Skip to content

Bring 3.0.1 into release/3.1.0 - #604

Open
Osiris-DevWorks wants to merge 59 commits into
release/3.1.0from
claude/project-thread-gjgx0d
Open

Osiris-DevWorks wants to merge 59 commits into
release/3.1.0from
claude/project-thread-gjgx0d

Conversation

@Osiris-DevWorks

Copy link
Copy Markdown
Owner

Requested by Amr · project thread

release/3.1.0 was cut from main at 3.0.0, before 3.0.1 shipped. So it is missing every 3.0.1 change, and the 3.1.0 release PR can't merge cleanly. This merges main into it.

Conflicts resolved

To check

  • The SCO tests (test_sco_overlay.py, test_sco_game_screen.py, test_global_hotkey.py) need Windows. I couldn't run them here.
  • Overlay drawer on a second monitor: it should still open on the game's screen, and clicks should pass through while you aim.

Generated by Claude Code

claude and others added 30 commits November 4, 2026 20:19
Branch-opening commit for the 3.0.1 integration target.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFfwPYBmRRqt4rroHKi
The Enhancements: Categories step now names the categories as the
checkboxes show them. Config: Appearance mentions the font size, and
Config: Tools mentions Restore user.ini, Preview Apply and Check for
Updates. Two new steps follow Tools: Settings Backup (Export / Import
Settings) and Star Citizen Tools (Clear Shader Cache). The tour is now 21
steps. AI translations refreshed in all eleven languages, keeping each
file's own words for theme names, font sizes and overrides.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFfwPYBmRRqt4rroHKi
#522 had no number while it was a draft, so comments, docstrings, the
CLAUDE.md guides and test descriptions called its work "perf proposal".
#522 and #529 shipped in 3.0.0, so each of the 184 occurrences, in 26
files, now names the PR that did the work. No code changed: every changed
Python file has the same AST once docstrings are taken out.

Fixes #547

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two test comments pointed at file and line numbers that have moved:
generate_enhancements_ini.py:2817 is a blank line now, and
blueprint_tracker_tab.py:325,349 are no longer the two lists. They now name
ContractVariant, scan_contract_generators and the two list attributes. The
first comment also described destructures _run_gen_missions no longer has.
Comments only: both files have the same AST.

Fixes #549

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every test in test_mission_variant_tuple.py built its tuple with the file's
own _make_variant, so a 13th field in ContractVariant still passed all four.
Three of them copied loops and v[N] reads that went in 415b0b4 (#240) and
ran no generator code. One test now holds _make_variant against the real
ContractVariant, field by field. Another runs the real
scan_contract_generators and checks both of its ContractVariant(...) calls,
the only code that still passes the fields by position.

Also corrects the file's tests/CLAUDE.md entry, which had the 1.4.1 regression
the wrong way round, and removes a clause from a generator docstring that
claimed destructures that do not exist. The generator's code does not change:
its AST is the same once docstrings are taken out.

Fixes #551

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
scan_contract_generators built each variant with two calls passing all 12
fields by position, and logged a contract that raised at DEBUG, which a
normal run does not show. On a LIVE cache, one wrong call skipped all 1,809
contracts it reached and lost 787 entries of
mission_rewards_enhancements.ini, and the run still finished as a success.

A variant is now built by keyword, and a sub-contract's comes from its
contract's with _replace. A skipped contract gets a WARNING naming it and the
error. More contracts skipped than giving a variant, or a walk that stops
before any gives one, raise ContractScanFailed, so the run fails with the old
INIs whole and the set not certified. The worker matches that class by name
and words its box with two new strings, in all 12 languages.

So game data cannot do that, a spawn count written as 4.0 now reads as 4, and
any other that is not whole counts as 0 with a WARNING: 52% of LIVE's released
contracts read a count. Older gaps in the same scan are fixed too, none hit on
LIVE today: a bad count in a handler's own spawns ended the whole scan, a
skipped contract left its blueprints and items behind, and
PointsToAward="inf" lost the contract.

On a LIVE cache, every file the generator writes matches the unchanged code
byte for byte across 11 settings combinations.

Built on #552, whose test wording this updates.

