Add acquisition automation modes and acqmode command#391
Add acquisition automation modes and acqmode command#391
Conversation
them powered and waveforms loaded. The biases are turned off and commands are not sent while inactive. untested
include the activate states for the camera controllers, and updates CalibrationTarget::configure to read the new format
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a89f02d2d3
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| auto wait_for_guiding = [&]() -> long { | ||
| ScopedState wait_state( wait_state_manager, Sequencer::SEQ_WAIT_ACQUIRE ); | ||
| this->async.enqueue_and_log( function, "NOTICE: waiting for ACAM guiding" ); |
There was a problem hiding this comment.
In auto modes 2/3, wait_for_guiding creates a ScopedState for SEQ_WAIT_ACQUIRE, which clears that bit as soon as the helper returns (on guiding, timeout, or error). The acquisition thread already sets the same SEQ_WAIT_ACQUIRE bit in dothread_acquisition, so this new scoped state can clear the bit while acquisition is still running. Any clients or logic watching the wait-state (telemetry/UI or other waits) will see acquisition complete early and may proceed to expose even though acquisition is still active. Consider removing this ScopedState or using a separate state/refcounted mechanism for the “waiting for guiding” phase.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Will add a new state to wait for inside the mode loops to not mess with the overall acquisition state
There was a problem hiding this comment.
- Added SEQ_WAIT_GUIDE instead of SEQ_WAIT_ACQUIRE while waiting
for guiding. - Added ACQ_FINE_TUNE_XTERM so fine-tune can run in its own xterm when
enabled so it can be properly monitored. - Fine-tune process is now launched in its own process group and is terminated on seq abort . works even if seq abort sent when ngps_acq in progress
seq acqmode 1,2,3 TESTED with NGPS simulator mode
Commit: a20b509
sequencerd is subscribing to that and waiting on that to start an exposure
updates sequencerd.cfg.in other minor cleanup
- Add acqmode command to control acquisition automation (modes 1-3) - Mode 1: legacy manual mode (wait for user at each step) - Mode 2: semi-automatic (start acquisition, wait for user to expose) - Mode 3: fully automatic (apply offsets automatically) - Add repeat target detection to skip unnecessary slew/acquisition - Add configuration parameters for acquisition automation - Skip acquisition when re-observing same coordinates (modes 2-3) Based on PR #391 acquisition automation features Applied to DH/sequencer-calib-work branch (observatory baseline)
- Qt Widgets GUI for managing target sets and targets - Direct database access via MySQL Connector/C++ (X DevAPI) - Features: table views, inline editing, drag/drop reordering - Search functionality and sequencer controls - Build with CMake, reads DB settings from Config/sequencerd.cfg
- Add seq-progress launch script and X11 GUI (utils/seq_progress_gui.cpp) - Implement seq_progress topic publishing in sequencer - Track fine-tune, offset, and target state changes in real-time - Publish progress at key workflow points: * Before/after fine-tuning (tracks fine_tune_pid) * Before/during/after offset operations (tracks offset_active flag) * On target activation and completion - Add seq_progress topic and keys to common/message_keys.h - Add offset_active atomic flag to track offset state - Update utils/CMakeLists.txt to build seq_progress_gui with X11 This enables full 5-phase progress visualization (SLEW → SOLVE → FINE → OFFSET → EXPOSE) for on-sky monitoring and testing of acquisition automation. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The seq_progress_gui.cpp includes common.h which needs logentry.h from utils directory. Add PROJECT_UTILS_DIR to target_include_directories to resolve build error. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move utilities library after logentry and network in link order, since those libraries depend on symbols from utilities. Linker requires dependencies to be specified after the libraries that use them. Fixes undefined references to: tmzone_cfg, timestamp_from, get_time, get_latest_datedir, is_owner, has_write_permission, strip_control_characters Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CalibrationTarget::get_info() returns a const reference (not a pointer) and throws exception if not found. Updated calib_set() to: - Remove unnecessary null check (get_info throws exception instead) - Use '.' instead of '->' to access calinfo members (it's a reference) - Store as 'const auto &' to match return type Fixes compilation errors in sequence.cpp:2332, 2341, 2342, 2353, 2380 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Update MySQL Connector/C++ paths in ngps_db_gui CMakeLists.txt to search: - /include (not bare ) - /lib64 (where library actually is on production) - Prioritize mysqlcppconn8 to match main build This matches the paths used successfully by utils/CMakeLists.txt Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Remove exposure_progress > 0.0 check so percentage always shows - Show 'EXPOSURE 0%' immediately when phase becomes active - Add frame count for NEXP>1: 'EXPOSURE 2/3 12%' - Ensures percentage is visible even when starting mid-exposure
