fix(tui): tier events priority + decouple --time from events#42
Merged
Conversation
Three real bugs surfaced when testing 5.2.2 against the maintainer's own data (65 daemon-lifecycle events vs 5 power events in the SQLite events table): 1. Events panel showed daemon noise instead of power events. PRIORITY_EVENTS lumped power + daemon into one tier, so the most- recent N rows in the cap were all daemon-restarts. Power events from days ago got pushed off-screen. Fix: split into POWER_EVENTS (always preserved within cap) and LIFECYCLE_EVENTS (fill remaining slots). Daemon noise can no longer push power events off-screen. 2. `eneru tui --once --events-only` silently fell back to log parsing for sparse-event installs. The default 1h time window was almost always empty -- SQLite returned [], code parsed /var/log/ups-monitor.log instead. The two surfaces showed different content and the operator never noticed. Fix: events query no longer takes a time-range parameter. Always queries the full events table (sparse, so this is fine). 3. `--time` accidentally scoped both the graph window AND the events panel. Cause of #2. Fix: --time and <T> are graph-only now. Anchor comments at tui.py:23-37 (TIME_RANGES) and cli.py:_add_monitor_args (--time) name the bug they prevent so this can't be re-coupled by accident. Knob change: - Removed --full-history (only added in unreleased 5.2.2; --length 0 covers the same use case more cleanly) - Added --length N (default 30, 0 = no cap; argparse type validator rejects negatives + non-numerics) - Live TUI events cap default raised 20 -> 30 (matches CLI default) Snapshot tail (run_once non-events-only path) caps at 10 rows unconditionally so --length 0 doesn't drown the status block -- operators wanting "all events" use --events-only. Test coverage: 13 new tests (981 total, was 968). Mirrors the maintainer's reproducer in test_tiered_trim_preserves_power_events (60 daemon + 3 power, asserts all 3 power events survive cap=20) plus pathological "more power events than cap" + length=0 / --length validation. E2E Test 31 swaps --full-history for --length assertions. Pre-push reviewed by agent-skills:code-reviewer (opus); P3 finding addressed (snapshot tail cap clamp). Co-Authored-By: Claude Opus 4.7 <[email protected]>
Discovered while testing the tier-trim fix on the maintainer's 24-row terminal: the data layer correctly returned 30 events with all 5 power events at the top of the chronological list, but the events panel anchored at the bottom only renders ~11 rows visible in that terminal size. Power events at positions 0-4 sat off-screen until the operator scrolled up. Fix: in run_tui's main loop, compute a visible-rows estimate from the panel layout (logs_end - logs_start - 4) and cap the data query to min(EVENTS_MAX_ROWS_NORMAL, visible_estimate). Tier-trim then guarantees power events fit inside what the panel actually shows. <M> still expands to EVENTS_MAX_ROWS_MORE (500) for full scrollable history. Verified manually on a 24x80 terminal: cap computes to 11; query returns 5 power + 6 most-recent daemon; panel displays all 11 with power events at the top. Co-Authored-By: Claude Opus 4.7 <[email protected]>
📝 WalkthroughWalkthroughThe pull request replaces the Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Owner
Author
|
@coderabbitai full review |
Owner
Author
|
@cubic-dev-ai review this pull request |
✅ Actions performedFull review triggered. |
@m4r1k I have started the AI code review. It will take a few minutes to complete. |
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #42 +/- ##
=======================================
Coverage 79.49% 79.49%
=======================================
Files 25 25
Lines 4515 4530 +15
Branches 841 843 +2
=======================================
+ Hits 3589 3601 +12
- Misses 726 729 +3
Partials 200 200
🚀 New features to boost your workflow:
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
🧹 Nitpick comments (1)
tests/e2e/groups/stats.sh (1)
264-273: Inline the exit-code check underset -e.If
eneru monitor --once --events-only --length 5 ...ever returns non-zero, the subshell exits before the$?branch runs, so the custom log dump never executes. Wrap the command directly in theifinstead.Suggested change
-eneru monitor --once --events-only --length 5 --config "$E2E_DIR/config-e2e-stats.yaml" > /tmp/test31-length.log 2>&1 -if [ $? -ne 0 ]; then +if ! eneru monitor --once --events-only --length 5 \ + --config "$E2E_DIR/config-e2e-stats.yaml" \ + > /tmp/test31-length.log 2>&1; then echo "FAIL: --length 5 returned non-zero" cat /tmp/test31-length.log exit 1 fi🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@tests/e2e/groups/stats.sh` around lines 264 - 273, The test currently runs `eneru monitor --once --events-only --length 5 --config "$E2E_DIR/config-e2e-stats.yaml"` and then checks `$?`, but with `set -e` a non-zero exit will abort before the check and log dump; change the test to run the command inside an `if` conditional (e.g., `if eneru ... > /tmp/test31-length.log 2>&1; then ... else ... fi`) so the failure branch can print the log and exit with failure; update the pass/fail branches around that `if` and keep the same log file `/tmp/test31-length.log` and messages.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@docs/changelog.md`:
- Around line 36-38: Update the changelog bullet about the Live TUI events cap
to reflect that it's height-bounded rather than a flat 30 rows: change the line
mentioning "Live TUI events cap raised from 8 to 30 rows" to say "Live TUI
events cap raised (up to 30 rows, bounded by visible_panel_rows)" or explicitly
note it uses min(30, visible_panel_rows) in normal mode so shorter terminals
show fewer than 30 rows.
In `@docs/configuration.md`:
- Around line 439-446: The examples block mixes invocation styles (e.g., "sudo
eneru validate ..." vs "sudo /opt/ups-monitor/eneru.py monitor ..."); update
every example so package-user commands consistently use the packaged wrapper
form "/opt/ups-monitor/eneru.py" (e.g., replace "sudo eneru ..." and any "eneru
..." invocations with "sudo /opt/ups-monitor/eneru.py ..."), keeping the same
flags and arguments (validate, run, monitor, --dry-run, --once, --events-only,
--verbose, --length, --config) so the block uniformly shows package-user usage.
In `@src/eneru/tui.py`:
- Around line 473-489: When capping rows (when max_events is set) the current
logic treats all non-POWER_EVENTS equally so verbose chatter can evict lifecycle
events; change the trim in the block that references rows, POWER_EVENTS,
max_events, kept_non_power and kept so lifecycle rows are prioritized: split
non_power into lifecycle = [r for r in rows if r[2] in LIFECYCLE_EVENTS] and
chatter = [r for r in rows if r[2] not in LIFECYCLE_EVENTS and r[2] not in
POWER_EVENTS], then when computing remaining fill from the most recent lifecycle
rows first (up to remaining) and only then backfill with the most recent chatter
to produce kept_non_power, re-merge with power and sort by timestamp as before.
In `@tests/test_tui.py`:
- Around line 1240-1252: Rename the one-letter comprehension variable "l" to a
clearer name like "line" in the test assertions around the
query_events_for_display call: update the three asserts that use "l" in their
any() comprehensions and the daemon_lines list comprehension to use "line"
instead (e.g., any("ON_BATTERY: real outage" in line for line in lines) and
daemon_lines = [line for line in lines if "DAEMON_RESTARTED" in line]); this
will satisfy Ruff E741 and improve readability without changing test logic in
the test_tui.py tests that assert output from query_events_for_display.
---
Nitpick comments:
In `@tests/e2e/groups/stats.sh`:
- Around line 264-273: The test currently runs `eneru monitor --once
--events-only --length 5 --config "$E2E_DIR/config-e2e-stats.yaml"` and then
checks `$?`, but with `set -e` a non-zero exit will abort before the check and
log dump; change the test to run the command inside an `if` conditional (e.g.,
`if eneru ... > /tmp/test31-length.log 2>&1; then ... else ... fi`) so the
failure branch can print the log and exit with failure; update the pass/fail
branches around that `if` and keep the same log file `/tmp/test31-length.log`
and messages.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 175e8f5c-88c2-4eef-aa32-9f7d8c52c49e
📒 Files selected for processing (7)
docs/changelog.mddocs/configuration.mdsrc/eneru/cli.pysrc/eneru/tui.pytests/e2e/groups/stats.shtests/test_cli.pytests/test_tui.py
Comment on lines
+36
to
+38
| - `--time` and `<T>` apply to the graph only. Use `--length` for events. | ||
| - Events panel defaults to priority-only. | ||
| - Live TUI events cap raised from 8 to 30 rows. |
There was a problem hiding this comment.