Fixes #553

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A hand-mineable gem rock that has not been fractured shows its
hud_mining_fps_rock_name_* string (_1 Aphorite, _2 Dolivine, _3 Hadanite),
not its items_commodities_* name, so it showed no [CF] tag until it broke.
The commodity crafting pass now gives each such HUD name the tag its
commodity's name got.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFgHVmZmwQTwPYiS2Vi
…xxjo

Tag gem rock HUD names with their crafting tag
…#546)

The readout panel is part of the 3D cockpit, so a wider field of view draws
it smaller (7 px digits on a 2560x1600 screen at max, against the 11 px the
reader expected) and nearer the middle of the screen (about 0.39 down, the
edge of the old scan box). The Aurora's readout is as small.

- The reader looks for digits at 1.0 and 0.7 of the expected size
  (DIGIT_SCALES), sharing one mask and blob pass per threshold.
- SCAN_REGION reaches 0.45 down instead of 0.40.
- A comma must sit on the row's baseline. The taller box took a small blob
  far below the digits for the comma and cut 21,200 short.
- Reference glyphs from two max field of view Golem shots, and ALT_COUNT 4,
  since a new small 3 pushed a scaled-down 8 to fourth place.
- Fixtures fov1-fov5 and tests for the small digits and the stray comma.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFinijz9yEfxUtqqe9g
…cyvy

SCO: read the scan number at wide field of view and in small cockpits (#546)
…eeqd

Bring the guided tour up to date with 3.0.0
…r7ce

docs: Linux note that the overlay needs X11, not Wayland
A ship whose description key is blank in base.ini filled that key with its
stats. The game also shows such a key on the Carrack's interior map, so
the stats showed up there.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFgZu1TscYhoHQ42RrD
…9uxr

Skip ship stats on a blank description key (#561)
On Microtech in an Argo MOLE, with Windows HDR on, the scan decoder
rarely read the signature. Lossless captures of the overlay's own scan
area showed why:

- Over snow or a bright sky the orange panel behind the digits is as
  red as the white digits (both over 230), and the reader only looked
  at red. When no red threshold decodes, it now reads the blue channel
  at BLUE_THRESHOLDS, where the panel stays at about 125-180.
- HDR clips the white digits and they bloom by a pixel:
  - the comma joins the next digit's foot and hangs below the
    baseline, so the row had no comma and was dropped. _peel_comma
    takes it off. A "1"'s foot ends on the baseline and is left alone.
  - 12 px digits carry a 6 px comma where 9.9 px digits were expected.
    The comma is now sized against its own row's digits.
  - two 0s touch along a column with more ink than either hollow, and
    "00" was cut into three slivers. No part of a cut may be narrower
    than a "1" of that blob's height.
- Every threshold of a channel is read and the best decoded match kept,
  not the first: a low threshold blurred a 6 into an 8 that decodes too.
- ALT_MARGIN 0.3 -> 0.15: numbers no deposit has (70,000, 87,000) were
  swapped for ones that decode, and "12.0km" under an UNKNOWN marker
  read 12,000.

On 30 real HDR frames: 9 read right and 2 wrong, against 2 right and
0 wrong before. Ten hdr* fixtures (SCO captures at 1440 px, read at
their own size) and unit tests for each rule. Two scaled-down known
misses now read right. A read takes about 16 ms median against 8 ms,
88 ms at most, well inside the 250 ms capture tick.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…he bright-panel test helper

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Over snow with HDR the panel's blue reaches 230-245 behind the digits, and
where it glows behind the first digits it clips to 255 like them. Blue now
also reads at 245, and green at 250 is a last pass. On 30 labelled HDR
captures this reads 15 right (was 9) and 2 wrong (as before; one is now
87,000 read as 67,000, which decodes to no deposit), plus three readouts
the labels had missed. Five new hdr fixtures cover it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
sco: read the scan number on bright planets and with HDR
…label

Replace the perf proposal label with #522 and #529 (#547)
Name the code in two test comments instead of line numbers (#549)
…check

Test the real ContractVariant, not a copy of its shape (#551)
…errors

Stop contract scan errors passing silently (#553)
After a patch the applied global.ini stays in place without the keys the
patch added, which the game shows as raw @key names until Smart Citizen is
run again. The Play button takes its label from that file, so Apply now adds
a second line to it, "Run Smart Citizen after every patch", which is on the
menu the first time the game starts after a patch.

It is added in _merge_for_apply beside the version stamp and keeps its
rules: an earlier reminder is stripped first, and a missing key is skipped.
The game left-aligns the button's lines, so the shorter one is padded with
no-break spaces to sit centred, measured with the widths of the font the
button is drawn in, read from Data.p4k: Blender Pro Medium, or Noto Sans in
the Chinese, Japanese and Korean slots, one of which Russian borrows.
_ApplyMergeInputs carries the language slot for that, as a required field.
The loc-string line break is now one named constant across main_window.py,
and the three patterns that strip a line with the breaks before it are built
from it (each compiles to the pattern it replaced).

HELP.md in all 12 languages, the translations log and the CLAUDE.md guides
describe it.

Fixes #566

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The game logs the "Received Blueprint" notification as it shows it, so a
player who plays in Spanish logs "Plano recibido: <name>" and the scan,
which looked for the English text alone, found nothing.

The scanner now finds every notification line by its "Added notification"
part and checks its text against templates: the text of the loc key
crafting_hud_notification_received_blueprint in each language to look for,
with %s where the name goes. English is always one of them, for older logs.
main_window names the files the text is read from, for the channel being
read (its base.ini for the selected language as the table reads it, and
the global.ini applied to that channel's game), without making any folder,
through two new settings helpers, get_base_ini_leaf and
get_applied_global_ini. The Scan and the overlay's live watch read them on
their own threads.

Fixes #564

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Owned set holds display names, in the language that was loaded when
each was added. After a language switch an item owned under its old name
stayed in Owned under that name and showed again in Available under its
new one, so a Spanish player's list held the S-38 Pistol twice (#569).

Each load now renames an owned name the loaded strings do not know to its
name in them, through its loc key: the other languages' cached base.ini
files map the old name to its keys, and the loaded strings map the keys to
today's name. It renames only when every key leads to one loaded name, and
merges into that name when it is already owned. A name that matches no key,
or keys of two items, stays as it is. A scan's log names and Import Owned
Blueprints are renamed the same way, before the #372 recovery.

The other languages' files are never read on the GUI thread: the async load
reads them on FileLoaderWorker's thread, and the synchronous reload on a
LanguageNameTablesRead thread started before the merge. One copy per
language is read (the active channel's when it has one), and each file only
once a session unless it changes.

#570: the #372 repair's catalogue held items' short labels (an armour
set's name on its own), so an old logged name such as "Piernas de la
armadura Dust Devil" was saved as "Dust Devil", which matches no
blueprint. known_item_names now leaves _short keys out, and a load drops
saved labels from the Owned set. It keeps a label's text that another
language uses as an item's own name: the Spanish file keeps the S-38's
short label as "S-38 Pistol", its English name, which the rename then
turns into "S-38 Pistola". While a language's file could not be read, the
drop waits for a load that reads it.

The root, src/gui, src/utils and tests CLAUDE.md guides describe it.

Fixes #569
Fixes #570

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
LanguageNameTablesRead is a plain daemon thread, not a QThread, because
perform_merge_and_reload leaves it running when the reload stops before
the rebuild, and a QThread dropped while it runs aborts the app. The root
CLAUDE.md's threading model now says so.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…-language

Find received blueprints in the language the game shows (#564)
…md edits)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFeoV1R61WxRTVFv3Qy
xxxxxxxxxxxxx and others added 25 commits November 14, 2026 22:19
- src/utils/test_plan.py: the plan now covers 3.0.1's changes over 3.0.0
  (#566, #564, #569/#570, #546/#563, #555, #561, #545).
- docs/3.0.1-RELEASE-NOTES.md: release notes.
- RWolfyo added to Contributors in docs/ABOUT.md, README.md and each
  translated ABOUT.md (#563).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFeoV1R61WxRTVFv3Qy
…qyma

3.0.1 pre-release: test plan, release notes, credit RWolfyo
Every release gets a build provenance attestation for the installer and
the portable zip, and a VirusTotal scan whose links are added to the
release notes once the VIRUSTOTAL_API_KEY secret is set. SECURITY.md says
how to report a problem privately and how to check a download.

Split out of #558 so these ship in 3.0.1; code signing stays in 3.1.0.


