PlotJuggler — Issue Tracking

Snapshot of the 156 open GitHub issues (as of 2026-04-15) mapped against commits reachable from branch plugin_issues.

• Source list: ~/Downloads/github_issues_summary.md (156 issues)
• Branch under review: plugin_issues (includes merged upstream/main and cherry-picked Feature: StringsSeries allowed)
• Last updated: 2026-04-21

Totals

Category	Open	✅ Fixed	⏳ Waiting
Obsolete / should be closed	10	0	10
Questions / user support	10	0	10
Plugin bugs	49	22	27
Core app	32	12	20
Feature requests	51	2	49
Total	156	36	120

Per-plugin scoreboard

Plugin	Open	✅	⏳
CSV (DataLoad + Toolbox + StatePublisher)	8	8	0
ROS / ROS2 (ParserROS, DataLoad/Stream)	18	0	18
MCAP (DataLoadMCAP)	5	5	0
Lua / Scripting	4	4	0
Parquet (DataLoadParquet)	2	2	0
MQTT (DataStreamMQTT)	3	0	3
ZMQ (DataStreamZMQ, StatePublisherZMQ)	3	2	1
ULog (DataLoadULog)	2	0	2
Protobuf (ParserProtobuf)	1	0	1
UDP (DataStreamUDP)	1	1	0
FFT (ToolboxFFT)	1	0	1
Quaternion (ToolboxQuaternion)	1	0	1

---

✅ Fixed

Plugin bugs — CSV (8 / 8)

#	Title	Evidence
#850	More than one header line in CSV	07451d6e, bd318377
#987	Load a directory tree with CSV files	2d57c875
#1000	Load CSVs from CLI without prefix	c91ccda5
#1068	Select filename on export	toolbox_ui.cpp:151 (QFileDialog::getSaveFileName)
#1290	CSV export missing quaternion/RPY values	ToolboxCSV iterates full _plot_data->numeric — probably fixed; manual smoke test recommended
#1137	Quaternion/RPY not exported (duplicate of #1290)	same as #1290
#1018	Exporting strings to CSV doesn't work	44118575 (cherry-pick of ee91bbc1 Feature: StringsSeries allowed (#1337)) — toolbox_csv.cpp reads _plot_data->strings in 8 places
#662	Multi-file prefix: can't add prefix to individually-added files	c91ccda5 — --auto-prefix CLI flag auto-assigns each file's basename as its prefix (commit message cites #1000 only; scope extends to #662)

Core app (12 / 32)

#	Title	Commit
#749	Multi-file prefix dialog buttons inaccessible	b827b3e7
#464	-l + -d flag data not reloaded	abb6a866
#1052	Plot Docker toolbar "X" segfault	ef5495f0
#1062	Build fails on clang ≥ 19	57fa83a6
#1080	Vertical zoom stuck on max 0.1	e4ee6288
#967	--buffer_size not applied on Ubuntu 20.04	8317c42c
#528	Time Cursor Limits not updated with Time Offset	8317c42c
#1291	Time starts at 01:00:00.0 on time offset	c0707de5
#1014	Layout XML attribute order not deterministic	c3c33c28
#1326	Fields with colons misinterpreted as time	6806a320
#1151	Colormap not saved in layout XML	19d06903
#994	Buffer value not saved in layout	09d7f6b9

Plugin bugs — MCAP (5 / 5 — category complete)

