_on_tree_open used tree.selection() which may differ from the opened
node; switched to tree.focus(). Network calls in _live_tree_expand now
run in a daemon thread so the UI stays responsive during lazy loading.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
After open_wiresheet rebuilds the DOM and tree, re-find the target
element by its comp id (not the old tree item ID which no longer
exists) and re-select it with the restoring flag to prevent recursion.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Use RLock instead of Lock for _close_lock so close() can call
_shutdown() without deadlocking — previously the session never
marked itself closed, causing infinite keepalive sends.
Mask kit checksums to unsigned 32-bit for hex display so negative
i4 values render as proper 8-digit hex (e.g. d3984c51 not -2c67b3af).
Tested against live device at 192.168.1.251 — full handshake,
schema read, and component tree load confirmed working.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New checkbox in Preferences (default off). When enabled, every DASP
send (>>>) and receive (<<<) prints message type, seq/ack numbers,
payload size, and remote address to stdout.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Guard tree.get_children() with tree.exists() so a stale item ID from
a COV or lazy-load rebuild does not raise TclError.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Port the DASP transport and Sox protocol stack from sedonaOntrol Java
to Python, enabling live device editing alongside the existing offline
SAX editor. Lazy-loads components on navigation, caches them, and
subscribes to COV events for the active wiresheet.
New files: dasp.py, sox.py, device_session.py, connect_dialog.py,
live_hooks.py. Modified: app_state.py (live-mode state), main.py
(Device menu, COV polling, lazy tree expand), properties.py (Sox
write-through), sax_file.py (structural mutations route through
live_hooks), schema.py (find_manifest helper).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Scan sedona/platforms/db/ to find which native kits are compiled for
the SAX's platform. Kits with native code for other platforms, kits
providing a PlatformService for a different device, and their
transitive dependents are placed in a "Not supported" section with a
red-cross checkbox that cannot be toggled.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The save dialog now suggests e.g. normal_kits_20260911_143200.xlsx
so successive exports never collide.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Saves the kit table to an Excel file with section headers, check
marks, and all four data columns. Default filename is based on the
open SAX file.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Renamed "Kits to be installed" to "Kits available on the computer"
for kits whose dependencies are all in the app. Kits with unmet
dependencies go into a greyed-out "Other kits on the computer" section
that cannot be selected or ticked.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
No columns stretch — sizes are set for the longest kit name
(ontrolModbusTcpSlaveSmart) and action text (Keep with no checksum).
A horizontal scrollbar appears when the pane is narrower than the
total column width.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The boxes dict was missing the (False, False) key — an unchecked kit
that is also required — which caused a KeyError when the Kit Manager
opened automatically for files with incomplete checksums.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When a SAX file is opened and any declared kit is missing a checksum,
the Kit Manager panel opens automatically and the palette stays empty
until the checksums are resolved.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The sidebar palette combo now lists kits from the open file's schema
rather than every kit installed in Sedona home, with their resolved
version labels. Falls back to all installed kits when no file is open.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Manage Kits now opens inside the editor pane (like Properties and Link
View) instead of a separate Toplevel window. Also normalize hex
checksums by stripping leading zeros so .kit and manifest filenames
that disagree on padding (e.g. 462d43e vs 0462d43e) still resolve to
a version.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Opening a Properties or Link view was invisible to the history, so Back
from one did two things at once: it took the panel down and stepped to the
sheet before the one it was opened over. There was no way back to a
picture you had just been reading.
A history entry is now a (kind, element) pair naming what the editor pane
held -- a sheet, a Properties view, or a Link View -- and restoring one
puts that back: a sheet entry selects its tree node, a panel entry
re-opens the panel. So opening a panel is a move like opening a sheet.
Back from a Properties view lands on whatever came before it and Forward
brings the panel back, and following a link out of the Link View now has a
way home: Back returns to the picture the link was followed from, and a
second Back to the sheet that was open before that.
