Skip to content

Dev - #208

Merged
Mzying2001 merged 49 commits into
masterfrom
dev
Nov 9, 2026
Merged

Dev#208
Mzying2001 merged 49 commits into
masterfrom
dev

Conversation

@Mzying2001

Copy link
Copy Markdown
Owner

This pull request introduces a new native command proxy component and a database abstraction with a SQLite implementation, updates project documentation to reflect these additions, and removes the obsolete EmptyExe subprocess. The changes improve the application's infrastructure for command forwarding and persistent storage, while also updating the documentation and project structure accordingly.

New Components and Features:

  • Added CefFlashBrowser.CommandProxy native project, which implements a command proxy executable with Flash diagnostic filtering. This includes a new main.cpp with logic for forwarding commands and handling diagnostics, and a corresponding MSBuild project file. [1] [2]
  • Introduced CefFlashBrowser.Database project, providing a database interface (IDatabase, IDatabaseSession) and a thread-safe, asynchronous SQLite implementation (SqliteDatabase). [1] [2] [3]

Project Structure and Documentation Updates:

  • Updated the documentation in AGENTS.md to describe the new Services responsibilities (including database and favicons), added entries for CefFlashBrowser.Database and CefFlashBrowser.CommandProxy, and listed new test and regression directories. [1] [2]
  • Removed the obsolete CefFlashBrowser.EmptyExe project and its entry in documentation, as it is no longer needed with the new command proxy. [1] [2] [3]

Documentation and Convention Improvements:

  • Refined documentation about commit conventions and removed outdated browser regression notes, streamlining contribution guidelines.

Build Configuration:

  • Added a custom property for CefSharp binaries directory in the main WPF application's project file.

Add an object dependency property for fallback content and render it in the control template. Preserve favicon priority and use the existing page icon when no custom default is set.
Store discovered page-to-icon URL mappings in a shared SQLite database and
refresh icons across favorites menus and the manager. Preserve navigation
ownership, image failure fallback, edit selection, and orderly shutdown.

Add database and favicon service tests, isolated WPF/CEF regressions, and
package smoke checks for x86 and x64 native SQLite loading.

Validation: x86/x64 Debug and Release builds, 212 Release unit tests per
architecture, complete browser regressions, and published archive smoke checks.
Favicons were embedded in header templates, leaving the MenuItem icon
column unused. Bind Icon to each website and render it with a dedicated
data template while keeping header templates text-only.

Extend browser regressions to cover icon placement, default icons,
header text, edited favorites, and live mapping updates.

Validation: x64 Debug solution build, targeted favicon regression,
and complete WPF/CEF browser regression suite.
Remove CefFlashBrowser.Database and System.Data.SQLite from the Costura exclusion list so they use the default assembly embedding behavior.
Normalize page keys by removing queries, fragments and trailing slashes while resolving relative icons against the original page URL. Let schemeless favorites match HTTP/HTTPS pages, then homepages and the latest updated icon for the same host and port.

Preserve mapping timestamps and deterministic fallback selection across initialization, retries and restarts. Cover the matching rules with unit tests and all three favorites surfaces with browser regression checks.

Validation: Debug x64 solution build, 247 unit tests, and WPF/CEF browser regression checks passed.
Centralize schema initialization and favicon persistence in DatabaseService so database consumers share a reusable service. Preserve lazy initialization, retry behavior, and connection ownership, and update IoC wiring and regression coverage.
Use exact page mappings before same-origin homepages for explicit and schemeless addresses. Remove the latest-site-icon fallback and index while preserving URL normalization and HTTPS-first schemeless matching.

Update lookup, initialization, and browser regression coverage. Debug|x64 build and 270 unit tests passed. Favicon browser checks passed; the full browser suite stopped at a keyboard foreground-window timeout.
Replace the dependency property with a get-only ICollection<string> initialized per browser instance. Keep ObservableCollection as the backing collection to preserve change notifications.
Disable the default dual-architecture native copy for the application and stage the selected library under Assets\SQLite. Remove legacy SQLite copies while preserving unrelated output files, and preload the native library by absolute path before creating the database.

Update package smoke checks for Costura embedded dependencies and verify that SQLite loads from the isolated package. Keep database and unit test project layouts unchanged.

Validation: Debug and Release builds for x86/x64, 270 unit tests, both package smoke checks, incremental cleanup, Clean, and native load failure diagnostics passed. Browser regression completed 294 assertions before the keyboard foreground check timed out.
Replace the static LanguageManager with a singleton LanguageService
registered and cleaned up by Program, preserving language behavior.

