materialization: pre-build Filter results per partition, with UI, poller and cache fixes by TheWitness · Pull Request #298 · Cacti/plugin_flowview

TheWitness · 2026-10-22T11:48:49Z

Summary

This branch adds materialization to FlowView: pre-computing each Filter's per-partition results into the Parallel Query shard cache, so reports over older data open quickly. It also fixes the bugs that turned up while running it on a large install (about 1,470 hourly partitions). The original time-shift arrow fix is still included.

What changed

Materialization

cli/materialize_queries.php (new) materializes every saved Filter, or --query-id=ID[,ID...], over the cold partitions, with NAT on and off.
- It runs only the partitions a Filter does not have cached yet. Each batch is its own Parallel Query and is sized from Max Concurrent Threads, because one query over every partition can't finish within the Maximum Run Time.
- A batch that times out is followed by a smaller one. What was cached is kept, so a re-run resumes where the last one stopped.
- Options: --partitions, --nat=on|off, --dry-run, --force, --status, --debug.
- Before scanning, it waits for any running partition ALTER TABLE and then applies the pending ones itself, using a shared planner (flowview_get_partition_alter()) that flowview_upgrade.php now also uses.
- Each Filter is registered in processes as materialize with the Filter id as the task id, so Filters can be materialized separately and at the same time.
- SIGTERM and SIGINT stop its workers and the parallel queries they started, and remove the registrations, as the flow collector does.
- When there is no terminal (started from the UI or the poller), its progress goes to the Cacti log.
Automatic: each poller cycle materializes the partitions that rotated since the previous cycle.
The live (newest) partition is never cached, because its result would be wrong as soon as more flows arrive. Instead, the table and charts of one report share a single scan of it for 60 seconds, behind a named lock, stored under map_partition = 'live'.

Filters page (`flowview_filters.php`)

New columns:
- Materialization: a Required / Running / Up to date pill, plus a "Running: N" pill, with tips listing the missing partitions.
- NAT Shards, Non-NAT Shards and Partitions.
The ID column now sits right after Filter Name.
New Materialize action. It skips Filters that are already being materialized and waits 2 seconds before redirecting, so the Running state shows.
A line under the table counts the cached shards that match a Filter (by md5 cache key) and those that match none.
A Prune button, after Clear, opens a non-modal jQuery UI dialog and POSTs action=prune_shards, which removes only the shards that match no Filter.
Saving a Filter with a change that affects its cache now:
1. opens a jQuery UI warning dialog with Cancel / Continue first;
2. on Continue, saves and then shows a Save Succeeded dialog telling the user to run Materialize.
This replaces the WARN message that Cacti showed as "Some Actions failed".

Report page (`flowview.php`)

The Report selector is removed for a Filter that has a report, because the report is part of the cache key. The server takes the report from the Filter for the page, the charts and the session, so a tampered request can't change it.
The time-shift backward/forward icons now work. Previously they called timeshiftFilterLeft/Right, which were undefined.

Fixes found along the way

Go service (service/flowview-query/scheduler.go): expireShardCache deleted every cached shard older than the Cached Data Time to Live (6 hours by default), which emptied the cache overnight. It now expires only the 60-second live rows. Cold shards are removed only when their partition is dropped.
Poller:
- It now converts every cold partition still on InnoDB to the fast engine (3 per cycle, never while an ALTER is running). Before, it only looked at the 2nd to 4th newest, so missed partitions stayed InnoDB.
- It does not start conversions while a Filter is being materialized.
Process names:
- Timeout cleanup and the runner's signal handler looked for db_query<id> and db_shard<id>, but the processes register as db_query_<id> and db_shard_<id>, so they were never found.
- parallel_database_query_cleanup() used SIGTERM, which is undefined under the web server. That was a fatal error that disabled the plugin. Signals now go through flowview_signal_process().
Request reuse:
- A finished request with the same md5 was reused by a materialization, so nothing was rescanned. Finished requests are now discarded first.
- A request left scheduled by a timeout made every later run wait out the full limit. Unfinished requests with no runner that are more than 2 minutes old are now removed.
flowview_runner.php loads setup.php itself, so it no longer depends on the plugin hooks to define flowview_connect().
Cacti 1.2.32: drop_multi already renders as select2, so the Filter editor no longer also calls .multiselect() on Protocols, which drew two controls. Earlier Cacti versions keep jquery-multiselect.
__() values are now passed before the flowview text domain in the new strings.
A printed report's title used the wrong array index.
The materializer pads its range start by a day, so the oldest partition gets cached.

