fix(dax): quote pipe source/sink properties so node descriptions survive pipewire-pulse by Ozy311 · Pull Request #3438 · aethersdr/AetherSDR

Ozy311 · 2026-06-06T19:37:05Z

fix(dax): quote pipe source/sink properties so node descriptions survive pipewire-pulse

Summary

On builds without libpipewire-0.3 (native PipeWire off), DAX audio nodes are created
through the legacy pactl module-pipe-source / module-pipe-sink path. That path passes the
per-channel name as source_properties=device.description="AetherSDR DAX N", but
pipewire-pulse's pulse-module argument parser splits the property value on unescaped
whitespace and does not honor the embedded quotes for grouping. The value is taken only up
to the first space, so device.description collapses to "AetherSDR" for every DAX RX node
and the TX node. node.description is derived from device.description, so qpwgraph, JACK
patchbays, WSJT-X and FLDIGI all show an indistinguishable "AetherSDR" and operators must
hand-route by guesswork.

The same truncation also orphans the trailing node.latency=256/48000 quantum hint (#1008),
which never reaches the node.

Fix

Wrap the property value in single quotes so the spaced device.description (and the
node.latency hint) survive the split. node.description is then derived correctly. 2 lines:

- QStringLiteral("source_properties=device.description=\"%1\" node.latency=256/48000").arg(sourceDesc),
+ QStringLiteral("source_properties='device.description=\"%1\" node.latency=256/48000'").arg(sourceDesc),
- QStringLiteral("sink_properties=device.description=\"%1\"").arg(sinkDesc),
+ QStringLiteral("sink_properties='device.description=\"%1\"'").arg(sinkDesc),

The native pw_stream path (PipeWireNativeRxSource) already sets PW_KEY_NODE_DESCRIPTION
via the API and is unaffected — this aligns the legacy fallback (and TX, which always uses the
legacy sink) with it.

Why (quoting A/B)

Exact argv the code emits, read back with pw-cli on pipewire-pulse:

`source_properties` value emitted	`node.description`	`node.latency`
`device.description="AetherSDR DAX 1" node.latency=256/48000` (before)	`AetherSDR`	dropped
`device.description=AetherSDR\ DAX\ 1` (backslash)	`AetherSDR`	n/a
`'device.description="AetherSDR DAX 1" node.latency=256/48000'` (fix)	`AetherSDR DAX 1`	`256/48000`

Test plan

Incremental ninja -C build clean (Nobara/FC43, Qt 6.10.3, gcc 15.2.1; native PipeWire off)
Live against a FLEX-6700, DAX enabled — pw-cli ls Node:
aethersdr-dax-1..4 → AetherSDR DAX 1..4, aethersdr-tx → AetherSDR TX
qpwgraph shows the named pipe nodes (screenshot)
FLDIGI Soundcard/Devices (PortAudio): Capture lists AetherSDR DAX 1..4, Playback AetherSDR TX (screenshots)
WSJT-X device list (maintainer/secondary — PulseAudio names follow the same device.description)

Screenshots (patched build, live on a FLEX-6700, Linux/pipewire-pulse)

FLDIGI → Soundcard/Devices — Capture lists each DAX channel by name:

FLDIGI Playback — AetherSDR TX selectable by name:

qpwgraph — named pipe nodes (AetherSDR DAX 1/2/4, AetherSDR TX; pre-fix all read AetherSDR):

Current DAX Implementation higher latency than SSDR DAX Latency #1008 — DAX RX latency; the node.latency=256/48000 quantum hint this fix un-orphans.
DAX RX: native pw_stream source on Linux — 400ms → 200ms (#1008) #2312 — added the node.latency=256/48000 token to this exact RX line (latency work) but left
the value unquoted; this fix makes both that hint and the spaced device.description survive.
It deliberately kept the legacy module-pipe-source/sink fallback that this patch repairs.