#	Title	Evidence
#1133	Error opening bag files since update to latest version	Likely fixed by 92b862b4 ("much faster MCAP pre-loading and update to 2.1.2", Feb 2026) or subsequent MCAP lib work. Manually verified on 2026-04-20. Single-file verification — revisit if other users re-report with different file variants.
#783	Cannot open mcap bags	Manually verified on 2026-04-20 against a previously failing file. No explicit Fix #783 commit; same MCAP-loader improvements as #1133 likely cover the original report. Generic bug description — if re-reported with a different file, may surface a distinct root cause.
#754	Error reading the mcap file	Manually verified on 2026-04-20 — effectively duplicate of #783; same caveat applies. Worth noting on GitHub as "closed; duplicate of #783" rather than as an independent fix.
#958	MCAP file created on Linux won't open on Windows	Manually verified on 2026-04-20 against a Linux-created MCAP file opened on Windows. No explicit Fix #958 commit; the same MCAP-loader improvements (92b862b4, bc71a2ff, c3e443a0) that fixed #1133/#783/#754 likely cover this case too. Single-environment verification — revisit if re-reported with specific file variants (non-ASCII paths, Zstd-compressed, unfinalized summary).
#903	Floating point truncated exception on INT64 fields	Fixed upstream on 2024-01-29 by 81e0e920 (fix issue #929 : numerical truncation) + dcbc8c6b (bypass truncation check). Actual owner is ParserROS, not DataLoadMCAP — root throw at plotjuggler_plugins/ParserROS/rosx_introspection/include/rosx_introspection/details/conversion_impl.hpp:209,314,334; runtime opt-out wired in ros_parser.cpp:148-161. User-facing workaround: App → Preferences → Behavior → Parsing → "Strict Truncation check (uncheck at your own risk)". Error message (ros_parser.cpp:169-173) guides users to this toggle. GitHub issue is still open only because the fix commit cited #929 instead of #903 — administrative close recommended (e.g. as duplicate of #929).

Plugin bugs — Lua / Scripting (4 / 4 — category complete)

#	Title	Evidence
#1328	Fails to build with Lua 5.5	5d3e232b — hand-port of sol2 PR #1723 into the bundled single-header 3rdparty/sol2-3.5.0/include/sol/sol.hpp (widens unsupported-version guard to accept 5.5, extends LUA_ERRGCMM shim to Lua ≥ 5.4, wraps lua_newstate for the 3-arg 5.5 signature). Regression-checked against Lua 5.1 (minimal sol2 program compiles and runs); real 5.5 build not verified locally (Lua 5.5 not installed on this machine — worth a real build once available).
#1312	New Lua version breaks quat_to_yaw / quat_to_roll	ae78a30b — default.snippets.xml: PR #1164 had swapped math.atan(y, x) for math.atan2(y, x) to support Lua 5.1/5.2, but math.atan2 was removed in Lua 5.4, so the snippets broke on the default bundled 5.4.7. Fix adds local atan2 = math.atan2 or math.atan at the top of both snippets so the correct name is picked at load time across Lua 5.1–5.5. quat_to_pitch uses math.asin and was unaffected.
#968	Misalignment in time series due to differing sizes with reactive scripts	8d9c2afb — reactive_function.{h,cpp}: TimeseriesRef::atTime(t) silently returns the clamped endpoint value when t falls outside the series' recorded range (because getIndexFromX in timeseries.h:152-175 clamps to size()-1/0). Feb 2026 storage refactor (48497feb, 9493434e) did not change that semantics. Fix exposes getIndexAtTime(t) on TimeseriesRef so reactive-script Lua (the ToolboxLuaEditor / ReactiveLuaFunction context) can compare the retrieved sample's x against the query time and detect out-of-range clamping. Scope note: applies only to reactive scripts, not to Lua custom functions (LuaCustomFunction has its own _lua_engine with no TimeseriesRef binding; its scripts receive pre-computed positional args from the C++ host and share the same clamping bug at python_custom_function.cpp:483,490 and the equivalent in lua_custom_function.cpp). Fixing that second surface requires a host-level range check — deliberately out of scope here since no user has reported it. Purely additive; matches firemark's suggestion on the closed PR #969. Fork-only for now (not sent upstream).
#800	Crash caused by reactive script errors	bdf39ec9 — reactive_function.{h,cpp}: ReactiveLuaFunction::calculate() showed a modal QMessageBox::warning for every failure (line present and untouched since 2022-07-25, af4f9d115); under streaming sources (UDP 20-100 Hz) a script that errors every tick piled dialogs faster than the user could dismiss them, exhausting Qt USER32 handles on Windows and crashing PJ. Fix adds a _disabled_after_error flag set before the modal QMessageBox::warning is shown (so re-entrant streaming timer firings during the nested modal event loop see the flag and short-circuit, instead of queueing more dialogs). Script is auto re-enabled on save via the existing editor lifecycle (lua_editor.cpp:271 constructs a fresh ReactiveLuaFunction on every save — new instance → default flag), so no explicit resume API was needed. Scope: reactive scripts only; CustomFunction streaming-path errors go to stderr via qWarning, no dialog, no parallel spam pathology.

Plugin bugs — ZMQ (2 / 3)

#	Title	Evidence
#705	ZeroMQ Subscriber filter	COMPLETED — fixed upstream by PR #730 (db90f70a ZMQ: Add topics filtering, merged 2022-12-18). Already on plugin_issues (and main). Adds semicolon/comma/whitespace-separated topic-filter parsing in datastream_zmq.cpp:355-372 (parseTopicFilters), plus a test publisher at plotjuggler_plugins/DataStreamZMQ/utilities/start_test_publisher.py. Safe administrative close — Davide merged this himself.
#941	Extraneous suffixes added to ZMQ IPC endpoints	COMPLETED — 752e1773 "datastream_zmq: drop bogus :port suffix on ipc:// endpoints". The socket-address assembly at datastream_zmq.cpp:153 always concatenated <protocol><address>:<port>, which is correct for tcp:// but corrupted ipc:// endpoints (IPC uses filesystem paths, no port). Fix guards the :port suffix on protocol and disables the port field in the dialog when ipc:// is selected so the user can't enter a value that would be silently ignored. Manually verified on 2026-04-21 with a pyzmq PUB on both tcp://*:9872 and ipc:///tmp/pj_zmq_test against PlotJuggler's ZMQ Subscriber — both paths plot as expected; IPC case no longer appends :9872. Two sibling bugs are intentionally left in-tree (separate tickets if re-reported): radioBind toggle forces "0.0.0.0" (TCP-specific), and the default stored address is "localhost" (also TCP-centric). Fork-only for now (not yet sent upstream).

Plugin bugs — UDP (1 / 1 — administrative close, not worth fixing)

#	Title	Evidence
#839	UDP JSON stream timestamp setting not saved with layout	Administrative close — no code. No sense fixing it in isolation. The reported symptom is one face of a design shared by all five "simple" streamers (DataStreamUDP, DataStreamMQTT, DataStreamZMQ, DataStreamWebsocket, DataStreamSerialPort): none override xmlSaveState / xmlLoadState, all rely on QSettings for cross-run memory. The complex bridges (DataStreamFoxgloveBridge, DataStreamPlotJugglerBridge, PluginsZcm) and every DataLoader (Parquet, CSV, ULog, MCAP, Zcm) do persist state to the layout XML. Fixing UDP alone would desync one plugin from its four siblings without solving the reporter's underlying workflow; broadening to all five is refactor scope out-of-bounds for a single ticket. Reopen if the reporter describes a concrete use-case that QSettings defaults cannot satisfy, or if layout-persistence is rebuilt as a base-class feature on PJ::DataStreamer.

Plugin bugs — Parquet (2 / 2 — category complete)

#	Title	Evidence
#863	Parquet file with BYTE_ARRAY / TIMESTAMP_MILLIS columns fails to load	4080e0eb — dataload_parquet.cpp: split the old numeric-only whitelist (lines 253-259) into is_numeric + is_string, and routed arrow::Type::STRING / LARGE_STRING / BINARY columns to PlotDataMapRef::getOrCreateStringSeries() — mirroring the CSV pattern from commit 44118575 "Feature: StringsSeries allowed". ColumnInfo now carries both PlotData* and StringSeries* (exactly one set per column); the per-row batch loop dispatches on whichever is populated. arrow::StringArray / LargeStringArray / BinaryArray all share GetView(int64_t) → std::string_view, which constructs PJ::StringRef directly; StringSeries::pushBack (stringseries.h:56-65) does the interning. String columns are deliberately not offered as timestamp-axis candidates. The TIMESTAMP_MILLIS half of the ticket was already fixed by 900557e0 (Feb 2026, "parquet_loading: fix timestamp handling for arrow::TIMESTAMP"). Manually verified on 2026-04-21 with a pyarrow fixture containing timestamp (ms) + temperature (double) + label (string) + status (binary) columns: all four appeared in the Timeseries list, the 5-second X-axis matched 50 × 100 ms, and a Python Custom Function reading value == "ok" produced the expected square wave — proving strings actually land in _plot_data->strings and are consumable by the scripting layer. Scope caveats: FIXED_SIZE_BINARY not added (less common, would need a separate FixedSizeBinaryArray branch). Not tested on very large files or on chunked-string arrays.
#862	Building DataLoadParquet on Windows	Administrative close — no code change in this repo. Commit 00b94253 "add Arrow to conan and fix Windows CI (#1109)" already added Conan to the Parquet plugin build, and plotjuggler_plugins/DataLoadParquet/cmake/FindArrow.cmake:35-62 covers MSVC toolchain detection, _static suffix handling, .dll lookup, and import-lib (.lib) resolution. Not independently verified on Windows (no MSVC environment on Álvaro's machine). Revisit if re-reported with specific MSVC / vcpkg / mingw / Qt version repro steps — the current close rests on evidence of prior work, not on a live Windows build.

