22a70eb803
* chore: bump chromium in DEPS to 91.0.4464.0 * chore: rebuild chromium/dcheck.patch with import-patches -3 Mechanical only; no code changes * chore: remove content_browser_main_loop.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2725153 The function being patched (BrowserMainLoop::MainMessageLoopRun()) no longer exists. NB: if removing this introduces regressions the likely fix will be to add a similar patch for ShellBrowserMainParts::WillRunMainMessageLoop() which has similar code and was added at the same time this was removed. * chore: rebuild chromium/put_back_deleted_colors_for_autofill.patch with import-patches -3 Mechanical only; no code changes * chore: rebuild chromium/disable_color_correct_rendering.patch with import-patches -3 Mechanical only; no code changes * chore: rebuild chromium/eat_allow_disabling_blink_scheduler_throttling_per_renderview.patch with patch Mechanical only; no code changes * chore: rebuild chromium/gpu_notify_when_dxdiag_request_fails.patch with import-patches -3 Mechanical only; no code changes * chore: rebuild chromium/ui_gtk_public_header.patch manually no code changes * chore: rebuild chromium/web_contents.patch with import-patches -3 Mechanical only; no code changes * chore: remove v8/skip_global_registration_of_shared_arraybuffer_backing_stores.patch Refs: https://chromium-review.googlesource.com/c/v8/v8/+/2763874 This patch has been merged upstream * chore: export patches * chore: update add_trustedauthclient_to_urlloaderfactory.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2757969 Sync with removal of render_frame_id_ * chore: sync chromium/put_back_deleted_colors_for_autofill.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2785841 SkColorFromColorId() no longer takes theme, scheme args * chore: sync chromium/put_back_deleted_colors_for_autofill.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2772143 Change new calls to GetDarkSchemeColor to fit our patched call signature * chore: update add_trustedauthclient_to_urlloaderfactory.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2757969 Sync with removal of render_frame_id_ in our mojom * chore: update chromium/frame_host_manager.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2740008 UrlInfo ctor now takes UrlInfo::OriginIsolationRequest instead of a bool * chore: update chromium/revert_remove_contentrendererclient_shouldfork.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2755314 Upstream has removed `history_list_length_` which we were comparing to 0 to calculate our `is_initial_navigation` bool when calling ShouldFork(). ShouldFork() is ours and none of the code paths actually use that param, so this commit removes it altogether. * chore: update permissions_to_register Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2789074 Replace all uses of APIPermission::ID enum with Mojo type * refactor: update return type of PreMainMessageLoopRun() Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2725153 Used to return void; now returns an int errorcode. Note: 2725153 also has some nice doc updates about Browser's "stages" * refactor: sync ElectronBrowserMainParts to MainParts changes Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2725153 RunMainMessageLoopParts has been replaced with WillRunMainMessageLoop so `BrowserMainLoop::result_code_` is no longer available to us for our exit_code_ pointer. This variable held a dual role: (1) of course, hold the exit code, but also (2) was a nullptr before the message loop was ready, indicating to anyone calling SetExitCode() that we were still in startup and could just exit() without any extra steps. exit_code_ still fulfills these two roles but is now a base::Optional. * chore: update ElectronBrowserMainParts::PreDefaultMainMessageLoopRun Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2725153 BrowserMainParts::BrowsePreDefaultMainMesssageLoopRun() has been removed; move that work to the new WillRunMainMessageLoop(). * refactor: stop using CallbackList; it has been removed. Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2785973 * refactor: update use of threadpools. Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2773408 The upstream code is still in flux (e.g. reverts and re-lands) but the tl;dr for this commit is (1) include thread_pool.h if you're using it and (2) don't instantiate pools directly. * refactor: remove routing_id from CreateLoaderAndStart Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2762858 NB: One logic branch in ProxyingURLLoaderFactory::CreateLoaderAndStart calls std::make_unique<InProgressRequest>, which needs a routing_id. This PR uses the member field `routing_id_` since there's no longer one being passed into CreateLoaderAndStart. * refactor: sync to upstream ParittionOptions churn Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2771318 PartitionOptions' enums have changed. * refactor: update Manifest::Location usage Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2771320 tldr: s/Manifest::FOO/ManifestLocation::kFoo/ * chore: bump chromium in DEPS to 91.0.4465.0 * update patches * refactor: update extensions::Manifest to upstream Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2771320 - extensions::Manifest::COMPONENT + extensions::mojom::ManifestLocation::kExternalComponent * refactor: sync with upstream UrlInfo ctor changes Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2740008 UrlInfo ctor now takes UrlInfo::OriginIsolationRequest instead of a bool * chore: update invocation of convert_protocol_to_json.py Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2792623 python3 is being used in parts of the upstream build, but the copy of convert_protocol_to_json.py invoked in v8/third_party/inspector_protocol is not python3-friendly. Node has a py2+3-friendly version of it in its tools directory, so call it instead. * chore: use extensions::mojom::APIPermissionID Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2791122 tldr: - extensions::APIPermission::kFoo + extensions::mojom::APIPermissionID::kFoo * chore: Remove support for TLS1.0/1.1 in SSLVersionMin policy Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2765737 Remove TLS v1.0 & 1.1 from our SSLProtocolVersionFromString() function. This is the same change made upstream at https://chromium-review.googlesource.com/c/chromium/src/+/2765737/8/chrome/browser/ssl/ssl_config_service_manager_pref.cc * fixup! chore: update ElectronBrowserMainParts::PreDefaultMainMessageLoopRun * chore: Use IDType for permission change subscriptions. Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2791431 tldr: {Subscribe,Unsubscribe}PermissionStatusChange's tag type used to be an int; now it's the new SubscriptionId type (which is an IdType64). * chore: sync PowerMonitor code to upstream refactor Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2752635 tldr: PowerMonitor has been split into PowerStateObserver, PowerSuspendObserver, and PowerThermalObserver to reduce number of tasks posted to consumers who only need notifications for one of those things instead of all of them. * chore: use PartitionOptions's new Cookies field Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2771318 * Revert "refactor: remove routing_id from CreateLoaderAndStart" This reverts commit 8c9773b87a3c84f9073a47089eb2b6889d745245. 8c9773b was only a partial fix; reverting to start & try again. * update patches * chore: bump chromium in DEPS to 91.0.4466.0 * chore: update chromium/accelerator.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2795472 tldr: sync patch with upstream renamed variable & macro names. * chore: update chromium/gtk_visibility.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2796200 tldr: no code changes; just updating the diff to apply cleanly. note: ooh upstream Wayland hacking! * chore: update chromium/picture-in-picture.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2710023 tldr: no code changes; just updating the diff to apply cleanly. * chore: update chromium/worker_feat_add_hook_to_notify_script_ready.