Notes / out of scope

node.nick is intentionally not set (the native path doesn't set it either; node.description
is what qpwgraph/JACK display).
Known related (separate follow-up): the app's own RX/TX audio streams (media.class
Stream/Output/Audio / Stream/Input/Audio) still carry the generic node.name="AetherSDR"
with no node.description, so they appear as AetherSDR / AetherSDR-<id> in device lists.
Different code path (main audio engine, not the DAX bridge) — tracked in App's own RX/TX audio streams show as generic "AetherSDR" (no node.description) in device lists #3440.
DaxIqModel.cpp builds its pactl command with the same backslash-escaped-space form (via
bash -c), so DAX-IQ descriptions are likely truncated the same way — separate follow-up.

Closes #3437

73, Ozy K6OZY
AI compute partnership: cloaked.agency — drafted with Don, our VP of Engineering agent, on the Linux dev stack (nobara-dell: Nobara/FC43, KDE Plasma 6.6.4, Qt 6.10.3, kernel 7.0.9, Wayland; Dell Precision 5570).

…ive pipewire-pulse The legacy module-pipe-source/sink path passes the channel name via source_properties=device.description="AetherSDR DAX N", but pipewire-pulse's pulse-module argument parser splits the property value on unescaped whitespace and does not honor the embedded quotes for grouping. The value is taken only up to the first space, so device.description collapses to "AetherSDR" for every DAX RX node and the TX node, and the trailing node.latency=256/48000 hint (aethersdr#1008) is orphaned. node.description is derived from device.description, so qpwgraph, JACK patchbays, WSJT-X and FLDIGI all show an indistinguishable "AetherSDR". Wrap the property value in single quotes so the spaced device.description (and node.latency) survive the split. node.description is then derived correctly: DAX RX nodes show "AetherSDR DAX 1".."DAX 4" and TX shows "AetherSDR TX". Verified live against a FLEX-6700 (pw-cli + qpwgraph + FLDIGI PortAudio device list). The native pw_stream path (PipeWireNativeRxSource) already sets node.description via the API and is unaffected; this fixes the legacy fallback (and TX, which always uses the legacy sink). Co-authored-by: Don @ cloaked.agency <don@cloaked.agency> & K6OZY

aethersdr-agent

Looks good — clean, minimal fix with thorough evidence.

A few things I verified:

runPactl calls QProcess::start("pactl", args), so each QStringList entry is its own argv element with no shell expansion. The single quotes therefore reach pipewire-pulse's module-arg parser as literal characters, where they group the spaced value — exactly matching your A/B table. ✅
sourceDesc / sinkDesc are built from a fixed literal plus an integer index, so there's no quote-injection surface to worry about from the description string itself.
The native pw_stream path (PipeWireNativeRxSource) uses PW_KEY_NODE_DESCRIPTION directly and is unaffected, as noted — this only repairs the legacy module-pipe-source/sink fallback, which is the path TX always takes and RX takes when libpipewire-0.3 is off.
The inline comments are the right amount of context — future readers will know why those single quotes are load-bearing instead of "stylistic."

The node.latency=256/48000 un-orphaning is a nice incidental win on top of the description fix (#1008).

The DaxIqModel.cpp follow-up note and the separate node.name=AetherSDR issue (#3440) are correctly scoped out of this PR.

Thanks for the careful write-up, the pw-cli readback table, and the screenshots — this was easy to review. 73!

— AetherClaude

_{🤖 aethersdr-agent · cost: $6.7861 · model: claude-opus-4-7}

aethersdr-agent

Thanks for the careful writeup — the empirical A/B table with pw-cli readback makes this very easy to verify. LGTM.

Confirmed locally:

runPactl() invokes pactl directly via QProcess::start("pactl", args) (no shell), so the single quotes are passed literally to pactl/pulse-module parser — which is the intended behavior here. PulseAudio's pa_modargs parser is the layer that respects the outer '…' for grouping, exactly as the patch description claims.
sourceDesc / sinkDesc are constructed from a hardcoded prefix plus an integer ("AetherSDR DAX %1" / "AetherSDR TX"), so there's no injection / quote-escape concern in the formatted value.
Change is scoped strictly to the two pactl argv lines in PipeWireAudioBridge.cpp; no behavior change for the native pw_stream path or for build configurations with libpipewire-0.3.
Comments anchor the why (parser splits on whitespace) at both call sites, which is exactly where a future reader would otherwise re-introduce the bug.