The live TUI cap is height-bounded, not a flat 30 rows.
Normal mode now uses min(30, visible_panel_rows), so shorter terminals still show fewer than 30 rows. This bullet should say "up to 30 rows" or mention the visible-panel bound explicitly.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@docs/changelog.md` around lines 36 - 38, Update the changelog bullet about
the Live TUI events cap to reflect that it's height-bounded rather than a flat
30 rows: change the line mentioning "Live TUI events cap raised from 8 to 30
rows" to say "Live TUI events cap raised (up to 30 rows, bounded by
visible_panel_rows)" or explicitly note it uses min(30, visible_panel_rows) in
normal mode so shorter terminals show fewer than 30 rows.
Comment on lines
439
to
446
| Example package commands: | ||
|
|
||
| ```bash | ||
| sudo eneru validate --config /etc/ups-monitor/config.yaml | ||
| sudo eneru run --dry-run --config /etc/ups-monitor/config.yaml | ||
| sudo eneru monitor --once --events-only --config /etc/ups-monitor/config.yaml | ||
| sudo /opt/ups-monitor/eneru.py monitor --once --events-only --verbose --full-history --config /etc/ups-monitor/config.yaml | ||
| sudo /opt/ups-monitor/eneru.py monitor --once --events-only --verbose --length 100 --config /etc/ups-monitor/config.yaml | ||
| ``` |
There was a problem hiding this comment.
Keep the package-user examples on one invocation style.
This block is headed as package-user examples, but it now mixes eneru ... and /opt/ups-monitor/eneru.py .... Please make the whole block use the packaged wrapper consistently.
Suggested change
-sudo eneru validate --config /etc/ups-monitor/config.yaml
-sudo eneru run --dry-run --config /etc/ups-monitor/config.yaml
-sudo eneru monitor --once --events-only --config /etc/ups-monitor/config.yaml
+sudo /opt/ups-monitor/eneru.py validate --config /etc/ups-monitor/config.yaml
+sudo /opt/ups-monitor/eneru.py run --dry-run --config /etc/ups-monitor/config.yaml
+sudo /opt/ups-monitor/eneru.py monitor --once --events-only --config /etc/ups-monitor/config.yaml
sudo /opt/ups-monitor/eneru.py monitor --once --events-only --verbose --length 100 --config /etc/ups-monitor/config.yamlAs per coding guidelines "When writing documentation, use /opt/ups-monitor/eneru.py invocation style for package users (README, troubleshooting), python -m eneru or eneru for developers (CONTRIBUTING, testing), and eneru for PyPI users".
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| Example package commands: | |
| ```bash | |
| sudo eneru validate --config /etc/ups-monitor/config.yaml | |
| sudo eneru run --dry-run --config /etc/ups-monitor/config.yaml | |
| sudo eneru monitor --once --events-only --config /etc/ups-monitor/config.yaml | |
| sudo /opt/ups-monitor/eneru.py monitor --once --events-only --verbose --full-history --config /etc/ups-monitor/config.yaml | |
| sudo /opt/ups-monitor/eneru.py monitor --once --events-only --verbose --length 100 --config /etc/ups-monitor/config.yaml | |
| ``` | |
| Example package commands: | |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@docs/configuration.md` around lines 439 - 446, The examples block mixes
invocation styles (e.g., "sudo eneru validate ..." vs "sudo
/opt/ups-monitor/eneru.py monitor ..."); update every example so package-user
commands consistently use the packaged wrapper form "/opt/ups-monitor/eneru.py"
(e.g., replace "sudo eneru ..." and any "eneru ..." invocations with "sudo
/opt/ups-monitor/eneru.py ..."), keeping the same flags and arguments (validate,
run, monitor, --dry-run, --once, --events-only, --verbose, --length, --config)
so the block uniformly shows package-user usage.