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2775573 tldr: no code changes; just updating the diff to apply cleanly. * chore: export_all_patches * chore: update chromium/feat_add_set_theme_source_to_allow_apps_to.patch Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2796511 tldr: NotifyObservers has been renamed to NotifyOnNativeThemeUpdated, so update the invocation in our patch. * chore: update ElectronBrowserClient w/upstream API Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2797454 tldr: GetDevToolsManagerDelegate() was returning an owned raw pointer. Replaced it with CreateDevToolsManagerDelegate() which uses unique_ptr<>. * chore: handle new content::PermissionType::FILE_HANDLING in toV8() Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2762201 `file-handling` string confirmed in https://chromium-review.googlesource.com/c/chromium/src/+/2762201/18/chrome/browser/ui/webui/settings/site_settings_helper.cc * refactor: remove routing_id from CreateLoaderAndStart pt 1 Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2762858 Part 1: the easiest ones * 2796724: Support Python3 https://chromium-review.googlesource.com/c/infra/luci/python-adb/+/2796724 * chore: bump chromium in DEPS to 91.0.4468.0 * 2668974: WebShare: Implement SharingServicePicker https://chromium-review.googlesource.com/c/chromium/src/+/2668974 * 2802766: Apply modernize-make-unique to media/ https://chromium-review.googlesource.com/c/chromium/src/+/2802766 * 2802823: Apply modernize-make-unique to gpu/ https://chromium-review.googlesource.com/c/chromium/src/+/2802823 * 2803041: Apply modernize-make-unique to remaining files https://chromium-review.googlesource.com/c/chromium/src/+/2803041 * 2798873: Convert GtkKeyBindingsHandler build checks to runtime checks https://chromium-review.googlesource.com/c/chromium/src/+/2798873 * 2733595: [ch-r] Parse ACCEPT_CH H2/3 frame and restart with new headers if needed https://chromium-review.googlesource.com/c/chromium/src/+/2733595 * chore: update patch indices * 2795107: Remove unused PermissionRequest IDs. https://chromium-review.googlesource.com/c/chromium/src/+/2795107 * chore: bump chromium in DEPS to 91.0.4469.0 * chore: fixup patch indices * chore: bump chromium in DEPS to 91.0.4469.5 * PiP 1.5: Add microphone, camera, and hang up buttons to the PiP window https://chromium-review.googlesource.com/c/chromium/src/+/2710023 * fixup! refactor: remove routing_id from CreateLoaderAndStart * refactor: use URLLoaderNetworkServiceObserver for auth requests from SimpleURLLoader * fixup! chore: fixup patch indices * 2724817: Expand scope of wasm-eval to all URLs https://chromium-review.googlesource.com/c/chromium/src/+/2724817 * Fixup patch after rebase * chore: bump chromium in DEPS to 91.0.4472.0 * 2797341: [ozone/x11] Enabled the global shortcut listener. https://chromium-review.googlesource.com/c/chromium/src/+/2797341 * 2805553: Reland Add GTK ColorMixers to ColorPipeline P1 https://chromium-review.googlesource.com/c/chromium/src/+/2805553 *2804366
: PiP 1.5: Label back to tab button with origin and center it https://chromium-review.googlesource.com/c/chromium/src/+/2804366 * 2784730: Fix crash on AX mode change in NativeViewHost without a Widget https://chromium-review.googlesource.com/c/chromium/src/+/2784730 * chore: update patch indices * 2810174: Add PdfAnnotationsEnabled policy. https://chromium-review.googlesource.com/c/chromium/src/+/2810174 * 2807829: Allow capturers to indicate if they want a WakeLock or not. https://chromium-review.googlesource.com/c/chromium/src/+/2807829 * chore: bump chromium in DEPS to 92.0.4473.0 * chore: bump chromium in DEPS to 92.0.4474.0 * chore: bump chromium in DEPS to 92.0.4475.0 * chore: update patches * chore: updates patches * chore: update is_media_key patch to handle new ozone impl Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2797341 * fix: ExecuteJavascript requests now need to be flagged as non-bf-aware Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2787195 * chore: icon_util_x11 is now icon_util_linux Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2791362 * build: update sysroots Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2628496 * build: fix missing symbols on linux build * use_ozone and use_x11 are not exclusive * new button view to build for pip Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2797341 Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2804366 * chore: fix broken gtk_util color patch * chore: remove patch