Claude-Session: https://claude.ai/code/session_011hqxFg4JmXaXwTRySyJEQW

Co-authored-by: Claude <noreply@anthropic.com>
While the player aims, Star Citizen hides its pointer but lets it run over
the whole screen, so a shot fired as it crossed the drawer went to the
drawer, which took focus from the game. While Star Citizen is in front
with its pointer hidden, the drawer now lets every click through to the
game (WS_EX_TRANSPARENT on its own layered window), and it takes clicks
again once the pointer shows (F held, the mobiGlas) or another window is
in front. ScoOverlay asks Windows which window is in front and whether the
pointer shows, at once when the drawer shows and then every 15 ms while
the game is in front, every 100 ms otherwise, since Qt runs a timer of
20 ms or less on a 1 ms Windows timer. Turning it on closes a showing
tooltip, which is a window of its own.

Every pixel the tabs paint now opens a tab: a press on a tab's outline
opens the tab whose outline is nearest, and a press on the tab bar's
outermost line of device pixels, which Qt hands to the drawer above 100%,
opens its tab too. A right or bottom drawer no longer ends a device pixel
past its screen at 125% and 150% where a monitor or the taskbar lies
beyond. The drawer is placed and sized in one step, so a monitor at
another scale beside the main one no longer rescales an opening drawer.

The docs and the Overlay tab's description say what the overlay now asks
Windows, in all 12 languages, and the 3.0.1 Test Plan has a section for
it.

Fixes #582

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A click on the overlay drawer gives it focus. Closing the drawer with
the hotkey handed focus back to the game, but closing it with its own
tab did not: the drawer kept focus, and the game needed a click before
it took a key or a click again.

The hotkey remembers the window in front before it opens the drawer. A
click could not, because by the time Windows sends WM_MOUSEACTIVATE it
has already taken focus from the game (a logging build showed no window
in front at that moment, in game). So the drawer now notes the window
in front each time the pointer moves over it (WM_MOUSEMOVE, never the
drawer itself, dropped when the pointer leaves, WM_MOUSELEAVE), and a
click hands that window to ScoOverlay as the one to give focus back to
when the drawer closes. The hotkey asks the same way, so with the
closed drawer itself in front (after a refused hand-back) it takes
that noted window too. The drawer's nativeEvent
returns (False, 0) for every message, so Qt handles each click as
before. It never calls super().nativeEvent(), which crashes PyQt6 6.11
when the window shows.

The hand-back waits until no mouse button is held, so a tab close hands
focus back the way the hotkey close does, and the game never gets a
button's release without its press. Windows refusing the hand-back, or
another window in front 250 ms after it, is logged at INFO, one check
per hand-back for its own window and none once the overlay stopped or
the drawer opened again.

A click or the hotkey that names no window (none in front and none
noted) leaves the window already kept, and hiding the drawer drops the
pointer's note, since Windows may send no WM_MOUSELEAVE for a window
hidden under the pointer. The 3.0.1 Test Plan gets a section for this.

Fixes #585

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
review_pr on #584 found four Minors, three of them in files:

- The Test Plan's second "drawer while aiming" step said the game keeps
  focus right after a drawer tab click, but that click gives the drawer
  focus. It now has the tester click back into the game before aiming.
- The overlay entries in src/gui/CLAUDE.md and tests/CLAUDE.md named the
  report by its date. They cite #582, as the guides' other entries cite
  their issues.
- src/gui/CLAUDE.md put the Windows-platform check in set_click_through.
  It is in BlueprintDrawer.set_pass_through, and the entry now says so.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Brings PR #586 (issue/585-drawer-focus-back at 137e50f) into PR #584's
branch (issue/582-drawer-pass-through at 87eac80), so the #587 fix can be
built on both: all three change the overlay's focus and hotkey code and
the same docs, and #587's key check runs on #584's pass-through tick.

The conflicts are the ones #586's PR body describes. In
src/gui/sco_overlay.py both sides add code at the same three places (the
constants after FOCUS_CHECK_MS, the new BlueprintDrawer methods and the
new ScoOverlay.__init__ state), so both are kept, with a blank line added
between the two sets of drawer methods. The sco_overlay.py entries in
src/gui/CLAUDE.md and tests/CLAUDE.md, one long line each, are merged
sentence by sentence against the merge base. No sentence was changed
differently by the two sides.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Since #524 the overlay drawer's hotkey also registered the keyboard's raw
input with RIDEV_INPUTSINK while it was held, because Star Citizen keeps
registered hotkeys away while it has focus. That sent Smart Citizen every
key press, whatever window had focus. The raw input, its message-only
window, the WM_INPUT handling and the raw key matching are gone.
RegisterHotKey, WM_HOTKEY and MOD_NOREPEAT are unchanged.

