Skip to content

slicecamd: expose fine-acquire goal and precision in the config file - #491

Closed
cfremling wants to merge 1 commit into
mainfrom
feature/expose-fineacquire-thresholds
Closed

cfremling wants to merge 1 commit into
mainfrom
feature/expose-fineacquire-thresholds

Conversation

@cfremling

@cfremling cfremling commented Oct 16, 2026 •

Copy link
Copy Markdown
Collaborator

Automatic acquisition in bad seeing requires setting the fine-acquire
convergence tolerance and scatter gate from the config file.

Both are compile-time constants in slicecam_interface.h:

  • goal_arcsec = 0.4
  • prec_arcsec = 0.35

Adds FINE_ACQUIRE_GOAL and FINE_ACQUIRE_PRECISION, parsed in
configure_interface alongside the other FINE_ACQUIRE_* keys.

Defaults in slicecamd.cfg.in are the current values; behaviour unchanged.

FINE_ACQUIRE_GOAL and FINE_ACQUIRE_PRECISION are parsed into
fineacquire_state.goal_arcsec and prec_arcsec, which were compile-time
constants. Defaults in slicecamd.cfg.in are the existing values, 0.4 and
0.35, so behaviour is unchanged.
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Oct 16, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-10-16T09:48:09.303764Z 32d6205 PR opened
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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 32d62050a0

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".

}

if ( config.param[entry] == "FINE_ACQUIRE_GOAL" ) {
try { this->fineacquire_state.goal_arcsec = std::stod( config.arg[entry] ); }

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Reject invalid fine-acquire goals

When FINE_ACQUIRE_GOAL is negative, nan, or inf, std::stod succeeds and the daemon accepts the value. In do_fineacquire, a negative/NaN goal can never satisfy the convergence comparison and there is no correction-cycle limit, while infinity marks every target as locked immediately; this can leave acquisition repeatedly commanding offsets or let observations proceed before the target reaches the slit. Require a finite, positive value before updating the state.

Useful? React with 👍 / 👎.

@cfremling

Copy link
Copy Markdown
Collaborator Author

Withdrawn. These are not the limiting parameters for acquisition in poor seeing; no change needed.

@cfremling cfremling closed this Oct 16, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant