![electron-roller[bot]](/assets/img/avatar_default.png)
* chore: bump chromium in DEPS to 126.0.6470.0 * 5492605: Migrate TODOs referencing old crbug IDs to the new issue tracker IDs |5492605
* 5513277: Move subresource-filter-ruleset to GCS |5513277
* 5512656: Remove CustomizeChromeSupportsChromeRefresh2023 |5512656
* 5516009: Accept mouse events in inactive window for Top Chrome WebUIs |5516009
* 5376861: Change references to RWHVB in RWHIER and RenderWidgetTargeter to RWHVI. |5376861
* 5490530: Use partition_alloc PA_BUILDFLAG(...) outside PA. #cleanup |5490530
* 5296870: network: Allow trusted loaders to learn the sent request cookies. |5296870
* 5453438: Delegate delegated ink trails to RWHI from RWHIER. |5453438
* chore: update patches * chore: bump chromium in DEPS to 126.0.6472.0 * chore: bump chromium in DEPS to 126.0.6474.0 * chore: update patches * chore: bump chromium in DEPS to 126.0.6476.0 * chore: bump chromium in DEPS to 126.0.6478.0 * chore: bump chromium in DEPS to 126.0.6478.3 * chore: bump chromium in DEPS to 126.0.6478.8 * update patches * only disable enterprise_cloud_content_analysis * 5403888: [api] support v8::Data in v8::TracedReference and v8::EmbedderGraph5403888
* chore: bump chromium in DEPS to 127.0.6484.0 * chore: bump chromium in DEPS to 127.0.6485.0 * 5539004: Use NOTREACHED_IN_MIGRATION() in remaining chrome/ |5539004
* src: cast to v8::Value before using v8::EmbedderGraph::V8Node | https://github.com/nodejs/node/pull/52638/files * chore: update patches * chore: update v8 patches * chore: bump chromium in DEPS to 127.0.6486.0 * chore: bump chromium in DEPS to 127.0.6488.0 * chore: bump chromium in DEPS to 127.0.6490.0 * chore: bump chromium in DEPS to 127.0.6492.0 * chore: update patches For some reason, `feat_expose_raw_response_headers_from_urlloader.patch` got messed up in an earlier commit. * chore: update patches printing.patch was updated due to5535938
* 5527572: Move Connectors prefs files to components/enterprise/connectors/5527572
* chore: bump chromium in DEPS to 127.0.6494.0 * chore: bump chromium in DEPS to 127.0.6495.0 * chore: bump chromium in DEPS to 127.0.6496.0 * 5465511: [api] Mark v8::ObjectTemplate::SetAccessor(..) for deprecation5465511
* chore: revert v8 deprecation See patch message for more details.5526611
* chore: update patches * 5538771: Remove srcdoc else-if block in CalculateOrigin()5538771
* 5522321: [devtools] Support saving base64 encoded files via host bindings5522321
* 5376861: Change references to RWHVB in RWHIER and RenderWidgetTargeter to RWHVI.5376861
* 5530163: [media] Use VideoFrame::Plane typed enum instead of nameless enum5530163
* 5463431: iwa: Only create IsolatedWebAppURLLoaderFactory for subresources in IWAs5463431
* fixup! 5465511: [api] Mark v8::ObjectTemplate::SetAccessor(..) for deprecation5465511
* 5512176: Remove OnEnvironmentEstimationComplete()5512176
* 5528282: Move Web Speech API .mojom files to //media/mojo/mojom5528282
* 5513740: Reland "[Extensions] Restructure extensions::ProcessMap"5513740
* 5483406: [PEPC] Make PEPC permission subscription take into account device status5483406
* 5526034: [DoH] Remove kDnsOverHttps feature flag5526034
The title is a bit misleading. They removed handling for the feature flag and generally intend to remove it but haven't yet. I only changed our code to address the flag that was removed. A quick search on GitHub for `DnsOverHttpsFallback` yielded a few results, but they were all C++ chromium code or patches, 0 app code or discussion results. Since I couldn't find any evidence of this flag being used in developer applications, I've chosen to exclude this change from the breaking changes docs. * chore: revert v8 removal5497515