The panel's own Close is deliberately not recorded. It means "leave this
view", and recording it would put a sheet entry after the panel's, so Back
would re-open the panel that had just been closed. Leaving the position
alone means Back steps to what came before the panel and Forward re-opens
it, which is what the arrows are for.
Both arrows still go through close_editor_panel first, so unsaved
Properties edits are asked about exactly as before and a refusal leaves
the view and the position where they were. Pruning now has two kinds of
entry to check -- a sheet needs a tree node, a panel needs its component
still in the document -- and collapses any pair it leaves naming the same
view, since a Back that visibly does nothing is worse than a shorter
trail.
sheet_history and its functions are renamed for what they now hold.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The picture was right and unreadable: eleven links round one component
drew twelve wire crossings, because each column was stacked in the order
the file happens to list its links and the lanes were handed out by how
far a wire had to travel.
Both columns are now stacked in the order of the chosen component's own
slots -- what feeds its first slot at the top, what feeds its fifth below
that. Nothing here is read from meta, so the view was always free to put
the peers where they read best, and this is the order that turns the set
of wires on one side into a ladder: both ends run the same way down the
page, so no wire has any reason to cross another.
That leaves which lane each wire turns in. For a ladder there is an order
that cannot produce a crossing, and read off the slots rather than off the
direction the line is drawn in, it is the same on both sides -- two fans,
each nesting outwards from the component's own row:
reaching a peer slot above its focus slot -> turns nearer the focus
the higher up the focus it lands
reaching a peer slot below its focus slot -> turns nearer the focus
the lower down the focus it lands
Two wires of one fan, the outer going round the inner, pass above and
below its vertical but never through it; and a wire of one fan can never
share a stretch of the page with a wire of the other, so the two fans
reuse each other's lanes and the gap stays as narrow as the picture
allows. Where several wires leave the same slot no order of focus heights
separates them, so the longest branch turns first and the rest carry on
along the shared run beside the pin -- a wiresheet's own fan-out. Where
two slots already line up the wire is one straight line and is settled
before anything has to be fitted past it; the layout is rounded to whole
pixels so that a pair that ought to line up does.
The reasoning is not trusted on its own: every wire is checked against the
ones already routed (routes_meet) and takes the next lane out if it
clashes. One thing genuinely breaks the ladder -- a peer whose own slots
run counter to the focus's, linking its row 4 to row 3 and its row 3 to
row 4. Ends inverted like that have to cross wherever they are drawn,
since neither box's slot order is this view's to change.
Measured over 461 linked components in four real files, 1113 wires: no
wire passes through a box at all, and ten crossings remain in six
components, every one of them an inverted pair of that kind. Before this
the same four files drew 45.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Right-clicking a component now offers Link View alongside Properties, from
the wiresheet and from the navigation tree alike. It takes over the editor
pane the same way and answers one question -- what is wired to this
component, and where does it go.
The chosen component sits in the middle wearing the stroke a selected box
wears on the sheet, with what links into it on the left and what it links
out to on the right, so every arrow points the way the signal travels.
Pale yellow ground and no grid: nothing here can be positioned, because
the layout is derived from the links every time it is drawn and no part of
it is stored in the file. Read-only throughout -- no drag, no add, no
delete -- so a component's wiring can be read without the risk of
rearranging it.
Every line is grey, since this is not a wiresheet and a wire-blue line
would invite the reader to treat it as one. A link on the same sheet is
solid; one that crosses is dashed, keeps the grey knob that sheet paints
on the slot, and has the far component drawn in with its sheet named
beneath it. A reference the document cannot resolve gets a placeholder
naming the path, because a link the view silently dropped would be the one
thing a reader could not find out from here.
Clicking a wire or a component follows the link: it opens the sheet of the
end that is not the component the view is about, with the link selected in
red. A link that stays on one sheet therefore goes back to that sheet
where both ends are drawn, and one that crosses opens the other sheet --
the end the view could only show as a knob.