Feature requests — addressed indirectly (2 / 51)

#	Title	Evidence
#1323	Python scripting + StringSeries support + UI improvements to Custom Functions	c04e376e, ee91bbc1, ea9635e6, 29f8b6cb
#1220	Option for dates & times stored in separate columns	8d1346c3 (upstream PR #1259)

---

⏳ Waiting

Plugin bugs — ROS / ROS2 (18)

#	Title	Status
#1334	UI freezes when starting ROS2 Topic Subscriber in massive topic environment	⏳
#1315	Segfault subscribing to AMCL Particle Markers	⏳
#1283	Snap version lacks ROS-O toolkit	⏳
#1262	ROS2 std_msgs/Header treated differently in custom msg vs. built-in	⏳
#1204	Problem with ROS2 Header in some topics	⏳
#1060	Plugin exception "package not found" on ROS action feedback topics	⏳
#991	DataLoad ROS bags exception: map::at	⏳
#881	ROS2 message with arrays and bounded types: not enough memory in buffer	⏳
#878	No parser for encoding [cdr] after MCAP conversion from ros2 db3	⏳
#860	ROS2 topic Re-Publisher not finding topic list while topics visible	⏳
#841	Cannot find ros2 topic subscriber when compiling from source	⏳
#759	Crash when selecting topics to republish in ROS2	⏳
#743	After plotting IMU data, can't see orientation/angular velocity	⏳
#729	ParserROS compile issues on Windows	⏳
#622	Extra metadata header appearing in topic tree	⏳
#537	Rosout dark-theme: debug messages invisible	⏳
#477	Segfault in ROS2 topic Re-Publisher	⏳
#458	ROS topic cannot be loaded	⏳

Plugin bugs — MCAP (0 waiting of 5) ✅

All five MCAP-category issues are marked fixed.

Plugin bugs — Lua / Scripting (0 waiting of 4) ✅

All four Lua / Scripting issues are marked fixed.

Plugin bugs — MQTT (3)

#	Title	Status
#1148	Unable to connect via MQTT	⏳
#876	MQTT connection creating malformed packets	⏳
#747	MQTT streamer only subscribes to single topic	⏳

Plugin bugs — ZMQ (1 waiting of 3)

#	Title	Status
#1126	ZMQ Subscriber lacks option to specify Protobuf Paths	⏳

Plugin bugs — Parquet (0 waiting of 2) ✅

All two Parquet-category issues are marked fixed.

Plugin bugs — ULog (2)

#	Title	Status
#796	Multiple improvements to PX4 plugin needed	⏳
#672	Cannot load large ULog file (2.5 GB)	⏳

Plugin bugs — single-issue plugins (3)

#	Plugin	Title	Status
#1144	ParserProtobuf	Parse multiple protobuf message types simultaneously	⏳
#741	ToolboxFFT	FFT Y-scaling not working	⏳
#1243	ToolboxQuaternion	Add Quaternion-to-RPY autofill for PX4	⏳

Core app (20)