See patch message for more details. * chore: cherry-pick Node.js patch for V8 API removal fix Node.js PR: https://github.com/nodejs/node/pull/52996 V8 API Removal CL:5539888
See the patch description for more details. * 5492183: Extensions: CodeHealth: Give enums some class5492183
* fixup! 5528282: Move Web Speech API .mojom files to //media/mojo/mojom5528282
* 5514687: Reland "Add a secret handshake to the base::Feature constructor"5514687
* fixup! 5530163: [media] Use VideoFrame::Plane typed enum instead of nameless enum5530163
* 5466238: PDF Viewer: add metrics to record if PDF is opened with a11y5466238
* 5502081: Migrate OnDisplayRemoved to OnDisplaysRemoved5502081
* 5539888: [api] Remove several APIs deprecated in version 12.65539888
This commit essentially only removes the `only_terminate_in_safe_scope` isolate creation parameter. This undoes some work that was originally done in #35766. * 5498236: Make browser_tests force full async initialization for OSCrypt Async5498236
* fixup! 5528282: Move Web Speech API .mojom files to //media/mojo/mojom5528282
* 5545807: Migrate most remaining NOTREACHED()5545807
I took a systematic approach to modifying all of our uses of `NOTREACHED` that were causing errors: * If there was a `return` or `break` (etc.) immediately after `NOTREACHED`, I removed the control flow instruction and left the `NOTREACHED` unmodified * All other instances were migrated to `NOTREACHED_IN_MIGRATION` We should revisit pretty much all usage of `NOTREACHED` as an upgrade follow-up item. * fixup! 5526034: [DoH] Remove kDnsOverHttps feature flag5526034
Turns out the feature flags were removed in the `.cc` file, but not the `.h` feature list file. This means that the feature flags are pretty much officially gone. (The leftover symbols in the header are likely an oversight from what I can gather.) We may potentially decide to put this in the breaking changes doc if we decide this feature flag is important enough to highlight. * chore: bump chromium in DEPS to 127.0.6498.3 * chore: bump chromium in DEPS to 127.0.6500.0 * chore: bump chromium in DEPS to 127.0.6502.0 * chore: bump chromium in DEPS to 127.0.6504.0 * chore: bump chromium in DEPS to 127.0.6505.0 * chore: bump chromium in DEPS to 127.0.6508.0 * build: use Sha256Sum in script/sysroots.json Xref:5506275
* chore: update chore_add_electron_deps_to_gitignores.patch Xref: no manual changes; patch applied with fuzz 2 * chore: update feat_allow_code_cache_in_custom_schemes.patch Xref: no manual changes; patch applied with fuzz 1 * chore: e patches all * fixup! build: use Sha256Sum in script/sysroots.json `sync` succeeds now * chore: replace absl::optional with std::optional Xref:5253843
* chore: update CalculatePreferredSize() to new upstream semantics Xref:5459174
Xref:5541220
Xref:5514708
Xref:5504212
Xref: https://chromium-review.googlesource.com/516542 * chore: replace absl::optional with std::optional Xref:5296147
* chore: add kPip to enumeration as a no-op5546257
* [Autofill] Remove RenderFrame::ElementBoundsInWindow() Xref:5553982
* chore: fix feat_add_streaming-protocol_registry_to_multibuffer_data_source.patch need new header to pick up definition of BLINK_PLATFORM_EXPORT macro Xref:5463143
* chore: bump chromium in DEPS to 127.0.6510.0 * chore: update patches * chore: fix include path for native_web_keyboard_event.h Xref:5541976
* chore: add currently-unused should_include_device_status arg to GetPermissionStatusForCurrentDocument() Xref:5545382