Supporting changes:
- draw_component_box is factored out of render_components, so the sheet
and this view paint a component with the same code and a box cannot read
differently between them. draw_knob takes its canvas for the same
reason.
- The editor pane's hand-over is now shared: _take_editor_pane,
_give_editor_pane_back and close_editor_panel replace the
Properties-only versions, so "you have not saved that" is still asked
once, in one place, whichever panel has the pane.
- Fixes the jump to a link's other sheet, which arrived with the link not
selected. An after_idle callback is not late enough -- ttk dispatches
its own selection event later than one of those, so the repaint wiped
the selection. The link is left in _pending_link_selection for
on_tree_select to honour once it has painted.
- Each gap is sized to hold a lane per wire crossing it, and lanes are
handed out nearest-target-first, so a fan of a dozen links nests instead
of crowding a dozen verticals into a 30px band.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The four .sax test files live in EditorTestSax under the Sedona test
apps directory; the copies here were byte-identical duplicates and no
code referenced them.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Names the four columns -- Name, Type, Value, Unit -- in grey at row size over a
hairline rule, so it labels the table without competing with it.
A fifth column of raw flags= letters was tried alongside it and taken out
again. c, a, s and o are the manifest's own shorthand, and a column of them
told the reader nothing the row was not already saying: c is what the white
field means, a is what an "action" value means. What the flags decide is still
decided by them; it is just not spelled out twice.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@hidden is now filtered by default. niagara.html says hidden slots "are
designed to be invisible to the user" and that interfaces should rarely
display them; every one under Sedona home is a void action a kit uses
internally, such as HdwAI.checkType. A "Hidden slots (n)" checkbox in the
footer brings them back and the answer sticks for the session. The checkbox
only appears on a type that has some, and the rows are always built -- the
toggle grid_removes and re-grids them, so an emptied grid row collapses rather
than leaving a gap.
A bool now says which state it is in by the colour of its field rather than by
a dot beside it that had to be learnt: green true, red false, blue null, with
the words left black so they read as easily as every other value. Dropping the
dot also lets a bool's field line up with the numbers above and below it.
Two things that needed changing to get there:
- a dropdown is offered only where there is a choice to make. A read-only
bool or enum gets a plain field, since a disabled dropdown is an arrow
that does nothing -- and a read-only bool is still filled, because a
runtime bool has a state too.
- the dropdown is a tk.OptionMenu, not a ttk.Combobox. On the Windows themes
a Combobox draws its field with a native element and ignores
fieldbackground outright, so a bool could never be filled with its colour.
An OptionMenu reports through its variable rather than an event, which means
loading a row looks exactly like choosing one -- so the load is flagged and the
unsaved-state tint left alone until every row is in, or the view would open
already claiming edits nobody made.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Right-click a component on the wiresheet or in the tree and Properties takes
over the editor pane, listing every slot the manifest declares rather than the
handful a box paints. Config slots that are not @readonly are white and
editable; everything else is grey.
Rows are built from the slot's facets, read against SedonaDoc/niagara.html:
@trueText/@falseText, @range, @precision, @unit, and @min/@max, which bound
what Save accepts. Three facets it had wrong:
- null is a value, not an absence. control::WriteFloat pairs 'in' with a
setNull action and the sample app stores val="null" on four of them, so
null now shows, parses back, and reads as @nullText where declared.
- @min/@max on a Buf count characters. TcpDevice.ipAddr is a Buf(17) with
@max 16; a longer string is refused rather than overrunning the device.
- an int's implicit range is noise. [-2147483648 - 2147483647] said no more
than "a whole number", so only byte and short print a range they did not
declare.
meta gets a row of its own: the '>>' opens its user groups as checkboxes and
its position as x and y, since typing 822411269 into a field sets neither.
Save writes only the rows that moved, as one undoable SlotValueCommand, and
refuses the lot if any row fails its range check. An edited field paints the
view in the canvas's own dirty tint, and every way of leaving asks before
discarding.
