[Emc-commit] [LinuxCNC/linuxcnc] 54f3fc: test: extend ui-smoke with g-code execution and en...

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

  Branch: refs/heads/master
  Home:   https://github.com/LinuxCNC/linuxcnc
  Commit: 54f3fc2ca3819e5f0ab094998da003612c561a9c
      https://github.com/LinuxCNC/linuxcnc/commit/54f3fc2ca3819e5f0ab094998da003612c561a9c
  Author: Luca Toniolo <107...@us...>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    M debian/control.top.in
    M tests/ui-smoke/_lib/drive.py
    M tests/ui-smoke/_lib/launch.sh
    M tests/ui-smoke/_lib/run-gui.sh
    A tests/ui-smoke/_lib/smoke.ngc
    M tests/ui-smoke/axis/test.sh
    M tests/ui-smoke/gmoccapy/test.sh
    M tests/ui-smoke/qtdragon/test.sh
    M tests/ui-smoke/touchy/test.sh

  Log Message:
  -----------
  test: extend ui-smoke with g-code execution and endpoint check

Each per-GUI test now also drives estop reset, machine on, home all,
mode auto, program_open + auto(RUN) on a tiny shared smoke.ngc, waits
for sustained INTERP_IDLE, and asserts stat.position delta against
--expect-delta-mm 1,1,0 converted via stat.linear_units so the same
arg works on inch (axis, touchy) and mm (gmoccapy, qtdragon) sims.

State/mode commands use ensure_state/ensure_mode helpers with a
retry-and-stability pattern: gmoccapy and qtdragon re-issue their own
mode commands during startup and can revert task_mode AUTO -> MANUAL
right after we set it. The helpers wait for the desired state, then
re-check after STATE_STABILITY_S; on revert they retry up to
STATE_RETRY_BUDGET times. Intermediate timeouts use a quiet variant
so spurious UI_SMOKE_FAIL lines do not pollute the log during retries
(checkresult.sh greps for ^UI_SMOKE_FAIL on any line).

smoke.ngc is G21 G91 G0 X1 Y1 G90 M2 - relative move in mm, sim-
agnostic. The driver snapshots stat.position[:3] after homing and
checks (final - start) against the converted delta, sidestepping each
sim's HOME offset.

Adds python3-zmq and python3-opencv to debian/control.top.in under
!nocheck: qtdragon's hal_bridge and the camview widget segfault on
startup without them, which is invisible to the connect-only Phase 1
smoke but breaks the run-program path before the program can start.

5 consecutive local runs all green at 2m43s wall each.

  Commit: 20fbc52ae38e18a680bf28a64bf75104561ece52
      https://github.com/LinuxCNC/linuxcnc/commit/20fbc52ae38e18a680bf28a64bf75104561ece52
  Author: Luca Toniolo <107...@us...>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    M tests/ui-smoke/_lib/run-gui.sh
    M tests/ui-smoke/qtdragon/test.sh

  Log Message:
  -----------
  test: qtdragon ui-smoke runs against writable config mirror

CI run hit a PermissionError in qtvcp's logger when it tried to open
configs/sim/qtdragon/qtdragon_xyz/qtdragon.log for write: the GitHub
Actions workspace is mounted read-only for the docker build user, and
qtvcp resolves LOG_FILE = qtdragon.log into the config dir. hal_bridge
then exits, linuxcnc tears down, and the driver retries ESTOP_RESET
until the budget is exhausted.

qtdragon test.sh now mirrors the qtdragon_xyz config dir to a mktemp
directory, seds LOG_FILE to ~/qtdragon.log, and passes the absolute
INI path to run-gui.sh. run-gui.sh treats any path starting with /
as absolute; everything else still resolves under configs/sim. Trap
cleans the tmp dir on exit so the working tree stays clean.

Does not touch the shipped qtdragon config to avoid changing default
behaviour for real users. The same fix would work for any other
config that turns out to write into its own dir on CI.

  Commit: cc1476b65f3bf738ea6bff5333e99de68359f30e
      https://github.com/LinuxCNC/linuxcnc/commit/cc1476b65f3bf738ea6bff5333e99de68359f30e
  Author: Luca Toniolo <107...@us...>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    M debian/control.top.in
    M tests/ui-smoke/_lib/launch.sh

  Log Message:
  -----------
  test: drop python3-opencv dep and set QT_XCB_GL_INTEGRATION=xcb_egl

Ubuntu 24.04 rip-and-test runs hit a qtvcp segfault after the log-
permission fix let qtvcp get further than Phase 1 had. Debian
package-arch passes the same code. Two known asymmetries match:

- python3-opencv on Ubuntu pulls Qt5 GUI bits whose cv2/qt/plugins
  directory overrides the system PyQt5 platform plugin path under
  xvfb (opencv-python issue #572, Qt Forum 119109). qtvcp's
  camview_widget tolerates ImportError on cv2 and just logs a
  warning, so dropping the dep restores the harmless fallback path
  Phase 1 was already exercising.
- xcb_glx is the historical fragile integration under xvfb
  (Launchpad #1761708, QTBUG-67537); xcb_egl is what software-GL
  stacks expect anyway. Set as defense in depth.