Housekeeping

fixprivate.php, import_flows.php and repair_tables.php moved to cli/. They are tombstoned in manifest.json so upgrades remove the old copies.
README updated. CHANGELOG lists the features only, since nothing has been released yet.

Reviewers should know

The Go service must be rebuilt and restarted. Until it is, an old binary keeps purging the shard cache after 6 hours.
The first materialization on a large install takes a while. It runs in batches and can be re-run to continue.
Known core gap, not part of this PR: Cacti's deepness theme does not skin jquery-multiselect.
- Its main.css styles .ui-button, .ui-selectmenu-* and select2, but not .ui-multiselect / .ui-multiselect-menu.
- Its jquery.multiselect.css is identical to the dark theme's.
- So pages still using jquery-multiselect look like the dark theme. The fix belongs in Cacti core.
Two Filters with the same map query share one cache key, so their shard counts are identical. That is expected.

Testing

php -l on every changed file; tests/bin/validate-manifest.php passes.
Pest: tests/Unit + tests/Security, 305 passed. New tests cover materialization batching, request reuse, live-shard sharing, prune, signal handling, the per-Filter registration, the save dialogs and the select2 gating.
Go: go vet and go test ./... pass.
The signal handling, batch loop and generated dialog JavaScript were checked with stub harnesses and node --check.
Exercised manually on a live install during development. The latest dialog and select2 changes still need a browser check.

The filter's timeshiftBackward/timeshiftForward handlers called timeshiftFilterLeft()/timeshiftFilterRight(), which neither the plugin nor Cacti core defines, causing an uncaught ReferenceError. Add a timeshift action that shifts date1/date2 by the selected interval and a plugin-owned timeshiftFilter() that applies it and reloads the report. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The time shift callback called selectmenu('refresh') unguarded, which throws when the timespan select is not an initialized selectmenu, aborting before applyFilter() ran. Use a guarded setCustomTimespan() helper (instance check, as core does) for both the time shift and the date change handler. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Add a CLI script that runs every saved query through the parallel query for the entire range of flow partitions with NAT on and off. It registers a process with an 8 hour limit, reuses cached partitions that still match their data, and rebuilds stale or dropped ones. Run each query in its own process so reaching the run limit does not end the rest. Move fixprivate.php, import_flows.php and repair_tables.php into cli/, tombstone the old paths in manifest.json, and locate the Cacti root relative to each script. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ached shard counts Add a Materialize View action to the Filters page that, after confirmation, starts cli/materialize_queries.php in the background for the selected filters; --query-id now accepts a comma separated list. Recommend a forced re-materialization when a filter is saved with a change that alters what it materializes, from both the filter form and the report page. Show the number of cached partitions with NAT and without for each filter next to its ID. run_flow_query() gains a plan-only mode that returns the cache keys without running the query, and the map query hashing is shared with the parallel query request. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Add a Partitions column after the NAT and Non-NAT shard columns so the progress of a materialization can be followed. Count only genuine raw partition tables, and use the same helper in cli/materialize_queries.php so its partition count matches the page. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…on of rotated partitions Add a Status column to the Filters page showing whether the automatic materialization of each filter is required, running or up to date. Each poller cycle now records the newest partition and starts cli/materialize_queries.php in the background for the partitions that went cold since the previous cycle, for every filter with NAT on and off. The script gains a --partitions option, and the cache check can be limited to those partitions. The cached shard lookup now returns the partitions. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…d show status pills Have cli/materialize_queries.php wait for any ALTER TABLE already running on the partitions it needs and do the alters still due before the parallel query starts, checking each partition again just before altering it. What a partition needs altered is decided by flowview_get_partition_alter(), which the schema upgrade now also uses. Skip the poller's engine conversion while an ALTER TABLE is already running on the partition, so each cycle no longer queues another rebuild of the same table. Print a parallel query timeout in words when run from a script instead of the web page's JavaScript. Render the Filters page status as a pill in the form of the device status. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…per filter The Running status only looked at the materializer's own registration, so it disappeared when the script ended, even though the parallel query it started kept working. Look for the runner and shard worker processes of requests made by a script too, ignoring rows left behind by processes that are gone, and do not start a new materialization while one is running. Make the parallel query cleanup and the runner's termination handler use the names the processes register, db_query_<id> and db_shard_<id>, so a timeout actually stops them and a terminated one removes its row. Show the number of running queries of each filter in a pill beside the status pill, matched to the filter by its cache key. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