Comment on lines
+473
to
+489
| # Tiered trim: power events always survive the cap. ``max_events`` | ||
| # in (None, 0) means no cap. | ||
| if max_events and len(rows) > max_events: | ||
| power = [r for r in rows if r[2] in POWER_EVENTS] | ||
| non_power = [r for r in rows if r[2] not in POWER_EVENTS] | ||
| if len(power) >= max_events: | ||
| # Pathological: more power events than the cap. Show the | ||
| # most-recent N -- still better than evicting them. | ||
| kept = power[-max_events:] | ||
| else: | ||
| remaining = max_events - len(power) | ||
| kept_non_power = non_power[-remaining:] | ||
| # Re-sort by ts so the panel never displays out-of-time-order | ||
| # rows after the tier merge. Python's sort is stable, so | ||
| # equal-timestamp rows preserve insertion order (power first). | ||
| kept = sorted(power + kept_non_power, key=lambda r: r[0]) | ||
| rows = kept |
There was a problem hiding this comment.
Keep lifecycle events above verbose chatter when the list is capped.
In verbose mode, non_power currently mixes LIFECYCLE_EVENTS with low-priority chatter, and the [-remaining:] slice is purely recency-based. That means a noisy verbose tail can still push DAEMON_START/DAEMON_RESTARTED out of the capped list, even though the new contract says lifecycle rows fill the remaining slots after power events.
Suggested change
- if max_events and len(rows) > max_events:
- power = [r for r in rows if r[2] in POWER_EVENTS]
- non_power = [r for r in rows if r[2] not in POWER_EVENTS]
+ if max_events and len(rows) > max_events:
+ power = [r for r in rows if r[2] in POWER_EVENTS]
+ lifecycle = [r for r in rows if r[2] in LIFECYCLE_EVENTS]
+ chatter = [
+ r for r in rows
+ if r[2] not in POWER_EVENTS and r[2] not in LIFECYCLE_EVENTS
+ ]
if len(power) >= max_events:
# Pathological: more power events than the cap. Show the
# most-recent N -- still better than evicting them.
kept = power[-max_events:]
else:
remaining = max_events - len(power)
- kept_non_power = non_power[-remaining:]
+ kept_lifecycle = lifecycle[-remaining:]
+ remaining -= len(kept_lifecycle)
+ kept_chatter = chatter[-remaining:] if remaining > 0 else []
# Re-sort by ts so the panel never displays out-of-time-order
# rows after the tier merge. Python's sort is stable, so
# equal-timestamp rows preserve insertion order (power first).
- kept = sorted(power + kept_non_power, key=lambda r: r[0])
+ kept = sorted(power + kept_lifecycle + kept_chatter, key=lambda r: r[0])
rows = kept🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@src/eneru/tui.py` around lines 473 - 489, When capping rows (when max_events
is set) the current logic treats all non-POWER_EVENTS equally so verbose chatter
can evict lifecycle events; change the trim in the block that references rows,
POWER_EVENTS, max_events, kept_non_power and kept so lifecycle rows are
prioritized: split non_power into lifecycle = [r for r in rows if r[2] in
LIFECYCLE_EVENTS] and chatter = [r for r in rows if r[2] not in LIFECYCLE_EVENTS
and r[2] not in POWER_EVENTS], then when computing remaining fill from the most
recent lifecycle rows first (up to remaining) and only then backfill with the
most recent chatter to produce kept_non_power, re-merge with power and sort by
timestamp as before.
Comment on lines
+1240
to
+1252
| lines = query_events_for_display(config, max_events=20) | ||
| # All 3 power-event rows MUST survive the cap. | ||
| assert any("ON_BATTERY: real outage" in l for l in lines), ( | ||
| f"ON_BATTERY pushed off by daemon noise -- tiered trim regressed. " | ||
| f"Got: {lines}" | ||
| ) | ||
| assert any("POWER_RESTORED: outage 10m" in l for l in lines) | ||
| assert any("EMERGENCY_SHUTDOWN_INITIATED" in l for l in lines) | ||
| # Result respects the cap. | ||
| assert len(lines) == 20 | ||
| # Remaining 17 slots are most-recent daemon rows. | ||
| daemon_lines = [l for l in lines if "DAEMON_RESTARTED" in l] | ||
| assert len(daemon_lines) == 17 |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
# Find the test file
fd -type f -name "test_tui.py"Repository: m4r1k/Eneru
Length of output: 226
🏁 Script executed:
# Check if the file exists and get line count
wc -l tests/test_tui.pyRepository: m4r1k/Eneru
Length of output: 78
🏁 Script executed:
# Read the specific line ranges mentioned in the review
sed -n '1240,1252p' tests/test_tui.pyRepository: m4r1k/Eneru
Length of output: 734
🏁 Script executed:
# Check the other mentioned ranges
sed -n '1271,1276p' tests/test_tui.pyRepository: m4r1k/Eneru
Length of output: 412
🏁 Script executed:
# Check the third range
sed -n '1525,1545p' tests/test_tui.pyRepository: m4r1k/Eneru
Length of output: 1021
Rename one-letter comprehension variables to improve readability.