conflict * build: update linux manifests * chore: build bttlb on all platforms for pip * chore: add thread_pool include for views delegate win * chore: fix lint * chore: add node patches for V8 changes * build: add missing base include on windows * fix: update frame host manager patch for new state transitions Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2714464 * chore: update windows zip manifests * chore: update mac zip manifests * chore: fix patch linting * refactor: implement missing URLLoaderNetworkServiceObserver methods It is against The Mojo Rules to leave hanging callbacks. These always have to be called. Refs:186528aab9
* spec: fix locale test on local linux * fix: pass the exit code correctly in new PreMainMessageLoopRun Refs:2622e91c44
* fix: ensure we early-exit when request_handler_ is not provided Refs:93077afbfb
* fix: strongly set result_code in the BrowserMainLoop * fix: invalid usage of non-targetted PostTask You must always either use a host threadpool or specify a target thread. In this case we did neither after this refactor. Refs:4e33ee0ad3
* chore: fix gn check * chore: remove stray .rej files in patch * chore: add mojo error code to url loader failure * build: ensure CI is truthy in arm test env * fix: handle windowCaptureMacV2 being enabled when fetching media source id Refs: https://chromium-review.googlesource.com/c/chromium/src/+/2709931 Co-authored-by: Charles Kerr <charles@charleskerr.com> Co-authored-by: Shelley Vohr <shelley.vohr@gmail.com> Co-authored-by: deepak1556 <hop2deep@gmail.com> Co-authored-by: Samuel Attard <samuel.r.attard@gmail.com> Co-authored-by: Samuel Attard <sattard@slack-corp.com>
109 lines
5.3 KiB
Diff
109 lines
5.3 KiB
Diff
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
|
|
From: Andy Locascio <andy@slack-corp.com>
|
|
Date: Tue, 18 Feb 2020 14:35:04 -0800
|
|
Subject: content: allow embedder to prevent locking scheme registry
|
|
|
|
The //content layer requires all schemes to be registered during startup,
|
|
because Add*Scheme aren't threadsafe. However, Electron exposes the option to
|
|
register additional schemes via JavaScript in the main process before the app
|
|
is ready, but after the //content layer has already locked the registry.
|
|
|
|
This allows embedders to optionally keep the scheme registry unlocked, and it
|
|
is their responsibility to ensure that it is not accessed in a way that would
|
|
cause potential thread-safety issues.
|
|
|
|
Previously upstreamed patch: https://chromium-review.googlesource.com/c/chromium/src/+/1637040
|
|
|
|
This change was lost during upstream refactor in
|
|
https://chromium-review.googlesource.com/c/chromium/src/+/1901591, we should try
|
|
re-submitting the patch.
|
|
|
|
diff --git a/content/app/content_main_runner_impl.cc b/content/app/content_main_runner_impl.cc
|
|
index 81e48e33cad43875634d4ab6a5625c16d4c37891..230f9348128d2630e80f1c0f346be23b15446dd7 100644
|
|
--- a/content/app/content_main_runner_impl.cc
|
|
+++ b/content/app/content_main_runner_impl.cc
|
|
@@ -738,7 +738,7 @@ int ContentMainRunnerImpl::Initialize(const ContentMainParams& params) {
|
|
}
|
|
#endif
|
|
|
|
- RegisterContentSchemes();
|
|
+ RegisterContentSchemes(delegate_->ShouldLockSchemeRegistry());
|
|
ContentClientInitializer::Set(process_type, delegate_);
|
|
|
|
#if !defined(OS_ANDROID)
|
|
diff --git a/content/common/url_schemes.cc b/content/common/url_schemes.cc
|
|
index dc37f121130d83e200d73dd1ad566847548ac0fd..63080c1bc486a488841fc5d2081f4d5d4a00bde3 100644
|
|
--- a/content/common/url_schemes.cc
|
|
+++ b/content/common/url_schemes.cc
|
|
@@ -49,7 +49,7 @@ std::vector<std::string>& GetMutableServiceWorkerSchemes() {
|
|
|
|
} // namespace
|
|
|
|
-void RegisterContentSchemes() {
|
|
+void RegisterContentSchemes(bool should_lock_registry) {
|
|
// On Android and in tests, schemes may have been registered already.
|
|
if (g_registered_url_schemes)
|
|
return;
|
|
@@ -106,7 +106,8 @@ void RegisterContentSchemes() {
|
|
// threadsafe so must be called when GURL isn't used on any other thread. This
|
|
// is really easy to mess up, so we say that all calls to Add*Scheme in Chrome
|
|
// must be inside this function.