Nice catch on the orphaned node.latency=256/48000 token too — that's a quiet regression from #2312 that this same edit un-breaks.

73 and thanks for the contribution!

_{🤖 aethersdr-agent · cost: $3.2189 · model: claude-opus-4-7}

## Summary The CW sidetone uses the PortAudio backend, which on Linux runs over PipeWire's JACK API and registers a JACK client under PortAudio's hardcoded default name **`PortAudio`**. The sidetone sink opens alongside RX and stays open for the whole RX session, so qpwgraph / JACK patchbays / WSJT-X / FLDIGI show a persistent node literally named `PortAudio` with no indication it belongs to AetherSDR — the most opaque label of any AetherSDR audio node. ## Fix Call `PaJack_SetClientName("AetherSDR CW Sidetone")` immediately before `Pa_Initialize()`. The `pa_jack` contract requires the call precede `Pa_Initialize` and references (does not copy) the string, so it uses a static-lifetime `const char[]`. Guarded with `#if defined(Q_OS_LINUX) && __has_include(<pa_jack.h>)` so macOS CoreAudio / Windows WASAPI builds (no `pa_jack.h`, no JACK node) compile unaffected. CW sidetone is AetherSDR's only PortAudio user, so naming the process-global JACK client is unambiguous. ## Why - Completes the "every AetherSDR audio node is identifiable in patchbays" line of work alongside the merged DAX/TX pipe-naming fix (#3438) and the still-open main RX/TX stream naming (#3440). - A node named `PortAudio` is indistinguishable from any other PortAudio app and confuses routing in WSJT-X / FLDIGI device pickers. ## Scope - 13 lines added in one file (`src/core/CwSidetonePortAudioSink.cpp`). - Linux/PipeWire-only effect (guarded); no protocol, persistence, or UX change. ## Test plan - [x] Local incremental `ninja -C build` clean on Linux (Nobara 43, Qt 6.10.3, gcc 15.2.1). - [x] **Live A/B on a FLEX-6700 (RX-only), `pw-dump` ground truth:** - Baseline (upstream/main `753eb689`): sidetone node `client.name` / `node.name` / `node.description` all read **`PortAudio`**. - This branch: all three read **`AetherSDR CW Sidetone`**. - [x] macOS CoreAudio / Windows WASAPI: compiles unaffected (guard excludes the call; no JACK node on those backends) — maintainer/secondary. Related: #3438, #3440 --- 73, Ozy **K6OZY** AI compute partnership: [cloaked.agency](https://cloaked.agency) — drafted with Don, our VP of Engineering agent, on Linux dev stack (`nobara-dell`).