* chore: bump chromium in DEPS to 127.0.6512.0 * chore: update mas_avoid_private_macos_api_usage.patch.patch No manual changes; patch applied with fuzz 1 * chore: update feat_add_streaming-protocol_registry_to_multibuffer_data_source.patch No manual changes; patch applied with fuzz 1 * chore: update webview_fullscreen.patch No manual changes; patch applied with fuzz 1 * chore=: remove cherry-pick-22db6918bac9.patch already present upstream * chore: remove nonexistent patchfiles from .patches * chore: remove cherry-pick-3e037e195e50.patch no longer needed; merged upstream * Update namespace for files moved to //components/input Xref:5563251
* Require client for InitParams to always specify an ownership mode. Xref:5532482
Xref:5578714
* chore: e patches all * fixup! Update namespace for files moved to //components/input * chore: remove profile_keyed_service_factory, profile_selections from chromium_src already being linked in via chrome browser for printing * chore: bump chromium in DEPS to 127.0.6515.0 * chore: bump chromium in DEPS to 127.0.6516.0 * chore: update render_widget_host_view_base.patch Xref:5547803
patch applied manually due to simple upstream shear * chore: update feat_allow_code_cache_in_custom_schemes.patch No manual changes; patch applied with fuzz 1 * chore: e patches all * Pull RWHIER and RWT to //content/common/input. Xref:5397681
* chore: bump chromium in DEPS to 127.0.6517.0 * chore: update patches * fixup: Update namespace for files moved to //components/input * Remove 0-arg (default) constructor for views::Widget::InitParams.5578714
* fixup: only disable enterprise_cloud_content_analysis The original commita5480accc2
, was due to this CL 5527572: Move Connectors prefs files to components/enterprise/connectors/ |5527572
* chore: bump chromium in DEPS to 127.0.6519.0 * chore: update patches * src: do not use deprecated V8 API https://github.com/nodejs/node/pull/53084 * src: remove dependency on wrapper-descriptor-based cpp heap https://github.com/nodejs/node/pull/53086 * 5344413: [DevTools] Add `getHostConfig` UI binding for sending status of `base::Features` to DevTools5344413
* 5585788: Extensions: ManifestHandler: Separate Registry like ExtensionRegistry5585788
* chore: update filenames.libcxx.gni * 5506857: Reland "Migrate clang-format to gcs first class deps"5506857
* fixup: 5539888: [api] Remove several APIs deprecated in version 12.6 * fixup: 5506857: Reland Migrate clang-format to gcs first class deps * chore: bump chromium in DEPS to 127.0.6521.0 * chore: update patches * spec: update navigator.keyboard should lock the keyboard * Block or allow all MIDI using the existing SysEx permission Refs5154368
Refs5499157
* spec: update test/parallel/test-v8-stats * views: remove CalculatePreferredSize() Refs5504212
* chore: update patches after rebase * 5560288: Re-enable ChromeOS XNNPack on Intel only5560288
* chore: add nan patches for v8 changes Refs 5539888: [api] Remove several APIs deprecated in version 12.6 |5539888
and 5539852: [heap][api] Remove deprecated v8::Isolate::IdleNotificationDeadline |5539852
* 5573603: Modularize //chrome/browser/themes5573603
* 5539888: [api] Remove several APIs deprecated in version 12.65539888
* chore: update patches * test: fixup navigator.keyboard.lock on Windows * chore: remove unneeded profile target --------- Co-authored-by: electron-roller[bot] <84116207+electron-roller[bot]@users.noreply.github.com> Co-authored-by: Keeley Hammond <khammond@slack-corp.com> Co-authored-by: VerteDinde <vertedinde@electronjs.org> Co-authored-by: Jeremy Rose <nornagon@nornagon.net> Co-authored-by: clavin <clavin@electronjs.org> Co-authored-by: Charles Kerr <charles@charleskerr.com> Co-authored-by: PatchUp <73610968+patchup[bot]@users.noreply.github.com> Co-authored-by: John Kleinschmidt <jkleinsc@electronjs.org> Co-authored-by: deepak1556 <hop2deep@gmail.com>
243 lines
8.2 KiB
C++
243 lines
8.2 KiB
C++
// Copyright (c) 2015 GitHub, Inc.