… they are for Take the running parallel queries from the db_query_<id> and db_shard_<id> registrations in the process table, whoever started them, counting a registration while it is recent or its process is alive. This no longer depends on the request belonging to user 0. Match each running request to its filter by cache key and show the number of running queries per filter, with the NAT and non-NAT split in the tip. Add --status to cli/materialize_queries.php to list the running requests, the filter each is for and their progress. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…olumn Stop cli/materialize_queries.php from showing each ALTER TABLE twice: the log message was also printed, as well as the script's own progress line. The alter is still logged, and the schema upgrade still prints it. Rename the Filters page Status column to Materialization and explain in its tip what materialization is and what the pills and the Running count mean. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ll and timeout The Running: N pill tip passed only the text domain to __esc() for a text with placeholders, which Cacti rejects with an ArgumentCountError and broke the Filters page whenever a query was running. The pill label and the Parallel Query timeout message had the same mistake and would have shown the domain and 0 seconds. Pass the values first, as the rest of the plugin does. Make the unit test __() and __esc() stubs treat every argument as a sprintf argument like Cacti, so a mistake like this fails a test. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

A shard is only cached when the range starts before the earliest flow in the partition. Flows can start before the window of their partition, but the range was padded by only a second, so the oldest partition, or a rotated one materialized on its own, was never cached and its filter stayed Required. Pad the start by a day, and report the first and last partition in the summary. Say which partitions are not materialized, with NAT and without, in the tip of the Required and Running pill. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The partition still receiving flows was scanned as a full scan by a full run and its result cached, which is wrong as soon as more flows arrive, and it was counted as materialized. Decide whether a shard is a full scan in flowview_is_full_scan(), which refuses the newest partition, so no query can cache it. cli/materialize_queries.php now only materializes partitions that have rotated, and removes anything cached for the live partition before. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The live partition is no longer cached, so the table and every chart of a report scanned it again with the same map query. Workers running the same map query and range against it now take a named lock and the first keeps its result for 60 seconds for the others, under map_partition 'live' so it is never counted as materialized. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…rs materialized, convert every cold partition The Go query service aged out every cached partition after the Cached Data Time to Live, emptying the NAT and non-NAT shards overnight, so it only expires the short lived live shard results now. The Filters page showed every filter as Running while any materialization ran, so it uses the filters the run is for. The poller only converted the 2nd to 4th newest partitions, so any it missed stayed InnoDB, so it converts every cold partition, three per cycle and one converter at a time. flowview_runner.php now loads setup.php itself instead of relying on the plugin hooks having loaded it. The bug entries of this session are removed from the CHANGELOG as nothing has been released. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…for cached shards that match no filter The report is part of the cache key, so the Report selector is disabled on the report page for a filter that has one, and the server ignores a report sent for it. The Filters page tells the cached shards that match a filter from the others by md5, and a Prune button removes the others after confirmation. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…tl constants from the web server cli/materialize_queries.php now handles SIGTERM and SIGINT like the flow collector: the script stops its worker and removes its process registration, and a worker cancels its parallel queries, which stops their runner and shard workers and removes their registrations. Requests are tracked per process for that. parallel_database_query_cleanup() used SIGTERM, which does not exist under the web server without pcntl, so a timed out report was a fatal error that disabled the plugin; signals are now sent through flowview_signal_process(). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…p the poller converting under it Each filter a materialization works on is registered in the process table as materialize with the filter id as the task id, so filters can be materialized separately and the Filters page reads which are running from the process table instead of a setting. The poller no longer starts ALTER TABLE conversions while a filter is being materialized, as the materializer waits for any alter on the partitions it needs and the poller kept one running, and the materializer logs its progress to the Cacti log when it has no terminal. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…d request, and add Prune to the Filters actions A parallel query request is reused for the same query and range, and a finished one only hands back its result, so a materialization after the shard cache was lost finished at once and cached nothing. The materializer now discards a finished request of the same query first. Prune Cached Shards is also an action in the Filters page dropdown, as well as the Prune button. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The Report selector is no longer shown, only a hidden input id=report that carries the value the requests use, unless the selected filter has no report and needs one. The title of a printed report named the report by the statistical index. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…e filter No Report control or field is on the page for a filter that has a report. The requests no longer send it, the server takes it from the filter for the page, the charts and the session, a printed report has its panel checkboxes disabled by the server, and Save no longer changes the report of a filter. A filter with no report still shows the selector as it needs one. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