The file name on the tree's root row goes red while anything is unwritten,
counting both a dirty document and unsaved fields. Two things this needed:
the nav tree runs on its own style, because a themed Treeview's selected
foreground beats a row tag's; and the tag is added to or removed from the row
rather than recoloured, because reconfiguring a tag does not repaint rows that
are already drawn.
Also fixes a latent hazard the panel exposed: Del and Backspace are bound
application-wide, so they now stand down while a text field has focus rather
than deleting the component behind it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Right-clicking a component in the navigation tree offered Delete Component
alone, while the same component on the wiresheet also offered Link Mark and
Link From / Link To. Which list a component was picked from should not change
what can be done with it, so both popups are now built by
fill_component_popup.
Its entries close over the component instead of reading a selection back,
which is what lets each popup act on the row or box that was actually clicked
-- the canvas commands went through app_state.selected_canvas_tag, which a
tree row never sets. do_link_mark, do_link_from_mark, do_link_to_mark and
delete_tree_selected_component fall out with it, and Link Mark now says in the
status bar what it marked, since nothing else on screen shows that.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Right-clicking a wire or a knob now offers Go to Link Wire Sheet "<name>",
naming the sheet the far end is on -- the end a grey knob stands in for. A link
with both ends on the open sheet has nowhere to go, so the entry is left off
rather than greyed out.
find_component_by_path turns a link's from/to reference back into the component
it points at, the inverse of get_component_path, and the sheet is that
component's parent; a parent of <app> is the file's root node, which the popup
names after the file.
Arriving selects the same link and scrolls it to the middle of the view. It is
found by its element rather than by the tag it was picked on, since a repaint
hands out fresh wire tags, and it waits for that repaint from an after_idle
callback: ttk queues its selection event rather than dispatching it, so the
sheet is not painted when the tree node is selected -- and painting a sheet
clears the selected link, which was wiping the selection this feature had just
made. The jump is an ordinary sheet opening, so Back returns to where the link
was picked.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Fits the whole painted sheet in the visible canvas and scrolls it into view,
from the View menu, Ctrl/Cmd+9, or the four-corners button after the zoom
percentage. The ladder gains 25% and 33% at the bottom, so a wide sheet has
somewhere to fit into.
Fit is the one scale that is not on the ladder -- a sheet fits at whatever
scale it fits at -- so zoom_step now takes the first level above or below the
scale in force rather than the neighbour of the nearest rung.
The sheet is measured from item coordinates rather than canvas.bbox(), which
is inflated by the outline stroke: a stroke is the same two pixels at every
zoom, so a bbox divided back to 1:1 reports a slightly smaller sheet the
further in it is zoomed, and fitting an already fitted sheet crept from 288%
to 291% instead of standing still. Anything painted left of or above the
canvas origin is measured from the origin, since no amount of scrolling brings
it into view.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A ladder of scales from 50% to 300%, stepped by Zoom In / Zoom Out / Zoom 100%,
by the two buttons beside the navigation arrows, and by Ctrl and the wheel,
which holds the point under the pointer still. A menu or keyboard zoom holds
the middle of the view instead, so it closes in on what is already being
looked at.
Painting and routing stay at 1:1, because that is where every metric is tuned:
box header and row heights, wire stubs, lane gaps, standoffs, knob radii. Zoom
scales the finished picture, so a zoomed sheet keeps the proportions of a 1:1
one and the router reaches the same decisions at every scale. Text comes along
because the three box fonts are resized once the boxes have been measured with
them.
redraw_links replaces the direct draw_links calls a drag or a drop makes: the
router reads box positions off the canvas, so the boxes go back to 1:1 for the
routing pass and the result is scaled again. Canvas coordinates carry the zoom,
which is why the drag, release and drop handlers now snap against
meta_scale * zoom -- a box dragged the same distance across the screen lands on
the same meta step at 50%, 100% and 200%.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Cmd+W closes the window on macOS, so the Command half of the pair is left
alone; this is the one shortcut that is not bound on both modifiers.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Joins the Command/Control shortcut pairs, so the chord works the same on
either platform, and the View menu entry now shows it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
It re-flows what is on screen rather than editing the document, which is where
the rest of the sheet's presentation controls now live.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Tree / Palette hides the left column so the wiresheet has the whole window.
ttk has no hide flag for a pane, so the column is dropped from the paned
window and re-inserted at the front, which is why the configured width is
re-applied afterwards.