The loop variable l in comprehensions violates Ruff's E741 rule (ambiguous variable name), as single lowercase l is easily confused with 1 or I. Rename to line across all affected test assertions to maintain PEP 8 compliance.
Suggested changes
- assert any("ON_BATTERY: real outage" in l for l in lines), (
+ assert any("ON_BATTERY: real outage" in line for line in lines), (
f"ON_BATTERY pushed off by daemon noise -- tiered trim regressed. "
f"Got: {lines}"
)
- assert any("POWER_RESTORED: outage 10m" in l for l in lines)
- assert any("EMERGENCY_SHUTDOWN_INITIATED" in l for l in lines)
+ assert any("POWER_RESTORED: outage 10m" in line for line in lines)
+ assert any("EMERGENCY_SHUTDOWN_INITIATED" in line for line in lines)
# Result respects the cap.
assert len(lines) == 20
# Remaining 17 slots are most-recent daemon rows.
- daemon_lines = [l for l in lines if "DAEMON_RESTARTED" in l]
+ daemon_lines = [line for line in lines if "DAEMON_RESTARTED" in line]
assert len(daemon_lines) == 17
@@
- assert all("ON_BATTERY" in l for l in lines)
- assert any("outage-49" in l for l in lines) # latest power event
- assert any("outage-40" in l for l in lines) # 10th-latest
+ assert all("ON_BATTERY" in line for line in lines)
+ assert any("outage-49" in line for line in lines) # latest power event
+ assert any("outage-40" in line for line in lines) # 10th-latest
@@
- lines = [l for l in out.splitlines() if "DAEMON_START" in l]
+ lines = [line for line in out.splitlines() if "DAEMON_START" in line]
assert len(lines) == 5
assert "row-49" in lines[-1]
assert "row-45" in lines[0]
@@
- lines = [l for l in out.splitlines() if "DAEMON_START" in l]
+ lines = [line for line in out.splitlines() if "DAEMON_START" in line]
assert len(lines) == 40, f"length=0 must show all rows; got {len(lines)}"🧰 Tools
🪛 Ruff (0.15.12)
[error] 1242-1242: Ambiguous variable name: l
(E741)
[error] 1246-1246: Ambiguous variable name: l
(E741)
[error] 1247-1247: Ambiguous variable name: l
(E741)
[error] 1251-1251: Ambiguous variable name: l
(E741)
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@tests/test_tui.py` around lines 1240 - 1252, Rename the one-letter
comprehension variable "l" to a clearer name like "line" in the test assertions
around the query_events_for_display call: update the three asserts that use "l"
in their any() comprehensions and the daemon_lines list comprehension to use
"line" instead (e.g., any("ON_BATTERY: real outage" in line for line in lines)
and daemon_lines = [line for line in lines if "DAEMON_RESTARTED" in line]); this
will satisfy Ruff E741 and improve readability without changing test logic in
the test_tui.py tests that assert output from query_events_for_display.
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
Addresses CodeRabbit's 5 findings (cubic.dev was clean): 1. docs/changelog.md — clarify the live TUI events cap is now min(30, visible_panel_rows), not a flat 30, since the second commit made it height-bound. 2. docs/configuration.md — make all four example commands use the same `sudo eneru ...` invocation form. Earlier draft mixed the wrapper-path form with the bare command; maintainer prefers the bare form (PATH-resolved) consistently. 3. src/eneru/tui.py — three-tier trim instead of two: POWER_EVENTS always survive; LIFECYCLE_EVENTS fill next; chatter rounds out only if room remains. In --verbose mode the previous code let chatter evict lifecycle rows when the cap was hit; the new tier matches the user-facing priority hierarchy. 4. tests/test_tui.py — rename single-letter loop var `l` to `line` in five spots (Ruff E741). 5. tests/e2e/groups/stats.sh — wrap the `--length 5` invocation in `if cmd; then ... fi`. Under `set -e` the previous bare `[ $? -ne 0 ]` branch would never run because the script aborts on the first non-zero exit, masking any regression with no log dump. Plus one new unit test (`test_tiered_trim_lifecycle_outranks_chatter`) covering the new 3-tier semantics: 1 power + 30 chatter + 4 lifecycle at cap=7 produces 1 power + 4 lifecycle + 2 chatter, NOT 1 power + 6 chatter. 982 unit tests pass (was 981). Co-Authored-By: Claude Opus 4.7 <[email protected]>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Three real bugs in the unreleased v5.2.2 surfaced when testing against the maintainer's own data (65 daemon-lifecycle events vs 5 power events in the SQLite events table — the priority filter as it shipped was actually pushing power events off-screen).