…ive pipewire-pulse (aethersdr#3438) fix(dax): quote pipe source/sink properties so node descriptions survive pipewire-pulse ## Summary On builds without `libpipewire-0.3` (native PipeWire off), DAX audio nodes are created through the legacy `pactl module-pipe-source` / `module-pipe-sink` path. That path passes the per-channel name as `source_properties=device.description="AetherSDR DAX N"`, but **pipewire-pulse's pulse-module argument parser splits the property value on unescaped whitespace and does not honor the embedded quotes for grouping.** The value is taken only up to the first space, so `device.description` collapses to `"AetherSDR"` for every DAX RX node and the TX node. `node.description` is derived from `device.description`, so qpwgraph, JACK patchbays, WSJT-X and FLDIGI all show an indistinguishable `"AetherSDR"` and operators must hand-route by guesswork. The same truncation also orphans the trailing `node.latency=256/48000` quantum hint (aethersdr#1008), which never reaches the node. ## Fix Wrap the property value in single quotes so the spaced `device.description` (and the `node.latency` hint) survive the split. `node.description` is then derived correctly. 2 lines: ```diff - QStringLiteral("source_properties=device.description=\"%1\" node.latency=256/48000").arg(sourceDesc), + QStringLiteral("source_properties='device.description=\"%1\" node.latency=256/48000'").arg(sourceDesc), - QStringLiteral("sink_properties=device.description=\"%1\"").arg(sinkDesc), + QStringLiteral("sink_properties='device.description=\"%1\"'").arg(sinkDesc), ``` The native `pw_stream` path (`PipeWireNativeRxSource`) already sets `PW_KEY_NODE_DESCRIPTION` via the API and is unaffected — this aligns the legacy fallback (and TX, which always uses the legacy sink) with it. ## Why (quoting A/B) Exact argv the code emits, read back with `pw-cli` on pipewire-pulse: | `source_properties` value emitted | `node.description` | `node.latency` | |-------------------------------------------------------------------------|--------------------|----------------| | `device.description="AetherSDR DAX 1" node.latency=256/48000` (before) | `AetherSDR` | dropped | | `device.description=AetherSDR\ DAX\ 1` (backslash) | `AetherSDR` | n/a | | `'device.description="AetherSDR DAX 1" node.latency=256/48000'` (fix) | `AetherSDR DAX 1` | `256/48000` | ## Test plan - [x] Incremental `ninja -C build` clean (Nobara/FC43, Qt 6.10.3, gcc 15.2.1; native PipeWire off) - [x] Live against a FLEX-6700, DAX enabled — `pw-cli ls Node`: `aethersdr-dax-1..4` → `AetherSDR DAX 1..4`, `aethersdr-tx` → `AetherSDR TX` - [x] qpwgraph shows the named pipe nodes (screenshot) - [x] FLDIGI Soundcard/Devices (PortAudio): Capture lists `AetherSDR DAX 1..4`, Playback `AetherSDR TX` (screenshots) - [x] WSJT-X device list (maintainer/secondary — PulseAudio names follow the same `device.description`) ## Screenshots (patched build, live on a FLEX-6700, Linux/pipewire-pulse) FLDIGI → Soundcard/Devices — **Capture** lists each DAX channel by name: ![FLDIGI Capture list](https://raw.githubusercontent.com/Ozy311/AetherSDR/pr3438-assets/pr-capture-list.png) FLDIGI **Playback** — `AetherSDR TX` selectable by name: ![FLDIGI Playback list](https://raw.githubusercontent.com/Ozy311/AetherSDR/pr3438-assets/pr-playback-list.png) qpwgraph — named pipe nodes (`AetherSDR DAX 1/2/4`, `AetherSDR TX`; pre-fix all read `AetherSDR`): ![qpwgraph](https://raw.githubusercontent.com/Ozy311/AetherSDR/pr3438-assets/qpwgraph-crop.png) ## Related - aethersdr#1008 — DAX RX latency; the `node.latency=256/48000` quantum hint this fix un-orphans. - aethersdr#2312 — added the `node.latency=256/48000` token to this exact RX line (latency work) but left the value unquoted; this fix makes both that hint and the spaced `device.description` survive. It deliberately kept the legacy `module-pipe-source/sink` fallback that this patch repairs. ## Notes / out of scope - `node.nick` is intentionally not set (the native path doesn't set it either; `node.description` is what qpwgraph/JACK display). - Known related (separate follow-up): the app's own RX/TX audio streams (`media.class` `Stream/Output/Audio` / `Stream/Input/Audio`) still carry the generic `node.name="AetherSDR"` with no `node.description`, so they appear as `AetherSDR` / `AetherSDR-<id>` in device lists. Different code path (main audio engine, not the DAX bridge) — tracked in aethersdr#3440. - `DaxIqModel.cpp` builds its `pactl` command with the same backslash-escaped-space form (via `bash -c`), so DAX-IQ descriptions are likely truncated the same way — separate follow-up. Closes aethersdr#3437 --- 73, Ozy **K6OZY** AI compute partnership: [cloaked.agency](https://cloaked.agency) — drafted with Don, our VP of Engineering agent, on the Linux dev stack (`nobara-dell`: Nobara/FC43, KDE Plasma 6.6.4, Qt 6.10.3, kernel 7.0.9, Wayland; Dell Precision 5570).