- Print received EXPTIME values to stderr - Show exposure_progress calculation - Help diagnose why percentage stays at 0% - Temporary debug output to be removed once issue resolved
- Remove percentage display (UDP EXPTIME not reliably received) - Show 'EXPOSURE X/Y' for NEXP>1 - Show 'EXPOSURE' for NEXP=1 - Keep EXPTIME parsing code for future use - Remove debug logging
- Show percentage on EXPOSURE label again - Debug: Print EXPTIME messages received via UDP - Debug: Print parsed values (remaining, total, percent) - Debug: Print exposure_progress after update - Debug: Print label being drawn every 10 frames - Will help diagnose if UDP is received but not parsed/displayed
- Print whether UDP listener will be started - Print msgport and msggroup from config - Print UDP listener file descriptor if started - Help diagnose why EXPTIME messages not received
Sequencer: - Parse EXPTIME UDP messages and republish to seq_progress ZMQ topic - Rate-limit: only publish when percent changes (reduces message spam) - Format: exptime_remaining_ms, exptime_total_ms, exptime_percent GUI: - Receive exptime_percent from seq_progress ZMQ topic - Apply smoothing (70% old, 30% new) to percentage - Works when UDP listener disabled (use_udp=0, msggroup=NONE) - Fixes issue where GUI shows 0% during entire exposure
- Set NGPS_ROOT=/home/developer/Software directly - Remove environment variable fallback logic - Simplifies launch: just run 'seq-progress' without setting NGPS_ROOT
- Find script's own directory using BASH_SOURCE - Set NGPS_ROOT to parent directory of script location - Works from any current working directory - Works for both /home/developer/Software and NGPS_tmp installs
- Use 'export' so child process inherits NGPS_ROOT - Fixes 'ERROR: opening configuration file Config/acamd.cfg' - seq_progress_gui needs NGPS_ROOT to find acamd.cfg for ACAM port
When waiting for user continue after acquisition/fine-tune failure,
the button now shows "OFFSET & EXPOSE" or "EXPOSE" depending on
whether the target has a database offset. A blinking red instruction
message tells the user exactly what clicking the button will do.
The sequencer already applies target offset before exposure in both
the guiding-failure and fine-tune-failure paths; the wait messages
are updated to reflect this ("apply offset and expose").
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sends 'native z' via tcsd before target_offset() in both the guiding-failure and fine-tune/post-fine-tune paths. This resets cumulative offsets from acquisition and fine-tuning so the observer sees only the intended target offset. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…d processes When sequencerd forks child processes (e.g. ngps_acq), the listening socket file descriptors were inherited by the child. If the parent was killed while a child was still running, the child held the port open, preventing restart. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Publish acq_automatic_mode in sequencer ZeroMQ progress messages and show it as a right-aligned label in the GUI title bar (MANUAL/SEMI-AUTO/AUTO). Also poll acqmode via TCP fallback for robustness. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
Cannot approve This PR introduces process management ( The NGPS system is built around:
This PR breaks those patterns by:
The system lockup you experienced is likely due to these blocking syscalls interfering with the multi-threaded daemon architecture - deadlocks are almost inevitable when mixing low-level process management with the existing async design. This PR is also relying upon an external C program whose code is not part of the repository. Recommendation: The approach of shelling out to external commands doesn't fit the NGPS architecture and creates significant reliability risks. |
I do agree that spawning a new process for acq isnt ideal. The ngps_acq communicates with acamd, sequencerd, and slicecamd. As such, I think a new daemon, acqd could make sense, but probably overkill. I think slicecamd is the place for this not acamd since its using slicecam image data |
cfremling
force-pushed
the
CF/auto-acq
branch
from
a20b509 to
456dcaf
Compare
|
@astronomerdave I added ngps_acq.c to slicecamd/ and got rid of the child process fork/spawning. slicecamd now has 'scam autoacq'. please re-evaluate the PR |
2 participants
This adds 3 mode settings for acquisition with various levels of automation.
mode 1 is identical to current (legacy acquisition)
mode 2 waits for seq continue between each step (goes through acquisition, fine tuning using scam, offset, expose)
mode 3 is full auto
the fine tuning step uses ngps_acq (now embedded in slicecamd)
needs comprehensive on-sky testing before pull approved