Resolve the service at application call sites and update browser
regression fixtures to use the registered instance. Supply the missing
language and settings icons required by the main-window checks.

Validated with Debug x64/x86 builds, 270 unit tests, 563 browser regression
assertions, and nine-language switching with cross-process settings restore.
Remove the unused interface and resolve the concrete singleton throughout the app. Update unit tests and regression scripts while preserving property notifications and service lifecycle behavior.
Initialize language resources and apply the saved language during service
construction so startup no longer needs an explicit initialization call.
Inject ILogger, track the applied language, and simplify resource lookups.

Use the public service methods in browser regression fixtures and cover
automatic initialization, language notifications, and fallback behavior.

Validation: Debug|x64 solution build, 270 unit tests, and 593 browser
regression checks passed. The browser suite passed on a fresh-process
rerun after an initial native keyboard timeout.
Move the favorites collection and JSON persistence out of GlobalData into
FavoritesService with an injected file path and logger. Load on construction
and register a lazy singleton so all views share the same collection.

Update view models, XAML bindings, and regression fixtures. Replace the
save message with direct service calls on manager close and application exit,
without resolving unused services during shutdown.

Validation: Debug|x64 solution build, 270 unit tests, and 612 browser
regression checks passed. The complete browser suite passed after two
runs timed out in native Ctrl+W checks.
Load settings and prepare data directories before App construction and resource loading. Keep initialization inside the primary-instance branch.

Validation: Debug|x64 solution build with Visual Studio MSBuild passed.
Move tab layout, scrolling and drag previews into a ListBox-based control. Expose an IsDragEnabled dependency property and bubbling MoveItem requests so consumers own collection changes.

Keep HandyControl colors in BrowserWindow and add standalone control regression coverage.

Validation: VS MSBuild Debug/x64, 270 unit tests, browser regression suite, and theme checks without HandyControl resources.
Move converter resources into App.xaml in alphabetical key order and share TwoThirdConverter across browser views. Remove the separate dictionary and update browser regression setup to register the application converters.
Register tab widths as dependency properties with the existing defaults and automatic layout invalidation. Rename MoveItem to RequestMoveItem, align its wrapper documentation, and update consumers and regression checks.
Tab selection previously cleared window fullscreen, including when a foreground tab was created. Preserve fullscreen across tab changes and skip forced fullscreen exit for context-menu actions that open tabs in the same window.

Validation: Debug|x64 build, 270 unit tests, and the full browser regression suite passed.
Replace per-language view models with language codes, a display-name converter,
and observable current-language state. Keep menu checks and dialog selection
synchronized, including virtualized initial selection and fresh event parameters.

Migrate browser regression callers to CurrentLanguage and cover language names,
property notifications, and synchronization between the menu and dialog.

Validation: VS MSBuild Debug|x64, 270 MSTest tests, and browser regression suite.
Keep localization keys private, extract display-name resolution, and reject null keys when constructing localized items.
Window.Activate can fail or lose foreground ownership during keyboard setup. Retry with a native activation fallback before holding Ctrl and always detach the temporarily joined input queues.

Add a competing-window regression while preserving foreground checks between consecutive Ctrl+W presses.

Validation: focused tab regressions and the full Debug|x64 browser regression suite.
Store versioned CurrentUser DPAPI ciphertext in Password and expose a nonserialized plaintext property for proxy authentication and editing. Discard legacy plaintext and unreadable credentials, remove sensitive configuration logging, and use a masked password input.

Add unit and browser regression coverage for encryption, settings persistence, password editing and clearing, and log privacy. Validated the Debug x64 solution build, 295 unit tests, and proxy settings regression checks.
Replace EmptyExe with a native Windows command proxy that suppresses only
the Flash sandbox diagnostic. Forward other raw command lines to system
cmd, preserve standard streams and full exit codes, and restore COMSPEC
for forwarded children.

Update solution packaging and add executable regression checks covering
both architectures, CRT callers, quoting, streams and console events.

Validation: x86/x64 Debug and Release builds, command proxy regressions,
295 unit tests per architecture, browser regressions and release package
smoke checks passed.
Serve the HTML from an embedded resource through a per-browser resource handler, retaining the virtual file URL for local SWF compatibility. Stop copying the player page into build outputs.

Extend browser and package regression checks for embedded resources, reloads, source changes and independent player windows.
Deploy CefSharp directly to Assets\CefSharp using the package targets
instead of moving a hard-coded file list. Extract the remaining post-build
work into a batch script with explicit build environment variables and
extraction error handling.