…hersdr#3517) ## Summary The CW sidetone uses the PortAudio backend, which on Linux runs over PipeWire's JACK API and registers a JACK client under PortAudio's hardcoded default name **`PortAudio`**. The sidetone sink opens alongside RX and stays open for the whole RX session, so qpwgraph / JACK patchbays / WSJT-X / FLDIGI show a persistent node literally named `PortAudio` with no indication it belongs to AetherSDR — the most opaque label of any AetherSDR audio node. ## Fix Call `PaJack_SetClientName("AetherSDR CW Sidetone")` immediately before `Pa_Initialize()`. The `pa_jack` contract requires the call precede `Pa_Initialize` and references (does not copy) the string, so it uses a static-lifetime `const char[]`. Guarded with `#if defined(Q_OS_LINUX) && __has_include(<pa_jack.h>)` so macOS CoreAudio / Windows WASAPI builds (no `pa_jack.h`, no JACK node) compile unaffected. CW sidetone is AetherSDR's only PortAudio user, so naming the process-global JACK client is unambiguous. ## Why - Completes the "every AetherSDR audio node is identifiable in patchbays" line of work alongside the merged DAX/TX pipe-naming fix (aethersdr#3438) and the still-open main RX/TX stream naming (aethersdr#3440). - A node named `PortAudio` is indistinguishable from any other PortAudio app and confuses routing in WSJT-X / FLDIGI device pickers. ## Scope - 13 lines added in one file (`src/core/CwSidetonePortAudioSink.cpp`). - Linux/PipeWire-only effect (guarded); no protocol, persistence, or UX change. ## Test plan - [x] Local incremental `ninja -C build` clean on Linux (Nobara 43, Qt 6.10.3, gcc 15.2.1). - [x] **Live A/B on a FLEX-6700 (RX-only), `pw-dump` ground truth:** - Baseline (upstream/main `753eb689`): sidetone node `client.name` / `node.name` / `node.description` all read **`PortAudio`**. - This branch: all three read **`AetherSDR CW Sidetone`**. - [x] macOS CoreAudio / Windows WASAPI: compiles unaffected (guard excludes the call; no JACK node on those backends) — maintainer/secondary. Related: aethersdr#3438, aethersdr#3440 --- 73, Ozy **K6OZY** AI compute partnership: [cloaked.agency](https://cloaked.agency) — drafted with Don, our VP of Engineering agent, on Linux dev stack (`nobara-dell`).

Ozy311 marked this pull request as ready for review June 6, 2026 19:44

Ozy311 requested a review from a team as a code owner June 6, 2026 19:44

aethersdr-agent Bot reviewed Jun 6, 2026

View reviewed changes

ten9876 self-assigned this Jun 7, 2026

ten9876 merged commit b245b1d into aethersdr:main Jun 7, 2026
6 checks passed

Ozy311 mentioned this pull request Jun 10, 2026

fix(audio): name CW sidetone JACK client "AetherSDR CW Sidetone" #3517

Merged

3 tasks

Ozy311 mentioned this pull request Jun 13, 2026

App's own RX/TX audio streams show as generic "AetherSDR" (no node.description) in device lists #3440

Open

Ozy311 deleted the fix/dax-pipewire-node-names branch June 13, 2026 23:27

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(dax): quote pipe source/sink properties so node descriptions survive pipewire-pulse#3438

fix(dax): quote pipe source/sink properties so node descriptions survive pipewire-pulse#3438
ten9876 merged 1 commit into
aethersdr:mainfrom
Ozy311:fix/dax-pipewire-node-names

Ozy311 commented Jun 6, 2026 •

edited

Loading

Uh oh!

aethersdr-agent Bot left a comment

Uh oh!

aethersdr-agent Bot left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

Ozy311 commented Jun 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Fix

Why (quoting A/B)

Test plan

Screenshots (patched build, live on a FLEX-6700, Linux/pipewire-pulse)

Related

Notes / out of scope

Uh oh!

aethersdr-agent Bot left a comment

Choose a reason for hiding this comment

Uh oh!

aethersdr-agent Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Ozy311 commented Jun 6, 2026 •

edited

Loading