#	Title	Status
#1331	Complete Timeseries list synchronization and custom series separation	⏳
#1276	Compiling error: no declaration matches	⏳
#1275	Hang on exit	⏳
#1253	UI inconsistencies on HiDPI displays	⏳
#1197	Displayed value is nearest datapoint, not current/previous	⏳
#1114	Scroll step size bug	⏳
#1111 / #1104	Show only current time range in X/Y plots	⏳
#1092	Qt6: initial port	⏳
#1084	Change the time source for ROS2 bags	⏳
#911	Crash when dragging data into plot area on macOS	⏳
#886	Request for flatpak distribution support	⏳
#883	Can't find one or more curves	⏳
#816	Errors when launching AppImage	⏳
#791	Broken scaling with multiple displays	⏳
#756	Layout doesn't restore XY plots when multiple sources with prefixes are used	⏳
#563	Plot refreshing bug with OpenGL enabled	⏳
#468	Menus overlapping on high DPI device	⏳
#449	New XY plot stuck in loop	⏳
#402	Auto-reload custom timeseries when changing new file	⏳
#213	Keyed topics not supported	⏳

Feature requests (49)

Waiting (in issue-number order): #1333, #1332, #1327, #1322, #1321, #1319, #1311, #1288, #1286, #1274, #1267, #1266, #1264, #1251, #1245, #1213, #1044, #957, #945, #847, #823, #809, #788, #787, #785, #782, #775, #755, #750, #748, #732, #708, #695, #691, #635, #616, #606, #572, #570, #511, #471, #444, #442, #414, #367, #365, #286, #201, #110

Obsolete / Questions (20) — administrative close, no code

Obsolete (close as wontfix/duplicate): #1335, #1013, #1261, #943, #851, #913, #1270, #980, #875, #799

Questions (close as question/support): #1336, #1017, #1167, #1116, #1170, #894, #766, #644, #427, #744

---

Next up — Parquet plugin complete

Plugin-category progress: #863 by the new BYTE_ARRAY → StringSeries path in dataload_parquet.cpp (plus the earlier 900557e0 TIMESTAMP fix), #862 by administrative close citing commit 00b94253's Conan/Arrow Windows CI work, #839 (UDP layout-persistence) closed administratively as systemic-not-UDP-specific, #705 (ZMQ topic filter) closed administratively via upstream PR #730 / db90f70a, and #941 (ZMQ IPC :port suffix) fixed by 752e1773. Four plugin categories fully closed — CSV (8/8), MCAP (5/5), Lua (4/4), Parquet (2/2) — plus UDP (1/1 administrative) and ZMQ (2/3 — only #1126 Protobuf-paths left). 22 plugin issues closed of 49 total.

Remaining plugin categories (27 waiting): ROS/ROS2 (18), MQTT (3), ULog (2), ZMQ (1), Protobuf / FFT / Quaternion (1 each).

Suggested next target

• #537 (ROS/ROS2, rosout dark-theme invisible debug messages): tiniest scope, pure stylesheet. Good warm-up.
• #741 (ToolboxFFT Y-scaling): bounded inside one toolbox.
• #991 (DataLoad ROS bags map::at): probably a missing .count() guard.
• ROS/ROS2 clustering pass: #477 / #759 / #860 all in the ROS2 Re-Publisher, #1204 / #1262 about ROS2 Header quirks — duplicates inside those clusters may drop the effective count to ~14.
• ULog (2): #672 (2.5 GB file) + #796 (PX4 improvements meta-ticket). Upstream has an experimental rewrite on test/ulog-parser-unit-tests (bde0ede5, DataLoadULog2 using ulog_cpp) that would likely close both — but it's not merged. Watch-and-wait.

---

Caveats

• "Fixed" here means a commit in history claims to fix the issue, or code review shows the capability is present. Not independently verified against the live GitHub tracker.
• #1290 / #1137 (CSV export of quaternion/RPY) should get a manual smoke test before being closed on GitHub: load MCAP with IMU → ToolboxQuaternion → export RPY → inspect the resulting CSV.
• #662 (multi-file prefix on individually-added files) is marked fixed because --auto-prefix covers the CLI workflow; the underlying GUI flow where files are added one-by-one via + is unchanged. Revisit if the issue is re-reported with GUI-specific repro steps.
• #1337 is a PR number (not in the 156-issue list); it closes issue #1018.
• Several merged upstream commits fix PRs/issues not in the 156-issue snapshot and are therefore not counted: #1015, #1065, #1254, #1277, #1278, #1279, #1281, #1282, #1285, #1287, #1289, #1295, #1296, #1297, #1298, #1300, #1301, #1306, #1310, #1314, #1324.