Back and Forward walk the sheets opened this session, with two arrow buttons
in a new toolbar on the same commands. History entries hold the element whose
children are painted rather than a tree item id: rebuilding the tree hands out
fresh ids, while the elements themselves survive. Recording runs from the tree
selection handler, the one route every sheet opening takes, and is suppressed
while a walk drives that selection so it cannot rewrite its own trail. Opening
a different file retires the history; a delete or an undo drops only the
entries whose element has left the document, keeping the position on screen.
open_file builds the tree inline instead of going through
refresh_treeview_from_dom, so it never raised the tree-filled callback and the
history had no way to see a new file. It and reset_to_startup_state now fire
it, which also gets the tree column fitted on load rather than on the first
resize.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The tree and palette column took a share of every extra pixel the window
gained. It is now pinned to a width set in Preferences, defaulting to
230 px, so widening the window only ever feeds the wiresheet.
That fixed width brought two existing weaknesses into view:
- Long paths and type names had nowhere to go. Both lists now size their
tree column to the widest visible row - ttk derives the horizontal
scroll range from the column width, so a column left stretching can
never scroll - and carry vertical and horizontal scrollbars that leave
the grid when they have nothing to scroll.
- A short window squeezed the navigation tree away entirely. Both lists
asked for ten rows, so the palette requested the most while the tree,
carrying the larger weight, gave up its height first. The requested
rows are now 6 and 5, a 300 px minimum window height keeps the
shortfall shallow, and clamp_left_pane_split holds a dragged sash
between the two floors.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Adds control::Add2 under /Fire and links /AHU20_1/RetPres/RetSet.value to its
in1. That slot already had two links inside RetPres, so it now carries both an
unreachable same-sheet link and one leaving the sheet -- the only arrangement
that draws the half-and-half knob, and one no sheet in this file produced
before. It caught that marker rendering wrongly at its real 8px size.
RetPres is also tidied, which moved seven meta positions onto one row.
The diff covers the whole file because saving re-serializes it: the schema and
every other component and link are unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Pushed into the wire-coloured half, the centre filled almost all of it: that
half is only KNOB_RADIUS wide and the centre is KNOB_CORE_RADIUS, so at the
real 8px knob size the wire colour was reduced to a sliver and the marker read
as plain grey. Centred, both halves keep enough colour to tell apart.
Found on a slot carrying two links on its own sheet plus one leaving it, which
is the only arrangement that produces this knob and did not occur in testSax
until now.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Layout rules for tidy and for dragging:
- Boxes snap to whole meta steps live while dragged, not only on release.
- The left-most column and top row clear the canvas edge by at least 30px,
and neighbours by at least 30px, each rounded up to the next step.
- Box widths round up to the grid so a right edge lands on a grid line.
- Column gaps widen only where wires need lanes, so most stay at exactly 30px.
- A sheet too wide for meta's 0..255 range is laid out at 2, then 3, then 4
grid squares per step. The multiplier is derived from the sheet's contents,
so painting and tidying agree on it without recording anything in the file.
The grid size in Preferences is untouched.
Wire routing:
- No wire may pass under a component. A link with no clear route is shown as
knobs on both slots instead of a line hidden behind a box.
- Vertical runs hug a box within 15px, never closer than 10px, and two runs
sharing a gap stand 10px apart.
- Horizontal runs of different links stay 5px apart. Runs from one output pin
may share, since a fan-out is one signal.
- Straight routes claim their rows first, as their segments are pinned to
their slots and cannot move later.