… left behind by a timeout One parallel query for every partition can not finish within the Maximum Run Time, and a timeout leaves the request behind as scheduled, so the next run found it, started nothing and waited again. The materializer now runs only the partitions a filter has not cached, a batch at a time, with smaller batches after a timeout, and a request that is not finished, has no runner and is old enough is removed when the same query is requested again. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ore filter editing Pruning is for the whole cache, so it is no longer an action of the Filters dropdown or a confirmation page. The Prune button after Clear opens a modal jQuery UI dialog saying what will be removed and kept, and its Prune button posts action=prune_shards, which only acts on a POST. Adding the prune actions had dropped the case 'edit' label, so editing a Filter from the Filters page showed the list instead. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…nd confirm the save in a dialog Saving a Filter with a change to what is cached for it raised a warning message, which Cacti showed as 'Some Actions failed'. The Save button now asks the server whether the change is material and, if so, opens a jQuery UI dialog with Cancel and Continue before the form is posted. Once saved, a 'Save Succeeded' dialog says to start a materialization with the Materialize View action. Other saves keep Cacti's usual Save Successful message. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

… the Filters page Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…p jquery-multiselect before it From 1.2.32 Cacti's drop_multi renders a select2 'N Selected' control itself, and the Filter editor also made it a jquery-multiselect, so both widgets were drawn. The editor now only calls .multiselect() before 1.2.32, or when core did not decorate the select for select2, and the Protocols field shows 'All Protocols' when none are picked. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The messages and documentation that name the action are renamed with it. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

🟡 Changes recommended

Worker cancellation, batch partition scoping, concurrent schema preparation, and printed-report panel state have unresolved correctness defects.

4 open issues

What changed in this PR

This PR adds per-partition materialization to FlowView’s Parallel Query cache, reducing report load times for historical flow data.

Changes:

Adds resumable materialization through a CLI, Filter actions, and automatic poller scheduling.
Adds cache status, pruning, and save dialogs; updates report controls and time shifting.
Revises cache expiration, process cleanup, and partition preparation, with supporting tests and documentation.

File	Description
`tests/Unit/FlowviewFilterControlsTest.php`	Tests report selection and printed-report controls.
`tests/Unit/CoreFunctionsTest.php`	Adds materialization, cache, process, and UI regression coverage.
`tests/bootstrap-unit.php`	Updates translation and logging stubs.
`tests/bin/patch-coverage.php`	Updates CLI entry-point exclusions.
`service/flowview-query/scheduler.go`	Limits shard expiration to short-lived live results.
`service/flowview-query/scheduler_test.go`	Tests revised shard expiration behavior.
`service/flowview-query/README.md`	Documents cold-cache retention and live-result expiration.
`README.md`	Documents materialization commands and UI workflows.
`manifest.json`	Tracks relocated CLI utilities and obsolete paths.
`includes/arrays.php`	Adds the select2 Protocols placeholder.
`flowview.php`	Adds time shifting and enforces saved report settings.
`flowview_upgrade.php`	Reuses the shared partition-alter planner.
`flowview_runner.php`	Loads required helpers and corrects process cleanup names.
`flowview_process.php`	Converts cold partitions and schedules materialization.
`flowview_filters.php`	Adds materialization actions, status, counts, and pruning.
`cli/repair_tables.php`	Updates bootstrap paths after relocation.
`cli/materialize_queries.php`	Implements batching, preparation, status, and worker management.
`cli/import_flows.php`	Updates bootstrap paths after relocation.
`cli/fixprivate.php`	Updates bootstrap paths after relocation.
`CHANGELOG.md`	Records materialization features and CLI relocation.
`.github/copilot-instructions.md`	Documents the CLI directory layout.