|
|
// Use of this source code is governed by the MIT license that can be
|
|
// found in the LICENSE file.
|
|
|
|
// Most code came from: chrome/browser/chrome_browser_main_posix.cc.
|
|
|
|
#include "base/notreached.h"
|
|
#include "shell/browser/electron_browser_main_parts.h"
|
|
|
|
#include <errno.h>
|
|
#include <limits.h>
|
|
#include <pthread.h>
|
|
#include <signal.h>
|
|
#include <sys/resource.h>
|
|
#include <unistd.h>
|
|
|
|
#include "base/debug/leak_annotations.h"
|
|
#include "base/posix/eintr_wrapper.h"
|
|
#include "base/threading/platform_thread.h"
|
|
#include "content/public/browser/browser_task_traits.h"
|
|
#include "content/public/browser/browser_thread.h"
|
|
#include "shell/browser/browser.h"
|
|
|
|
namespace electron {
|
|
|
|
namespace {
|
|
|
|
// See comment in |PreEarlyInitialization()|, where sigaction is called.
|
|
void SIGCHLDHandler(int signal) {}
|
|
|
|
// The OSX fork() implementation can crash in the child process before
|
|
// fork() returns. In that case, the shutdown pipe will still be
|
|
// shared with the parent process. To prevent child crashes from
|
|
// causing parent shutdowns, |g_pipe_pid| is the pid for the process
|
|
// which registered |g_shutdown_pipe_write_fd|.
|
|
// See <http://crbug.com/175341>.
|
|
pid_t g_pipe_pid = -1;
|
|
int g_shutdown_pipe_write_fd = -1;
|
|
int g_shutdown_pipe_read_fd = -1;
|
|
|
|
// Common code between SIG{HUP, INT, TERM}Handler.
|
|
void GracefulShutdownHandler(int signal) {
|
|
// Reinstall the default handler. We had one shot at graceful shutdown.
|
|
struct sigaction action;
|
|
memset(&action, 0, sizeof(action));
|
|
action.sa_handler = SIG_DFL;
|
|
RAW_CHECK(sigaction(signal, &action, nullptr) == 0);
|
|
|
|
RAW_CHECK(g_pipe_pid == getpid());
|
|
RAW_CHECK(g_shutdown_pipe_write_fd != -1);
|
|
RAW_CHECK(g_shutdown_pipe_read_fd != -1);
|
|
size_t bytes_written = 0;
|
|
do {
|
|
int rv = HANDLE_EINTR(
|
|
write(g_shutdown_pipe_write_fd,
|
|
reinterpret_cast<const char*>(&signal) + bytes_written,
|
|
sizeof(signal) - bytes_written));
|
|
RAW_CHECK(rv >= 0);
|
|
bytes_written += rv;
|
|
} while (bytes_written < sizeof(signal));
|
|
}
|
|
|
|
// See comment in |PostCreateMainMessageLoop()|, where sigaction is called.
|
|
void SIGHUPHandler(int signal) {
|
|
RAW_CHECK(signal == SIGHUP);
|
|
GracefulShutdownHandler(signal);
|
|
}
|
|
|
|
// See comment in |PostCreateMainMessageLoop()|, where sigaction is called.
|
|
void SIGINTHandler(int signal) {
|
|
RAW_CHECK(signal == SIGINT);
|
|
GracefulShutdownHandler(signal);
|
|
}
|
|
|
|
// See comment in |PostCreateMainMessageLoop()|, where sigaction is called.
