.claude/skills/system-management/tasks/load-history.md

@@ -21,8 +21,10 @@ item storage group of type *Event/Item* or *Event/Event*. The sample project pro
   when an older unit's time range covers them (verified 2026-10-01, ISS-006). A restart closes
   the active unit and starts a new one, so load old data **before** restarting, or not at all.
   Check the unit times first: skill `history-management`, task `history-checks.md`.
-- A dropped sample can still become the initial value of the next unit at the next rollover
+- A dropped sample dated **shortly before** the first unit after the restart (same day) becomes
-  (ISS-038): check the result after loading.
+  the initial value of the next unit at the next rollover, instead of the item's live value
+  (reproduced twice; a sample one day older did not do this, ISS-038). Check the result after
+  loading, and do not load samples older than the active units.
 - Loaded samples appear in `ITEM_HISTORY` a few seconds after `ithloahis` returns.
 
 ## Input files

knowledge/issues.md

@@ -57,7 +57,7 @@ development system (see "Test system" in `progress.md`); they do not exist on ot
 | ISS-035 | 2026-10-01 | ALARM (`ALDPRT`) | `ALDPRT` is running and connected (`almdbg` 1 shows `Events-send-to` rising with each alarm), `aldprt.sup` sets no log destination (so per IM 36K10L15 §4.3 it should log via UMH) and `EVENTS` includes `alarm`/`ackn`, but no alarm line appears in `tls\log\umhlog.log` for the CI-Eng-Sample HIGH / HIGH HIGH / acknowledge events. No other log file changed. Cause unknown (UMH filtering? output only with `LOGUMH` set explicitly?). | open | Ask the user whether alarm lines were ever seen in the message log; possibly test `LOGUMH` (or `LOGSCN`) in `aldprt.sup` (admin, restart) in a later session. Low priority: history and overviews are not affected. | `.claude/skills/alarm-management/tasks/item-alarming.md` (last line); `knowledge/manuals/IM36K10L15-01EN.md` §8 |
 | ISS-036 | 2026-10-01 | ALARM (setup) | Installed vs manuals: `alh.sup` has `DB-JRNL = WAL` and `DB-SYNC = OFF`, not described in IM 36K10L15 §3.4; storage group `ALH` has lifetime **52 weeks** while IM 36K10C20 §A7 says the default is 2 days (may have been changed on this system earlier — unknown); `aldprt.sup` uses `TIMEFMT` where the manual text also writes `TIMFMT`. | closed | Installed software is the reference (standing user decision). Close when the user confirms. Closed 2026-10-01 (confirmed by user). | `knowledge/manuals/IM36K10L15-01EN.md` §7, §8; `knowledge/manuals/IM36K10C20-01EN.md` §A7 |
 | ISS-037 | 2026-10-01 | ALARM (history fields) | `ALARM_CHRONO` records contain leftover or unreadable text: INFO events generated with `almdbg` 16 carry `INSTALL`/`UNIT`/`TAG`/`SUB` values from earlier events and non-ASCII bytes in `SOURCE`, `CATEGORY`, `CONDITION`; item events of `CI-Eng-Sample.TANK1.LEVEL` show `SUB` = `RoC` although the item has only the sub-item `MAX`. | open | Check whether the alarm overview shows the same (phase 10, needs a display). Treat these columns as unreliable for non-OPC events until then. | `.claude/skills/alarm-management/tasks/alarm-checks.md` §4 |
-| ISS-038 | 2026-10-01 | HISTORY (`ithloahis` + rollover) | An `ithloahis` sample that was dropped (dated 11:50, before the active unit) still became the **initial value** of the next unit: after a forced rollover at 12:05:26 `ITEM_HISTORY` shows `CI-Eng-Sample.TANK1.INFLOW` = 95 (`INITIAL_VALUE` 1) although the item's value was 112. History after loading old samples can therefore show a wrong value at the next rollover. | open | Accept as a caution (documented in `load-history.md`), or retest after a CI Server restart to see whether the wrong initial value persists. | `.claude/skills/system-management/tasks/load-history.md`; `knowledge/manuals/IM36K10C40-01EN.md` §3 |
+| ISS-038 | 2026-10-01 | HISTORY (`ithloahis` + rollover) | An `ithloahis` sample that was dropped (dated 11:50, before the active unit) still became the **initial value** of the next unit: after a forced rollover at 12:05:26 `ITEM_HISTORY` shows `CI-Eng-Sample.TANK1.INFLOW` = 95 (`INITIAL_VALUE` 1) although the item's value was 112. History after loading old samples can therefore show a wrong value at the next rollover. | open | Retested 2026-10-01 after a CI Server restart (first unit 18:05:46). (a) Sample dated one day earlier (95), live value 112, forced rollover: initial value 112 — correct. (b) Sample dated 18:00 (96, a few minutes before the first unit), live value 113, forced rollover: initial value **96** — wrong again; the live item value stayed 113. So a dropped sample dated shortly before the first unit (same day) becomes the next unit's initial value; a much older one did not. Also found: a forced rollover within 2 minutes of the previous one is refused (`DSI-E-ROLLDONE`). Documented as a caution; behaviour confirmed, ready to close. | `.claude/skills/system-management/tasks/load-history.md`; `knowledge/manuals/IM36K10C40-01EN.md` §3 |
 | ISS-039 | 2026-10-01 | HISTORY (setup) | Installed vs manuals: `hisdbg` menu runs to option 22 (manual text: `-1 - 10`, list to 22) and lists history types 4 (`STATION_STAT`) and 36 (`SYS_LOG`) not in IM 36K10L20 §3.1; `ithsag.sup` keyword `run_interval` (manual: *Heartbeat*) and `queue_size` 500 KB (manual default 200); the option-type table in IM 36K10L20 §3.1 is garbled in the PDF (observed: Scan/Time `1 1`, Event/Item `2 2`, Aggregation `4 4`). | closed | Installed software is the reference (standing user decision). Close when the user confirms. Closed 2026-10-01 (confirmed by user). | `knowledge/manuals/IM36K10L20-01EN.md` §3; `knowledge/manuals/IM36K10C40-01EN.md` §2 |
 | ISS-040 | 2026-10-01 | HISTORY (aggregation) | Aggregation group `CIENG_AGG` (30 min + hour, `TANK1.LEVEL` max/min/avg) ran (`ithsag.log`: *Aggregation Completed*) but stored no values: retrieval gave `0x09 = ITM error` per period. `TANK1.LEVEL` was in `CIENG_EVENT` and `CIENG_SCAN10S`, both priority 0. Also `INTERPOLATION = "30 minutes"` is refused by `dssqld`; the installed name is `30Minutes`. | closed | Closed 2026-10-01 (confirmed by user). 2026-10-01 14:02: after `CIENG_SCAN10S` got `PRIORITY` 10, the 13:30–14:00 period stored max 77, min 33, avg 52.7, quality 1000 (`ithdbg` 20 and `dssqld` with `INTERPOLATION = "30Minutes"`). Cause: equal source priorities (C40 §2.1 says the highest-priority group is the source, but not that a tie gives nothing). Sample fixed (priority 10). The hourly value for 13:00–14:00 was not stored (first half had no source); not rechecked. Proposed close. | `.claude/skills/history-management/tasks/storage-groups.md` §4 (done); `knowledge/manuals/IM36K10C40-01EN.md` §3 (done); `examples/CI-Eng-Sample/03_his_group_df.qli`, `README.md` (done) |
 | ISS-041 | 2026-10-01 | GRAPHICS (HTML trend) | In the HTML display (`localhost:8080/mobileoperator/?d=CIENG_Tank`) the trend shows the pen legend `CI-Eng-Sample.TANK1.LEVEL [CIENG_SCAN10S]` and time axis but draws no line, although `ITEM_HISTORY` has 420 samples in the shown hour. Browser console: `Mismatch between number of labels (Time,CI-Eng-Sample.TANK1.LEVEL [CIENG_SCAN10S]) and number of columns in array (51)`. The trend has 4 further empty pens. 2026-10-01 later: the trend also had the property `chart.attribute[00]` connected to `TANK1.LEVEL` (added by mistake with the pen); the Operator Interface (desktop) showed *Short attribute name ( 61.0) is invalid*. After removing it the desktop trend draws the line, but the HTML trend still logs the mismatch error. | open | HTML only: try a trend with one pen (remove the empty ones) or a Micro Trend; check the HTML-supported trend properties (IM 36K10R30 §4.8.7); check whether the browser still uses a cached copy (`MO_CACHE_DISPLAYS` in `jwuiisg.sup`). | `.claude/skills/graphics/tasks/create-display.md`; `knowledge/manuals/IM36K10R30-01EN.md` §4 |