Bugs fixed
Events panel showed daemon noise instead of power events.
PRIORITY_EVENTSlumped power transitions and daemon-lifecycle into one undifferentiated set, so the most-recent N rows in the cap were allDAEMON_RESTARTED/DAEMON_UPGRADEDfrom package iteration. Power events from days ago got pushed off-screen.Fix: split into
POWER_EVENTS(always preserved within cap) andLIFECYCLE_EVENTS(fill remaining slots).query_events_for_displaytrim logic preserves all power events first, then fills with most-recent lifecycle events. Daemon noise can no longer push power events off-screen.eneru tui --once --events-onlysilently fell back to log parsing for sparse-event installs. The default 1h time window was almost always empty, so SQLite returned[]and the code parsed/var/log/ups-monitor.loginstead. Live TUI and--onceshowed structurally different content and the operator never noticed.Fix: events query no longer takes a time-range parameter at all. Always queries the full events table (events are sparse, this is fine).
--timeaccidentally scoped both the graph window AND the events panel. Cause of fix: correct legacy Discord config detection in validate_config #2.Fix:
--timeand<T>are graph-only now. Anchor comments attui.py(TIME_RANGES) andcli.py(--timeflag def) name the bug they prevent so this can't be re-coupled by accident.Knob change
--full-history(only added in unreleased 5.2.2; replaced cleanly by--length 0).--length Nfor--once: cap events output (default 30,0= no cap; argparse type validator rejects negatives + non-numerics).min(EVENTS_MAX_ROWS_NORMAL, visible_panel_rows)). Without this, the data cap was 30 but the panel only renders ~11 rows on a 24-row terminal, so the tier-trim's preserved power events sat above the visible window.<M>still expands to 500 rows.Sequencing
This rides on the existing unreleased v5.2.2 in main. No version bump or changelog re-tag needed — the
[5.2.2]entry is updated in this PR to reflect the final design (tier-trim,--length, no--full-history). Maintainer asked the changelog be kept tiny since this is a bugfix release.Test plan
pytest -q --timeout=30in uv venv); 13 new tests covering tier preservation (incl. the maintainer's exact reproducer: 60 daemon + 3 power, asserts all 3 power events survive cap=20), pathological "more power events than cap",--lengthvalidation, and snapshot-path tail behaviordnf reinstall):eneru tui(live, 24×80) shows the 5 power events at top of panel;--once --events-only --length 5returns the 5 most-recent power events;--length -1rejects with exit 2;--full-historyrejected as unknown by argparse--length 0) addressed before pushStatsexercises the new--lengthvalidation + priority-only default🤖 Generated with Claude Code
Summary by cubic
Fixes the TUI events list with three-tier priority so power events never get pushed off-screen, and makes
--timegraph-only. Adds--lengthfor--once, removes--full-history, and sizes live events to the visible rows.Bug Fixes
--verbose. Prevents chatter from evicting lifecycle rows in verbose mode.--timeand<T>now affect only the graph.min(30, visible_panel_rows);<M>still expands to 500).Migration
--full-historywith--length 0.--timeto size events, switch to--length.--verbose.Written for commit 7cdd1e7. Summary will update on new commits. Review in cubic
Summary by CodeRabbit
Release Notes
Documentation
--timecontrols only graph time range, not events queryChanged
--full-historywith new--length Nparameter to cap event output--verbose/-vnow includes low-priority eventsBug Fixes
--events-onlynow falls back to log parsing when default window is empty