|
|
void SIGTERMHandler(int signal) {
|
|
RAW_CHECK(signal == SIGTERM);
|
|
GracefulShutdownHandler(signal);
|
|
}
|
|
|
|
class ShutdownDetector : public base::PlatformThread::Delegate {
|
|
public:
|
|
explicit ShutdownDetector(
|
|
int shutdown_fd,
|
|
base::OnceCallback<void()> shutdown_callback,
|
|
const scoped_refptr<base::SingleThreadTaskRunner>& task_runner);
|
|
|
|
// disable copy
|
|
ShutdownDetector(const ShutdownDetector&) = delete;
|
|
ShutdownDetector& operator=(const ShutdownDetector&) = delete;
|
|
|
|
// base::PlatformThread::Delegate:
|
|
void ThreadMain() override;
|
|
|
|
private:
|
|
const int shutdown_fd_;
|
|
base::OnceCallback<void()> shutdown_callback_;
|
|
const scoped_refptr<base::SingleThreadTaskRunner> task_runner_;
|
|
};
|
|
|
|
ShutdownDetector::ShutdownDetector(
|
|
int shutdown_fd,
|
|
base::OnceCallback<void()> shutdown_callback,
|
|
const scoped_refptr<base::SingleThreadTaskRunner>& task_runner)
|
|
: shutdown_fd_(shutdown_fd),
|
|
shutdown_callback_(std::move(shutdown_callback)),
|
|
task_runner_(task_runner) {
|
|
CHECK_NE(shutdown_fd_, -1);
|
|
CHECK(!shutdown_callback_.is_null());
|
|
CHECK(task_runner_);
|
|
}
|
|
|
|
// These functions are used to help us diagnose crash dumps that happen
|
|
// during the shutdown process.
|
|
NOINLINE void ShutdownFDReadError() {
|
|
// Ensure function isn't optimized away.
|
|
asm("");
|
|
sleep(UINT_MAX);
|
|
}
|
|
|
|
NOINLINE void ShutdownFDClosedError() {
|
|
// Ensure function isn't optimized away.
|
|
asm("");
|
|
sleep(UINT_MAX);
|
|
}
|
|
|
|
NOINLINE void ExitPosted() {
|
|
// Ensure function isn't optimized away.
|
|
asm("");
|
|
sleep(UINT_MAX);
|
|
}
|
|
|
|
void ShutdownDetector::ThreadMain() {
|
|
base::PlatformThread::SetName("CrShutdownDetector");
|
|
|
|
int signal;
|
|
size_t bytes_read = 0;
|
|
do {
|
|
const ssize_t ret = HANDLE_EINTR(
|
|
read(shutdown_fd_, reinterpret_cast<char*>(&signal) + bytes_read,
|
|
sizeof(signal) - bytes_read));
|
|
if (ret < 0) {
|
|
NOTREACHED_IN_MIGRATION() << "Unexpected error: " << strerror(errno);
|
|
ShutdownFDReadError();
|
|
break;
|
|
} else if (ret == 0) {
|
|
NOTREACHED_IN_MIGRATION() << "Unexpected closure of shutdown pipe.";
|
|
ShutdownFDClosedError();
|
|
break;
|
|
}
|
|
bytes_read += ret;
|
|
} while (bytes_read < sizeof(signal));
|
|
VLOG(1) << "Handling shutdown for signal " << signal << ".";
|
|
|
|
if (!task_runner_->PostTask(FROM_HERE,
|
|
base::BindOnce(std::move(shutdown_callback_)))) {
|
|
// Without a valid task runner to post the exit task to, there aren't many
|
|
// options. Raise the signal again. The default handler will pick it up
|
|
// and cause an ungraceful exit.
|
|
RAW_LOG(WARNING, "No valid task runner, exiting ungracefully.");
|
|
kill(getpid(), signal);
|
|
|
|
// The signal may be handled on another thread. Give that a chance to
|
|
// happen.
|
|
sleep(3);
|
|
|
|
// We really should be dead by now. For whatever reason, we're not. Exit
|
|
// immediately, with the exit status set to the signal number with bit 8
|
|
// set. On the systems that we care about, this exit status is what is
|
|
// normally used to indicate an exit by this signal's default handler.