Knobs:
- Wire colour with a small white centre when the far end is on this sheet,
grey when it is on another, half and half when a slot carries both.
- Selectable and deletable like a wire, and raised above the boxes so a click
reaches them.
- Selecting a same-sheet pair draws a red dashed line between the two knobs.
Also:
- get_box_bounds returns the rectangle actually drawn rather than canvas.bbox,
which was inflated by the outline stroke and by selection.
- Preferences offers to round the box width to a multiple of the grid.
- The tidy coordinate-limit warning said to lower the grid size; raising it is
what helps, since the reach is 255 * grid size.
- Tidy Layout moves from its own menu into Edit.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Saved back out of the editor, so the fixture now exercises what the kit dialog
writes: every kit that was declared bare carries the checksum of its latest
installed build, and ontrolModbusSlaveSmart is declared at 929218bf -- the file
used its ModbusSlaveSmart component without declaring the kit, which left five
component types unresolvable on load.
Its .bak copy goes with it; the file it backed up is two rewrites old.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Placing a palette component no longer edits <schema> straight away. The kit is
noted as pending, its manifest is compiled so the box can be painted, and the
status bar says the kit will be declared on save. The <schema> block is then
reconciled when the file is written, in the one place that can report the whole
kit list the file will carry, with each kit's version and checksum.
What a kit's manifest depends on counts as used too, transitively, so placing a
component can bring in more than one entry: hvac pulls in types, ontrolPion
pulls in ontrolPlatform.
New Kits > Manage Kits... dialog lists every kit in three groups -- needed by
the app, declared but unused, and merely installed -- with the pinned build, the
newest installed build, and an Action dropdown of every build a kit ships.
The tick means the schema declares the kit: grey and fixed for kits the app
needs, unticked to remove, ticked to add at the newest build. "Upgrade all kits
to latest version" moves every declared kit; "Update kits" applies the ticks and
the Action column, additions and removals included.
Every schema change is applied as a batch, then verified: a missing manifest or
a component type that stops resolving reverts the whole block and reports why,
and a kit the app needs cannot be removed. Changes are one undo step, which
recompiles the slot registry in both directions since a repin can change a
component's slots.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Parse the readonly facet from kit manifests and use it to disable
slots that cannot participate in a link: read-only slots are greyed
on the target side only (they remain selectable as sources), action
and void slots are greyed on both sides, and type mismatches are
greyed when the other side has a selection. A status label previews
the link path at the bottom of the dialog.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implement full link workflow between component slots on the wiresheet:
- Link Mark / Link From / Link To context menu items for creating links
- Link dialog with slot type validation and source/target swap button
- Selectable link wires (click to highlight in red) with right-click delete
- Undo/redo support for both add and delete link operations
- Component deletion already cascades to remove related links
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The relative path failed when the working directory differed from
the script directory at runtime.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
A kit with no manifest under the Sedona home folder called sys.exit(1),
tearing down the whole app over one unopenable file.
schema now raises MissingManifestError instead, collecting every
unresolved kit first so one dialog names them all rather than the user
fixing them one reopen at a time. describe_missing_manifests builds the
body, pointing at the configured Sedona home and at Preferences.
open_file catches it, reports it, and calls the new
reset_to_startup_state to return to the just-launched state: no file,
empty tree, empty sheet, cleared registries and undo stack. The generic
except now resets too, since a half-parsed file previously left
current_file_path and a stale root behind.
Placing a palette component hits the same loader, where a full reset
would be too destructive: it rolls back the <kit> entry it just added,
rebuilds the previous registry, and abandons only that placement.
schema no longer imports sys or tkinter, so it reports nothing itself
and stays testable without a display. base_window_title moves to
app_state so the reset can restore the title without duplicating it.
README: document the new behaviour, drop the fixed gap, and correct the
stale "no undo, no keyboard shortcuts for save" claim.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
ArdaandClaude Opus 4.6