_{Review effort: Balanced}

Every alter of a raw partition now goes through flowview_alter_partition(), which holds a per-table named lock and works out the plan after taking it, so materializers of different Filters, the poller and the schema upgrade never apply a stale plan to a table another process already altered. The poller does not wait for the lock. A materialization worker is given its batch's partitions and scans only those, instead of every partition in the padded range, and it is started with proc_open() from an argument list, not through a shell, so SIGTERM reaches it. A printed report's saved panels are made to agree with the table-only request before they are merged, so they no longer re-enable the charts. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

display_domain() returned the resolved source and destination names as they were stored, and the report table printed them as HTML. They come from reverse lookups of the flow data, which whoever controls a reverse zone can set, so a crafted PTR name could inject markup. They are now escaped with html_escape(). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

… validate the stream override run_flow_query() returned a Filter's cache keys in plan mode, so its return value could, as far as static analysis can tell, carry request values to the report table, the summary and the CSV export. The plan is now given through an out parameter behind flowview_plan_filter(), and what run_flow_query() returns is only ever the report. The ex_addr stream override was bound to the query as it came; it is now only taken when it is an IP address, as exporter addresses are, and otherwise the Filter's own stream is used. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The pull request gate requires every changed line that the unit suite measures to be covered. This adds tests for the Filter editor and its saves, the chart data, planning a Filter's cache keys, the parallel and serial report paths, the request and shard lifecycle, the live partition shard sharing, a query that times out, and the partition bookkeeping. To make those paths testable: - flowview_exit() wraps exit so the unit bootstrap can see it instead of the process ending; save_filter(), flowview_get_chartdata() and the parallel query timeout use it. - includes/arrays.php is measured, as it changed. - The unreachable "No valid report was selected" branch is removed, the report is known to be statistical or printed by then, and cacti.pot is regenerated. - The printed report branch inside the two map query loop, which only runs for statistical reports, is removed. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The web page branch of the timeout report depends on CACTI_WEB, a constant a unit test cannot change, so the report moves into flowview_report_parallel_timeout() with whether it is a web request as an argument, and both ways are tested. The cleanup of a request's map table and db_debug() are covered too, as the change moves the diff onto them. The test stub of form_start() records its arguments instead of printing them, which SonarCloud took for a reflected XSS. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

sonarqubecloud · 2026-10-22T13:26:48Z

Quality Gate passed

Issues
0 New issues
1 Accepted issue

Measures
0 Security Hotspots
0.0% Coverage on New Code
1.5% Duplication on New Code

