatabricksSQLException.java and it is correct: the (reason, DatabricksDriverErrorCode) constructor sends the vendor code through getVendorCode(reason), and only the (reason, sqlState, DatabricksDriverErrorCode) constructor uses internalError.getCode().
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The preamble line just above this heading still says "No learnings have been recorded yet.", but now 30+ dated sections follow it. The author phase reads this file as its knowledge source (author.knowledge_log), so an untrue opening line is confusing. Also, ## Entries comes right after the paragraph with no blank line before it. Please delete or rewrite the "No learnings…" sentence, for example to "Dated sections are appended below by the retrospective flow.", and add a blank line before ## Entries. If the retrospective flow regenerates the header, fix it there too so the next rolling PR doesn't bring the stale text back.
| - **Context:** On PR #1754, the SEA session-create replay fixtures still matched the old `/api/2.0/sql/sessions/` URL with `warehouse_id` in the body, but the driver now POSTs to `/api/2.0/sql/warehouses/{id}/sessions`. WireMock `bodyPatterns.equalTo` matches exactly, so any change to the URL or body breaks replay. | ||
| **Rule:** When changing a SEA request's URL or JSON body, grep `src/test/resources/sqlexecapi/**/mappings/*.json` for the old path. Update the `url` and exact `equalTo` body patterns, and delete mappings for behavior that was removed. | ||
| - **Context:** On PR #1742, the engineer-bot follow-up was re-triggered four times and failed the same way each time (`CalledProcessError` while finalizing the bundle commit, "non-bot commit at HEAD"). Only on the fifth run did it escalate as NEEDS HUMAN DECISION. The requested change was a few unit tests. | ||
| **Rule:** If a bot follow-up fails twice with the same internal or commit-setup error, stop re-triggering it. Mark the thread blocked or needs-human, and have a person clean up the branch or add the small fix by hand. |
There was a problem hiding this comment.
🔵 Low — This PR #1742 rule covers the same failure (a CalledProcessError saying "non-bot commit at HEAD" while finalizing the bundle commit) as the 2026-10-29 PR #1739 entry, but its threshold is worded differently. The older entry says to stop "more than once … after the first repeat"; this one says to stop after the error "fails twice". These probably mean the same thing, but the bot reads both and the wording is inconsistent. It would be clearer to merge them into one rule with one threshold, or to have this entry point back to the earlier one and only add the new part (a person cleans up the branch or makes the small fix by hand).
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 1 Low
Looks good: this PR only appends dated learning entries, and the new 2026-11-03 entry (separate Thrift/SEA HTTP clients and IdleHttpConnectionExpiry) is clear and actionable. One low-severity issue: the file's preamble still says no learnings have been recorded.
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The preamble right above this new ## Entries heading still says "No learnings have been recorded yet." With more than 15 dated sections now added, that line is wrong. The author phase reads this file (author.knowledge_log), so a stale sentence at the top could confuse it. There is also no blank line between that paragraph and the ## Entries heading. Please drop or reword the sentence (e.g. "Dated sections are appended below by the retrospective flow.") and add a blank line before the heading.
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 1 Low
This PR only changes the learning log, and it looks good apart from one low-severity issue: the seed intro still says no learnings have been recorded. I checked the new 2026-11-05 entries and the earlier ones for accuracy and for rules that contradict each other, and found no conflicts.
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The new ## Entries section is appended right under the seed text that still says "No learnings have been recorded yet." Once this PR merges, that sentence is false, and the author phase reads this file (author.knowledge_log), so it could be misled. Drop the "No learnings have been recorded yet." sentence and keep only the note that dated sections are appended below. Also add a blank line before ## Entries so it matches the spacing used elsewhere in the file.
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 1 Low
This docs-only PR adds the 2026-11-06 learnings and looks good, with one low-severity issue: the file still opens with a stale "No learnings have been recorded yet" line, and the new ## Entries heading has no blank line before it. I did not find any contradictions or factual errors in the new rules.
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The new ## Entries heading comes right after the preamble line "No learnings have been recorded yet". That line is now false, because this PR adds about 60 learnings below it. The author phase reads this file as author.knowledge_log, so leaving a sentence that contradicts the content is confusing for both people and the bot. The heading also has no blank line before it, so some Markdown renderers will merge it into the previous paragraph instead of showing a heading.
Suggested fix: remove or reword the "No learnings have been recorded yet" sentence (for example, "Dated sections are appended below by the retrospective flow.") and add a blank line before ## Entries. The retrospective flow's seeding or append logic may need to handle this too, so later rolling PRs don't run into it again.
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 1 Low
Looks good. This PR only adds entries to the learning log, and they are consistent with each other and with CLAUDE.md. There is one low-severity issue: the file's opening text still says no learnings have been recorded. Nit (no inline comment): the 2026-10-29 entry says "the same cast pattern is still in getArray(long, int)". That is a claim about the current code, so it will go out of date and should be checked or reworded as history.