|
|
// This mechanism isn't a de jure standard, but even in the worst case, it
|
|
// should at least result in an immediate exit.
|
|
RAW_LOG(WARNING, "Still here, exiting really ungracefully.");
|
|
_exit(signal | (1 << 7));
|
|
}
|
|
ExitPosted();
|
|
}
|
|
|
|
} // namespace
|
|
|
|
void ElectronBrowserMainParts::HandleSIGCHLD() {
|
|
// We need to accept SIGCHLD, even though our handler is a no-op because
|
|
// otherwise we cannot wait on children. (According to POSIX 2001.)
|
|
struct sigaction action;
|
|
memset(&action, 0, sizeof(action));
|
|
action.sa_handler = SIGCHLDHandler;
|
|
CHECK_EQ(sigaction(SIGCHLD, &action, nullptr), 0);
|
|
}
|
|
|
|
void ElectronBrowserMainParts::InstallShutdownSignalHandlers(
|
|
base::OnceCallback<void()> shutdown_callback,
|
|
const scoped_refptr<base::SingleThreadTaskRunner>& task_runner) {
|
|
int pipefd[2];
|
|
int ret = pipe(pipefd);
|
|
if (ret < 0) {
|
|
PLOG(DFATAL) << "Failed to create pipe";
|
|
return;
|
|
}
|
|
g_pipe_pid = getpid();
|
|
g_shutdown_pipe_read_fd = pipefd[0];
|
|
g_shutdown_pipe_write_fd = pipefd[1];
|
|
#if !defined(ADDRESS_SANITIZER)
|
|
const size_t kShutdownDetectorThreadStackSize = PTHREAD_STACK_MIN * 2;
|
|
#else
|
|
// ASan instrumentation bloats the stack frames, so we need to increase the
|
|
// stack size to avoid hitting the guard page.
|
|
const size_t kShutdownDetectorThreadStackSize = PTHREAD_STACK_MIN * 4;
|
|
#endif
|
|
ShutdownDetector* detector = new ShutdownDetector(
|
|
g_shutdown_pipe_read_fd, std::move(shutdown_callback), task_runner);
|
|
|
|
// PlatformThread does not delete its delegate.
|
|
ANNOTATE_LEAKING_OBJECT_PTR(detector);
|
|
if (!base::PlatformThread::CreateNonJoinable(kShutdownDetectorThreadStackSize,
|
|
detector)) {
|
|
LOG(DFATAL) << "Failed to create shutdown detector task.";
|
|
}
|
|
// Setup signal handlers for shutdown AFTER shutdown pipe is setup because
|
|
// it may be called right away after handler is set.
|
|
|
|
// If adding to this list of signal handlers, note the new signal probably
|
|
// needs to be reset in child processes. See
|
|
// base/process_util_posix.cc:LaunchProcess.
|
|
|
|
// We need to handle SIGTERM, because that is how many POSIX-based distros
|
|
// ask processes to quit gracefully at shutdown time.
|
|
struct sigaction action;
|
|
memset(&action, 0, sizeof(action));
|
|
action.sa_handler = SIGTERMHandler;
|
|
CHECK_EQ(sigaction(SIGTERM, &action, nullptr), 0);
|
|
|
|
// Also handle SIGINT - when the user terminates the browser via Ctrl+C. If
|
|
// the browser process is being debugged, GDB will catch the SIGINT first.
|
|
action.sa_handler = SIGINTHandler;
|
|
CHECK_EQ(sigaction(SIGINT, &action, nullptr), 0);
|
|
|
|
// And SIGHUP, for when the terminal disappears. On shutdown, many Linux
|
|
// distros send SIGHUP, SIGTERM, and then SIGKILL.
|
|
action.sa_handler = SIGHUPHandler;
|
|
CHECK_EQ(sigaction(SIGHUP, &action, nullptr), 0);
|
|
}
|
|
|
|
} // namespace electron
|