Validated x86/x64 Debug and Release builds, incremental builds, batch
environment/path and failure cases, and x64 Debug browser regression.
Move the proxy output to Assets/CmdProxy and update runtime paths, regression scripts, and package checks.
Separate deployed runtime dependencies from application resources by moving
CefSharp, CmdProxy, SQLite and plugins into Binaries. Update runtime paths,
project deployment settings, regression scripts and documentation.

Extract platform archives under source Assets in isolated temporary folders,
then overwrite deployed DLLs and copy the plugin readme in postbuild. Keep
archives out of build output and clean extraction files on failure or success.

Validated with 295 unit tests, browser regression checks, x86/x64 Release
package smoke checks, and postbuild failure and readme-copy scenarios.
Bring in PR #206 per-tab mute support while preserving the dev branch's
blocked SWF collection, language service, and browser tab refactors.
Resolve regression resource and menu conflicts, migrate new mute labels
to LanguageService, and restore the required WPF dependency-property import.

Validation: VS MSBuild Debug/x64, all 295 unit tests, and the complete
browser regression suite passed. The browser suite passed on retry after
a foreground-window assertion interrupted the first run.
Scope mute icon lookup to its tab and verify the same icon in both states.
Check that SVG resources are readable and produce nonempty drawings, and
construct fixture icon URIs correctly instead of casting argument arrays.
Exercise background-tab mute and unmute through the actual menu command,
including command parameters, shared-menu retargeting and tab isolation.
Wait for the source page title before evaluating its JavaScript.

Validation: VS MSBuild Debug/x64, all 295 unit tests, the complete browser
regression suite, and five mutation checks passed.
Remove the mute dependency property and toggle method from
ChromiumFlashBrowser. Keep mute state and switching in BrowserViewModel,
and move native audio control to ChromiumWebBrowserEx on the CEF UI thread.
Update toolbar and tab bindings and route page menu commands through the
owning view model.

Check view model state against native CEF audio in regression scenarios,
including background tabs, auxiliary windows, immediate repeated toggles,
and missing or closed browsers.

Validation: VS MSBuild Debug/x64, all 295 unit tests, and the complete
browser regression suite passed.
Remove mute from the toolbar and page context menu. Place tab mute last
behind a separator, and add mute to the actions menu after new window.
Bind Ctrl+M to mute and Alt+Home to show the main window, with matching
menu hints and state-dependent mute labels.

Update regression checks for menu order, selected-tab targeting, native
CEF mute state, page and address-bar keyboard focus, and removed entries.
Locate the actions mute item by command to support menu separators.

Validation: VS MSBuild Debug/x64, all 295 unit tests, and the complete
browser regression suite passed for the menu changes. After adding
Alt+Home, the solution build and focused shortcut regression passed.
Base the local mute icon style on the application SVG style so its Fill
follows PrimaryTextBrush instead of retaining the SVG's fixed dark color.
Keep the current tab icon spacing and visibility trigger formatting.

Mirror the application SVG style in the regression environment and check
both the icon fill and rendered drawing colors across dark/light switches.

Validation: VS MSBuild Debug/x64 and the focused theme regression passed.
The complete browser regression suite passed on retry after an initial
native Ctrl+M keyboard timeout.
Reuse existing SVG assets for new-tab, new-window, and mute actions in
the actions, tab, and page context menus. Switch mute and volume icons
with the current mute state while retaining theme-aware colors.

Use 12px new-tab icons and 14px menu window and audio icons, preserving
icon dimensions elsewhere in the interface.

Validation: VS MSBuild Debug/x64 and focused menu icon, resource, size,
and light/dark rendering checks passed before the final menu window
icon adjustment from 15px to 14px. The final adjustment passed XML and
diff validation.
Register mute-mode.svg as an application resource and use it for mute
actions in the actions and tab context menus. Retain the 14px menu size,
volume icon for unmute, and existing tab mute indicator.
Add the URI to the browser regression environment.

Validation: VS MSBuild Debug/x64 with isolated output and focused menu
resource, size, and light/dark rendering checks passed. The normal output
directory was locked by a running application instance.
Use the supplied star SVG for the actions-menu favorites entry.
Register the new star and zoom SVG assets and their application and
regression URI resources. Keep the zoom row without an icon.

Validation: Debug/x64 build and focused resource, theme, and zoom-control
checks passed during development. The final zoom-icon removal passed
XML and diff validation.
@Mzying2001
Mzying2001 merged commit 238d6c5 into master Nov 9, 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