x01dcandClaude Sonnet 5
28eac31ab0
CI for csaxs_bec / test (push) Failing after 5s
Mirko, once Phase 1 (tomo-queue backend) was live-confirmed working: "is
the way to move to the widget phase" -- this adds OMNY as a third setup to
the tomo params GUI widget, alongside flomni/lamni.
Key design decision: replaced the module-level TOMO_TYPES dict (shared
across setups, couldn't express "type 1 means 2 sub-tomograms here but 8
there") with a per-profile "tomo_types" dict:
{tomo_type_int: {"label": str, "n_subtomos": int | None}}. n_subtomos is
not None became the single, uniform test for "is this an equally-spaced
type" everywhere (combo box, section visibility, validation, queue-table
projections display), since OMNY has three such flavors (types 1/4/5 --
2/4/8 sub-tomograms) instead of flomni's/lamni's one (always 8).
Critical bug caught by verifying the design against the live file, not by
the initial design itself: _validate() hardcoded
`tomo_type not in (1, 2, 3)` -- would have made the rest of Phase 2 look
like it worked (types 4/5 selectable, sections visible, math correct)
while silently rejecting every actual submit/queue attempt for an OMNY
type-4/5 job. Fixed to check profile membership instead. Also fixed three
UI strings that hardcoded "8 equal sub-tomograms" to read the active
type's real count.
New _compute_type1_omny/_requested_to_stepsize_omny (parameterized by
tomo_type via {1:2,4:4,5:8}, matching
OMNY._equally_spaced_subtomo_angles()'s own mapping exactly) and
_compute_fermat_positions_omny (calls OmnyFermatScan's real static
method, same pattern as flomni's). _format_projections() gained a 3-way
setup duck-type (OMNY's param set is a strict subset of flomni's -- no
positively-unique OMNY-only key exists, so a negative-then-positive check
is necessary) that resolves into SETUP_PROFILES rather than duplicating
the type-count mapping a third time.
Confirmed needing zero changes (verified, not assumed): _job_tooltip(),
_build_type23_section(), compute_fermat_positions call sites,
_detect_setup() (osamroy as OMNY's discriminator), the busy-banner
detection methods (already generic via the shared tomo_progress global
var from Phase 1), and the Qt Designer plugin registration files.
New test_omny_tomo_params_widget_math.py (29 pure-function tests, no
QApplication/QWidget, same style as the existing flomni/lamni test
files). Full suite: 761 passed. Not live-tested (needs a real display,
same as every GUI change this branch) -- next step is Mirko trying it
against a real OMNY session.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QLrD7sVYGLAzsQjLVJpCgt