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The preamble right above the new ## Entries heading (lines 11–12) still says "No learnings have been recorded yet", but this PR adds about 200 lines of learnings below it. The author phase reads this file as author.knowledge_log, so the stale sentence contradicts the content that follows. Also, ## Entries comes straight after the paragraph with no blank line, which some Markdown renderers handle inconsistently.
Suggested fix: have the retrospective flow replace or drop the "No learnings have been recorded yet." sentence the first time it appends, and put a blank line before ## Entries.
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 2 Low
Docs-only PR with two low-severity accuracy issues. The seeded intro now contradicts the new entries, and one rule misstates how Level.equals() behaves. Nit: the 2026-10-29 and 2026-10-30 rules give different thresholds for when to stop re-triggering a failing bot follow-up ("after the first repeat" vs "fails twice"). Consider making them say the same thing.
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The new ## Entries section goes right under the seeded intro, and that intro still says "No learnings have been recorded yet." Once this PR merges, the canonical log will contradict itself. The author phase reads this file as knowledge, so the contradiction can mislead it. Remove or reword that sentence in the same change, for example: "Dated sections are appended below by the retrospective flow." Also add a blank line before ## Entries, because right now the heading is directly attached to the paragraph above it.
| - **Context:** PR #1670 fixed later logging-enabled JDBC connections being unable to emit logs because an earlier `LogLevel=OFF` connection had already flipped the one-time `isLoggerInitialized` static guard (and installed no usable handler). The fix returns from the OFF/no-op path *without* setting the guard, and also leaves it unset when handler installation throws (`testFailedInitializationCanBeRetried`), so a later capable caller can still initialize. | ||
| **Rule:** Set a one-time initialization guard (static `isInitialized` flag / singleton) only after a fully-functional configuration is actually installed — never on a disabled/no-op early path or before an operation that can throw — otherwise the first weak or failed caller permanently poisons initialization for every later caller. | ||
| - **Context:** In PR #1670 a reviewer flagged that detecting the "logging off" case should compare `level.intValue() == Level.OFF.intValue()` rather than by object reference/equality, since `java.util.logging.Level` allows distinct instances that share the same integer value. | ||
| **Rule:** Compare `java.util.logging.Level` (and similar level/enum-like value objects) by `intValue()`, not by `==` reference or `.equals()`, because semantically-equal levels can be different object instances. |
There was a problem hiding this comment.
🔵 Low — This rule gets .equals() wrong. java.util.logging.Level.equals() already compares the integer value field, so two separate Level instances with the same intValue() are already .equals(). Only == reference comparison has the problem the rule describes. As written, the rule could lead future author runs to rewrite correct .equals() checks, or reviewers to flag them. Suggested wording: "Compare java.util.logging.Level by intValue() (or .equals()), never by == reference, because semantically-equal levels can be distinct instances."
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 2 Low
The learnings themselves are fine. This PR only changes docs. Two low-severity issues: the preamble still says no learnings have been recorded, which is now stale, and two entries give slightly different stop thresholds for repeated bot commit failures.
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The preamble on line 11 still says "No learnings have been recorded yet." The PR then adds more than 20 dated sections right below it. That line is now false, and the author phase reads this file (author.knowledge_log) as context, so the statement contradicts what follows it. There is also no blank line between that paragraph and the new ## Entries heading. Some renderers handle a heading with no blank line before it inconsistently.
Fix: delete or reword the "No learnings have been recorded yet" sentence (for example, "Dated sections are appended below by the retrospective flow.") and put a blank line before ## Entries. If the retrospective flow generates this text, fix it there so it doesn't come back.
| - **Context:** On PR #1754, the SEA session-create replay fixtures still matched the old `/api/2.0/sql/sessions/` URL with `warehouse_id` in the body, but the driver now POSTs to `/api/2.0/sql/warehouses/{id}/sessions`. WireMock `bodyPatterns.equalTo` matches exactly, so any change to the URL or body breaks replay. | ||
| **Rule:** When changing a SEA request's URL or JSON body, grep `src/test/resources/sqlexecapi/**/mappings/*.json` for the old path. Update the `url` and exact `equalTo` body patterns, and delete mappings for behavior that was removed. | ||
| - **Context:** On PR #1742, the engineer-bot follow-up was re-triggered four times and failed the same way each time (`CalledProcessError` while finalizing the bundle commit, "non-bot commit at HEAD"). Only on the fifth run did it escalate as NEEDS HUMAN DECISION. The requested change was a few unit tests. | ||
| **Rule:** If a bot follow-up fails twice with the same internal or commit-setup error, stop re-triggering it. Mark the thread blocked or needs-human, and have a person clean up the branch or add the small fix by hand. |
There was a problem hiding this comment.
🔵 Low — Two rules give different thresholds for the same situation, a repeated bundle-commit CalledProcessError:
- Line 181 (2026-10-29) says to escalate "after the first repeat".
- Line 193 (2026-10-30) says to stop once it "fails twice".
These can be read as the same threshold (two failures in total) or as two different ones (escalate on the 2nd failure vs. after the 2nd). The author bot gets this log as instructions, so it should get one clear rule. Combine the two entries into a single statement, e.g. "after the second identical failure, stop and mark NEEDS HUMAN DECISION".
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 2 Low
Looks good overall; this PR only touches documentation. I found two low-severity problems: one learned rule gets Level.equals() wrong, and the log's opening text still says no learnings exist. The rest of the rules (JDBC FLOAT/DOUBLE mapping, getTypeInfo ordering, first-wins column lookup, Julian/Gregorian cutover window) are accurate.
| - **Context:** PR #1670 fixed later logging-enabled JDBC connections being unable to emit logs because an earlier `LogLevel=OFF` connection had already flipped the one-time `isLoggerInitialized` static guard (and installed no usable handler). The fix returns from the OFF/no-op path *without* setting the guard, and also leaves it unset when handler installation throws (`testFailedInitializationCanBeRetried`), so a later capable caller can still initialize. | ||
| **Rule:** Set a one-time initialization guard (static `isInitialized` flag / singleton) only after a fully-functional configuration is actually installed — never on a disabled/no-op early path or before an operation that can throw — otherwise the first weak or failed caller permanently poisons initialization for every later caller. | ||
| - **Context:** In PR #1670 a reviewer flagged that detecting the "logging off" case should compare `level.intValue() == Level.OFF.intValue()` rather than by object reference/equality, since `java.util.logging.Level` allows distinct instances that share the same integer value. | ||
| **Rule:** Compare `java.util.logging.Level` (and similar level/enum-like value objects) by `intValue()`, not by `==` reference or `.equals()`, because semantically-equal levels can be different object instances. |
There was a problem hiding this comment.
🔵 Low — The 2026-09-08 rule says not to compare java.util.logging.Level with .equals(). That is wrong: Level.equals already compares the integer value, so level.equals(Level.OFF) and level.intValue() == Level.OFF.intValue() give the same answer. Only == (reference comparison) is unsafe, because two distinct instances can share a value.
The author phase reads this log, so this wording could push future fixes to replace correct .equals() calls without need, or a reviewer could flag them. Suggested rewording: "Compare Level with intValue() or .equals(), never with ==, because semantically equal levels can be different instances."
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The new ## Entries section goes directly under the preamble line "No learnings have been recorded yet." That line is now false, because this PR adds dozens of entries. The heading also has no blank line before it, so the paragraph runs straight into it. Drop or reword the stale sentence (for example, "Dated sections are appended below by the retrospective flow.") and add a blank line before ## Entries. Otherwise readers and the author-phase consumer get a self-contradicting log.
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 1 Low
Looks good. This PR only adds entries to the knowledge log, and there is one low-severity issue: the file's intro still says no learnings have been recorded. I spot-checked the new entries against the repo conventions and found no contradictions with CLAUDE.md or driver-pitfalls.md.
|
|
||
| No learnings have been recorded yet. Dated sections are appended below by the | ||
| retrospective flow. | ||
| ## Entries |
There was a problem hiding this comment.
🔵 Low — The file now has a lot of dated learnings under ## Entries, but the intro paragraph right above it (lines 11–12) still says "No learnings have been recorded yet." The author phase reads this file as its knowledge log (author.knowledge_log), so the stale sentence contradicts the content that follows it.
The heading is also attached directly to the previous paragraph with no blank line between them. Suggested fix: drop the "No learnings have been recorded yet." sentence, or reword it to something like "Dated sections are appended below by the retrospective flow.", and add a blank line before ## Entries.
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Verdict: 2 Nit
Looks good. This PR only changes the learning log. The new 2026-11-16 rules are accurate, specific and consistent with earlier entries. There are only two nits: a stale preamble sentence and an overlap with the 2026-11-05 batch-failure rule.
Other findings
- ⚪ Nit — The file preamble still says "No learnings have been recorded yet. Dated sections are appended below by the retrospective flow." That is no longer true now that entries exist below it. Consider dropping the first sentence. Also add a blank line before
## Entries, which currently sits directly against the preamble paragraph. - ⚪ Nit — The 2026-11-16 entry on PR #1625 (separate post-commit result-parsing failures from
EXECUTE_FAILED, then clear the batch) mostly repeats the 2026-11-05 rule from PR #1623 / #1625. Consider merging them or cross-referencing so the author-phase reader doesn't get the same guidance twice.
Signed-off-by: peco-engineer-bot[bot] <peco-engineer-bot[bot]@users.noreply.github.com>
Rolling retrospective learnings
This PR accumulates one dated section of learnings per day (from merged PRs and engineer-bot author runs) until it is merged; merging it starts a fresh one. The bot never writes the canonical log directly.
Latest update 2026-11-17: 2 new learning(s) since 2026-11-16T17:26:45Z.