|
|
- url::LockSchemeRegistries();
|
|
+ if (should_lock_registry)
|
|
+ url::LockSchemeRegistries();
|
|
|
|
// Combine the default savable schemes with the additional ones given.
|
|
GetMutableSavableSchemes().assign(std::begin(kDefaultSavableSchemes),
|
|
diff --git a/content/common/url_schemes.h b/content/common/url_schemes.h
|
|
index 3038f9d25798f36811b6398f8cc0e7d83ecc41b0..68189c36c47ef85b345b0ccc40c456f889977bee 100644
|
|
--- a/content/common/url_schemes.h
|
|
+++ b/content/common/url_schemes.h
|
|
@@ -16,7 +16,7 @@ namespace content {
|
|
// parsed as "standard" or "referrer" with the src/url/ library, then locks the
|
|
// sets of schemes down. The embedder can add additional schemes by
|
|
// overriding the ContentClient::AddAdditionalSchemes method.
|
|
-CONTENT_EXPORT void RegisterContentSchemes();
|
|
+CONTENT_EXPORT void RegisterContentSchemes(bool should_lock_registry = true);
|
|
|
|
// Re-initializes schemes for tests.
|
|
CONTENT_EXPORT void ReRegisterContentSchemesForTests();
|
|
diff --git a/content/public/app/content_main_delegate.cc b/content/public/app/content_main_delegate.cc
|
|
index 17d256b852b6e7615cf4e52e4a2d7fa326a4df44..bd0512902df4484e8714cda9ab1352b1c672a1cd 100644
|
|
--- a/content/public/app/content_main_delegate.cc
|
|
+++ b/content/public/app/content_main_delegate.cc
|
|
@@ -42,6 +42,10 @@ bool ContentMainDelegate::ShouldHandleConsoleControlEvents() {
|
|
}
|
|
#endif
|
|
|
|
+bool ContentMainDelegate::ShouldLockSchemeRegistry() {
|
|
+ return true;
|
|
+}
|
|
+
|
|
bool ContentMainDelegate::ShouldCreateFeatureList() {
|
|
return true;
|
|
}
|
|
diff --git a/content/public/app/content_main_delegate.h b/content/public/app/content_main_delegate.h
|
|
index 4162083f789b6cf5e29e52e4bd498c377cab1a20..d6010bc0e4dba560d10c3ed4729d77b5a60eddcd 100644
|
|
--- a/content/public/app/content_main_delegate.h
|
|
+++ b/content/public/app/content_main_delegate.h
|
|
@@ -65,6 +65,20 @@ class CONTENT_EXPORT ContentMainDelegate {
|
|
// returning initialization error code. Default behavior is CHECK(false).
|
|
virtual int TerminateForFatalInitializationError();
|
|
|
|
+ // Allows the embedder to prevent locking the scheme registry. The scheme
|
|
+ // registry is the list of URL schemes we recognize, with some additional
|
|
+ // information about each scheme such as whether it expects a host. The
|
|
+ // scheme registry is not thread-safe, so by default it is locked before any
|
|
+ // threads are created to ensure single-threaded access. An embedder can
|
|
+ // override this to prevent the scheme registry from being locked during
|
|
+ // startup, but if they do so then they are responsible for making sure that
|
|
+ // the registry is only accessed in a thread-safe way, and for calling
|
|
+ // url::LockSchemeRegistries() when initialization is complete. If possible,
|
|
+ // prefer registering additional schemes through
|
|
+ // ContentClient::AddAdditionalSchemes over preventing the scheme registry
|
|
+ // from being locked.
|
|
+ virtual bool ShouldLockSchemeRegistry();
|
|
+
|
|
// Allows the embedder to perform platform-specific initialization before
|
|
// creating the main message loop.
|
|
virtual void PreCreateMainMessageLoop() {}
|