Local 4/4 still green with both changes.

  Commit: ed2c5effd4c6b41be9ce79d68b937a73bfeb3793
      https://github.com/LinuxCNC/linuxcnc/commit/ed2c5effd4c6b41be9ce79d68b937a73bfeb3793
  Author: Luca Toniolo <107...@us...>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    M tests/ui-smoke/qtdragon/test.sh

  Log Message:
  -----------
  test: force QT_QPA_PLATFORM=offscreen for qtdragon

xvfb + xcb + xcb_egl was not enough for Ubuntu 24.04 rip-and-test:
qtvcp still segfaults during widget construction even with opencv
and qtwebengine paths quiet, and the same code passes on Debian
package-arch. Offscreen renders entirely in memory and exercises a
different Qt plugin entirely, dodging the xcb-stack instability.

scripts/linuxcnc itself forces QT_QPA_PLATFORM=xcb unless
LINUXCNC_OPENGL_PLATFORM is set to a non-glx value, so pin both.
Only qtdragon needs this; axis (Tk), touchy and gmoccapy (GTK) are
unaffected. Trade-off: no Phase 3 screenshot from qtdragon under
this config; Phase 3 would need an opt-out for offscreen tests.

  Commit: 218b74e851ab6cc9f41d11c52663357ced4a0b93
      https://github.com/LinuxCNC/linuxcnc/commit/218b74e851ab6cc9f41d11c52663357ced4a0b93
  Author: Luca Toniolo <107...@us...>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    M tests/ui-smoke/qtdragon/test.sh

  Log Message:
  -----------
  test: disable QtWebEngine sandbox+GPU for qtdragon ui-smoke

qtdragon embeds QWebEngineView. On rip-and-test (gcc) CI it racy-crashed
during Chromium browser-process spawn under offscreen + xvfb, no GPU,
no user namespaces. rip-and-test-clang got past it by luck.

Force --no-sandbox --single-process --no-zygote --disable-gpu so the
renderer runs in-process with software rendering.

  Commit: 98a028804cc90d06064587d8d4cb8822e5c3ee22
      https://github.com/LinuxCNC/linuxcnc/commit/98a028804cc90d06064587d8d4cb8822e5c3ee22
  Author: Luca Toniolo <107...@us...>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    M tests/ui-smoke/qtdragon/test.sh

  Log Message:
  -----------
  test: block QtWebEngine import in qtdragon ui-smoke via meta_path shim

QtWebEngine browser-process init segfaults inside the qtvcp process on
Ubuntu 24.04 CI even with --no-sandbox --single-process --disable-gpu.
The smoke test never touches the WebWidget, so block the
qtpy.QtWebEngineWidgets import via a sitecustomize meta_path finder;
WebWidget already has a fallback that swaps in a plain QWidget when
that import fails. No Chromium spawn, no segfault.

The previous chromium-flags attempt was retracted: 'Sandboxing disabled
by user.' confirmed Chromium got the flags but still crashed during
init, so we are not going to win that race.

  Commit: 939cdd5064d1358249afd0a1d55f36a063c49b51
      https://github.com/LinuxCNC/linuxcnc/commit/939cdd5064d1358249afd0a1d55f36a063c49b51
  Author: Luca Toniolo <107...@us...>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    A tests/ui-smoke/_lib/checkresult-quit.sh
    A tests/ui-smoke/_lib/launch-env.sh
    M tests/ui-smoke/_lib/launch.sh
    A tests/ui-smoke/_lib/qtdragon-prepare.sh
    A tests/ui-smoke/_lib/quit-launch.sh
    A tests/ui-smoke/gmoccapy-quit/checkresult
    A tests/ui-smoke/gmoccapy-quit/skip
    A tests/ui-smoke/gmoccapy-quit/test.sh
    A tests/ui-smoke/qtdragon-quit/checkresult
    A tests/ui-smoke/qtdragon-quit/skip
    A tests/ui-smoke/qtdragon-quit/test.sh
    M tests/ui-smoke/qtdragon/test.sh
    A tests/ui-smoke/touchy-quit/checkresult
    A tests/ui-smoke/touchy-quit/skip
    A tests/ui-smoke/touchy-quit/test.sh

  Log Message:
  -----------
  tests: add ui-smoke quit-path coverage (SIGTERM)

Adds a quit-path smoke test per GUI that boots the GUI, waits for the
NML task to come up, sends SIGTERM to the GUI process alone, and asserts
the GUI exits on its own within a short grace. This guards the
clean-shutdown handlers: a GUI that absorbs SIGTERM and has to be
SIGKILLed fails the test.

The new _lib/quit-launch.sh shares the headless environment (software
GL + audio silencing) with launch.sh by sourcing a new
_lib/launch-env.sh rather than copying it, so the two launchers cannot
drift apart. Results go through _lib/checkresult-quit.sh (pass on
UI_SMOKE_QUIT_OK). The GUI process is identified by matching a python
argv[0], so the linuxcnc launcher and xvfb-run wrappers that also carry
the GUI name on their command line are not mistaken for it. Per-GUI
dirs: touchy-quit, gmoccapy-quit, qtdragon-quit.

The qtdragon quit test needs the same CI workarounds the qtdragon smoke
test already carries (writable config mirror with a patched LOG_FILE,
the offscreen Qt platform, and the QtWebEngine import shim). Those move
out of qtdragon/test.sh into _lib/qtdragon-prepare.sh, sourced by both
qtdragon test.sh files, so the quit test reuses them instead of leaving
qtvcp to crash on startup.

Requires the SIGTERM handlers in #4076 (gmoccapy), #4077 (touchy) and
#4078 (qtvcp); without them the GUIs ignore SIGTERM and these tests
fail by design.

(cherry picked from commit aff5991dc3189b48a747c7c7e8962d4d16c5cbe3)

  Commit: 5098eeb4ffc3be6a83e198b179835ed14aeb19b2
      https://github.com/LinuxCNC/linuxcnc/commit/5098eeb4ffc3be6a83e198b179835ed14aeb19b2
  Author: Luca Toniolo <107...@us...>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    M tests/ui-smoke/_lib/drive.py
    M tests/ui-smoke/_lib/launch-env.sh

  Log Message:
  -----------
  ui-smoke: fail fast and explain when the GUI process dies

The driver polled NML for up to 60s after a GUI crash, then blamed
whatever stage timed out (e.g. homing); a dead task keeps serving its
last stat buffer. Watch the launcher PID and fail in ~1s pointing at
the crash. Enable PYTHONFAULTHANDLER for a Python traceback on fatal
signals.

  Commit: 93ad82040d521e3532b7a8d89e72d04d16faea09
      https://github.com/LinuxCNC/linuxcnc/commit/93ad82040d521e3532b7a8d89e72d04d16faea09
  Author: Luca Toniolo <107...@us...>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    A tests/ui-smoke/_lib/crashdump.sh
    M tests/ui-smoke/_lib/launch.sh
    M tests/ui-smoke/_lib/quit-launch.sh

  Log Message:
  -----------
  ui-smoke: capture a native backtrace on GUI crash

PYTHONFAULTHANDLER stops at the Python frame; a Qt/dbus/GL segfault
needs the C stack. Arm a core dump in the launchers and, if the GUI
leaves a core, gdb its backtrace into the log. Failure-path only, so
green runs pay nothing.

  Commit: a9ff6a4b4895ccd814b01db8e90862fbc33e6cf5
      https://github.com/LinuxCNC/linuxcnc/commit/a9ff6a4b4895ccd814b01db8e90862fbc33e6cf5
  Author: Luca Toniolo <107...@us...>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    M tests/ui-smoke/_lib/quit-launch.sh

  Log Message:
  -----------
  ui-smoke: show the driver's stderr in the quit path

quit-launch.sh catted ui-smoke.out but not ui-smoke.err, hiding the
driver's failure reason (e.g. the GUI-crashed message) on a quit-path
failure. Cat it like launch.sh already does.

  Commit: 81c3d892ba843993e3eae78d6efd54abaece848e
      https://github.com/LinuxCNC/linuxcnc/commit/81c3d892ba843993e3eae78d6efd54abaece848e
  Author: andypugh <an...@bo...>
  Date:   2026-06-05 (Fri, 05 Jun 2026)

  Changed paths:
    M debian/control.top.in
    A tests/ui-smoke/_lib/checkresult-quit.sh
    A tests/ui-smoke/_lib/crashdump.sh
    M tests/ui-smoke/_lib/drive.py
    A tests/ui-smoke/_lib/launch-env.sh
    M tests/ui-smoke/_lib/launch.sh
    A tests/ui-smoke/_lib/qtdragon-prepare.sh
    A tests/ui-smoke/_lib/quit-launch.sh
    M tests/ui-smoke/_lib/run-gui.sh
    A tests/ui-smoke/_lib/smoke.ngc
    M tests/ui-smoke/axis/test.sh
    A tests/ui-smoke/gmoccapy-quit/checkresult
    A tests/ui-smoke/gmoccapy-quit/skip
    A tests/ui-smoke/gmoccapy-quit/test.sh
    M tests/ui-smoke/gmoccapy/test.sh
    A tests/ui-smoke/qtdragon-quit/checkresult
    A tests/ui-smoke/qtdragon-quit/skip
    A tests/ui-smoke/qtdragon-quit/test.sh
    M tests/ui-smoke/qtdragon/test.sh
    A tests/ui-smoke/touchy-quit/checkresult
    A tests/ui-smoke/touchy-quit/skip
    A tests/ui-smoke/touchy-quit/test.sh
    M tests/ui-smoke/touchy/test.sh

  Log Message:
  -----------
  Merge pull request #4054 from grandixximo/ui-tests-phase2

test: ui-smoke phase 2, g-code execution and endpoint check

Compare: https://github.com/LinuxCNC/linuxcnc/compare/106519fb3e01...81c3d892ba84

To unsubscribe from these emails, change your notification settings at https://github.com/LinuxCNC/linuxcnc/settings/notifications