See analysis details on SonarQube Cloud

…8442) * fix(themes): skin jQuery multiselect in the Deepness theme The Deepness main.css repainted select2 and the jQuery UI selectmenu onto its --dp-* palette but never the jQuery multiselect button or popup menu, so those fell through to the jquery-ui.css textured surfaces and the generic uppercase button rule and looked like the Dark theme. Extend main.css (the stock jquery.multiselect.css is left untouched) so the multiselect button, menu, header, hover/active/disabled labels and group labels use the same colors as the theme's select2 controls. Reported in Cacti/plugin_flowview#298. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(themes): drop the ineffective Deepness multiselect border-radius .ui-corner-all forces border-radius with !important in Deepness, and the multiselect button carries that class, so the declaration never applied. Let the button follow the theme's jQuery UI corner rule like the other widgets. Also tag the CHANGELOG entry with the pull-request number. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: TheWitness <copilot@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…n four themes (develop) (#8443) * fix(themes): skin jQuery multiselect in Deepness and the 3.x markup in four themes Deepness: main.css repainted select2 onto its --dp-* palette but never the jQuery multiselect button or popup menu, so those fell through to the jquery-ui.css textured surfaces and the generic uppercase button rule and looked like the Dark theme. Repaint them to match the select2 controls. Cacti, Carrot, Hollyberry, Raspberry: these themes ship the old 1.x jquery.multiselect.css while include/js/jquery.multiselect.js is 3.x, so optgroup menus showed list bullets, a nested indent, compounded 0.9em option text and an unstyled group label. Add the missing 3.x rules. All changes extend each theme's main.css; the stock jquery.multiselect.css files are left untouched. Reported in Cacti/plugin_flowview#298. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(themes): satisfy the develop CHANGELOG and theme coverage gates - Move the CHANGELOG entries into changelog.d/8443.issue, as develop expects fragments rather than direct CHANGELOG edits. - Drop the Deepness multiselect border-radius; .ui-corner-all forces border-radius with !important in this theme, so it never applied. - Add Chromium computed-style tests for the new multiselect rules in Deepness, Cacti, Carrot, Hollyberry and Raspberry so the changed CSS lines are exercised by the theme correctness suite. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: TheWitness <copilot@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

TheWitness and others added 30 commits October 20, 2026 14:02

fix: Wrap checkboxes in spans for 1.2.32

b03c757

fix: Format the shard and partition counts with number_format_i18n

9e7e05d

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

fix: Put the ID column right after the Filter Name

a18e47b

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

fix: Open the prune confirmation dialog as non-modal

39638d9

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

fix: Wait two seconds after starting a materialization before showing…

f19ee56

… the Filters page Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

fix: Rename the Filters action Materialize View to Materialize

7d5e1a4

The messages and documentation that name the action are renamed with it. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot AI balanced review requested due to automatic review settings October 22, 2026 11:48

Copilot started reviewing on behalf of TheWitness October 22, 2026 11:49

View session

github-advanced-security Bot found potential problems Oct 22, 2026

View reviewed changes

Comment thread tests/Unit/CoreFunctionsTest.php Fixed

Copilot AI reviewed Oct 22, 2026

View reviewed changes

Comment thread cli/materialize_queries.php Outdated

Comment thread cli/materialize_queries.php

Comment thread cli/materialize_queries.php Outdated

Comment thread flowview.php

TheWitness and others added 3 commits October 22, 2026 08:18

xmacan previously approved these changes Oct 22, 2026

View reviewed changes

locales: Regenerate cacti.pot for the materialization strings

33b6485

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

TheWitness dismissed xmacan’s stale review via 01280ad October 22, 2026 13:19

TheWitness force-pushed the fix-toggles branch from 01280ad to 139592d Compare October 22, 2026 13:19

github-advanced-security Bot found potential problems Oct 22, 2026

View reviewed changes

Comment thread tests/Unit/FilterEditAndSaveTest.php Fixed

TheWitness requested a review from xmacan October 22, 2026 13:30

xmacan approved these changes Oct 22, 2026

View reviewed changes

TheWitness merged commit 772a94a into develop Oct 22, 2026
6 checks passed

TheWitness deleted the fix-toggles branch October 22, 2026 16:00

This was referenced Oct 22, 2026

fix(themes): skin jQuery multiselect in the Deepness theme (1.2.x) Cacti/cacti#8442

Merged

fix(themes): skin jQuery multiselect in Deepness and fix 3.x markup in four themes (develop) Cacti/cacti#8443

Merged

Navigation Menu

Repository navigation

materialization: pre-build Filter results per partition, with UI, poller and cache fixes - #298

materialization: pre-build Filter results per partition, with UI, poller and cache fixes#298
TheWitness merged 37 commits into
developfrom
fix-toggles

TheWitness commented Oct 22, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sonarqubecloud Bot commented Oct 22, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

TheWitness commented Oct 22, 2026

Summary

What changed

Materialization

Filters page (flowview_filters.php)

Report page (flowview.php)

Fixes found along the way

Housekeeping

Reviewers should know

Testing

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sonarqubecloud Bot commented Oct 22, 2026

Quality Gate passed

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Filters page (`flowview_filters.php`)

Report page (`flowview.php`)