So the hotkey still works in game, the overlay's pass-through tick now
asks Windows, while Star Citizen is in front, whether the hotkey's own key
is down (GetAsyncKeyState), and only on the look that finds it newly down,
whether Shift, Ctrl, Alt and Win are held, stopping at the first that
settles it. It asks about no other key, and nothing with another window
in front, where WM_HOTKEY works as before. No key press is sent to the
app, and the game still gets the keys. Pause, which Windows reports down
and up at once, counts by its "pressed since" bit. A press the check finds
waits 10 ms, so a WM_HOTKEY for the same press is handled first, and one
press toggles the drawer once however it came. The first check after each
registration is logged.

A test now fails on any module in src/ or scripts/ that names a raw input
or hook call, or, outside global_hotkey.py, a key state call, and a real
Windows test checks that no keyboard raw input is registered while a
hotkey is held.

The Overlay tab's description, HELP, the FAQ's privacy paragraph and the
README say what the overlay asks Windows now, in all 12 languages. HELP
adds that the keys need a moment's hold in game, and that Smart Citizen
has to run as administrator when Star Citizen does. The 3.0.1 Test Plan
has a section for it.

Fixes #587

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
From the review_pr walk of #586:
- The Test Plan's second #585 item now says to click the item with F
  held, since in game the pointer stays hidden until F is held.
- The overlay entries of src/gui/CLAUDE.md and tests/CLAUDE.md cite
  #585, as the guides' other entries cite their issue.
- The ten new click and hand-back tests that drive the overlay take
  conftest's slot_errors, so an exception in one of their slots fails
  that test instead of aborting the whole run.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…rough

Stop the overlay drawer taking focus while the player aims (#582)
Keeps both PRs' additions in sco_overlay.py and the overlay guide entries.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFeoV1R61WxRTVFv3Qy
Hand focus back to the game when a clicked drawer closes (#585)
Brings in #586's last commit (guide citations, Test Plan wording, slot test guard).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFeoV1R61WxRTVFv3Qy
Stop the drawer hotkey taking the keyboard's raw input (#587)
…RussoJr

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFeoV1R61WxRTVFv3Qy
…qyma

3.0.1 release notes: overlay drawer fixes, credit RussoJr
…Total action

- docs/LEGAL.md and the 11 translated LEGAL.md files: a "The Smart Citizen
  Overlay" subsection says what the overlay reads and asks Windows, and that
  it never reads game memory, hooks the game or sends input.
- release.yml: crazy-max/ghaction-virustotal pinned to the v4.2.0 commit,
  since it receives the VirusTotal key and GITHUB_TOKEN.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFeoV1R61WxRTVFv3Qy
…qyma

Legal tab: describe the overlay in the privacy section; pin the VirusTotal action
…qyma

3.0.1 release notes: Highlights for the scan decoder and overlay
Release 3.0.1
release/3.1.0 was cut from main at 3.0.0, before 3.0.1 shipped, so it
lacked every 3.0.1 change. Conflicts resolved:
- VERSION.TXT: keep 3.1.0.
- sco_overlay.py: keep #573's screen following beside 3.0.1's drawer
  pass-through (#582) and hotkey key check (#587); the drawer docks to
  the game's screen and keeps within_device_pixels on that screen.
- src/gui/CLAUDE.md, tests/CLAUDE.md: 3.0.1's text plus #573's lines.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011hqxFjEDADY2PnfwNXEq9T
@Osiris-DevWorks Osiris-DevWorks mentioned this pull request Nov 20, 2026
@Osiris-DevWorks Osiris-DevWorks added the 3.1.0 label Nov 20, 2026 — with Claude
@Osiris-DevWorks Osiris-DevWorks self-assigned this Nov 20, 2026
@claude

claude Bot commented Nov 20, 2026

Copy link
Copy Markdown

Code review

No issues found. Checked for bugs and CLAUDE.md compliance.

This is a main→release merge. I checked the sco_overlay.py conflict resolution (the drawer docks to the game's screen and within_device_pixels uses that same screen). I did not run the Windows-only SCO tests.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants