[browser] native - no sockets, no threads, no blocking #123962

pavelsavara · 2026-02-03T20:38:00Z

dotnet-policy-service · 2026-02-03T20:39:04Z

Tagging subscribers to this area: @karelz, @dotnet/ncl
See info in area-owners.md if you want to be subscribed.

Copilot

Pull request overview

This pull request adds support for the browser (WASM) target by providing stub implementations for features not available in the browser environment, specifically: no sockets, no threads, and no blocking operations.

Changes:

Added browser-specific stub implementations for networking, interface addresses, network statistics, and console APIs
Updated error messages in WASM-specific files from "WASI" to "WASM" for clarity
Modified CoreCLR PAL to prevent blocking operations and thread suspension in single-threaded WASM
Updated build configuration to disable socket support for browser targets and include browser-specific source files
Added perftracing configuration for browser builds

Reviewed changes

Copilot reviewed 20 out of 21 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
src/native/libs/System.Native/pal_signal_wasm.c	Updated error messages from "WASI" to "WASM"
src/native/libs/System.Native/pal_networkstatistics_wasm.c	New file with stub implementations returning "not supported" for network statistics APIs
src/native/libs/System.Native/pal_networking_browser.c	New file with comprehensive stub implementations for all networking APIs, returning ENOTSUP
src/native/libs/System.Native/pal_io.c	Added early return for pipe() on WASM targets with ENOTSUP
src/native/libs/System.Native/pal_interfaceaddresses_browser.c	New file with stub implementations for interface enumeration APIs
src/native/libs/System.Native/pal_dynamicload_wasm.c	Updated error messages from "WASI" to "WASM"
src/native/libs/System.Native/pal_console_wasm.c	New console implementation for WASM with some working features (stdin/isatty) and stubs for unsupported features
src/native/libs/System.Native/CMakeLists.txt	Updated build configuration to differentiate between browser and WASI targets, using browser-specific source files
src/native/eventpipe/ds-ipc-pal-websocket.c	Added string duplication helper functions for eventpipe
src/mono/mono/profiler/log.c	Changed sys/socket.h include guard from HOST_WIN32 to HAVE_SYS_SOCKET_H
src/mono/mono/profiler/helper.c	Changed sys/socket.h include guard from HOST_WIN32 to HAVE_SYS_SOCKET_H
src/mono/mono/profiler/CMakeLists.txt	Disabled AOT profiler for browser targets
src/mono/mono/mini/cfgdump.c	Added guard for sys/socket.h include
src/mono/browser/runtime/dotnet.d.ts	Formatting fix: removed space before parentheses in function signatures
src/coreclr/pal/src/thread/threadsusp.cpp	Added guard to prevent thread suspension in single-threaded WASM
src/coreclr/pal/src/thread/thread.cpp	Added guard to prevent CREATE_SUSPENDED flag in single-threaded WASM
src/coreclr/pal/src/synchmgr/wait.cpp	Added guards to prevent blocking waits and made Sleep return immediately in WASM
src/coreclr/pal/src/synchmgr/synchmanager.cpp	Added assertion to prevent waiting on non-signaled objects in WASM
src/coreclr/pal/src/debug/debug.cpp	Added WASM-specific implementation of PAL_ProbeMemory using emscripten heap size
eng/native/tryrun.browser.cmake	Disabled HAVE_SYS_SOCKET_H for browser targets
eng/native.wasm.targets	Added perftracing configuration flags for browser builds

Copilot · 2026-02-03T20:47:01Z

src/coreclr/pal/src/debug/debug.cpp

+#if defined(__EMSCRIPTEN__)
+    if ((uintptr_t)((PBYTE)pBuffer + cbBuffer) <= emscripten_get_heap_size())
+    {
+        return TRUE;
+    }
+    return FALSE;


The fWriteAccess parameter is not used in the Emscripten implementation. This means the function cannot distinguish between read-only and writable memory validation. The function should either use this parameter to check write permissions (though this may be difficult in WASM), or at minimum document this limitation, or mark the parameter as unused with (void)fWriteAccess.

Copilot · 2026-02-03T20:47:01Z

src/native/libs/System.Native/pal_console_wasm.c

+    assert(buffer != NULL || bufferSize == 0);
+    assert(bufferSize >= 0);
+
+     if (bufferSize < 0)


There is an extra leading space before the 'if' statement. This appears inconsistent with the indentation in the existing code. The line should start with 4 spaces for proper indentation alignment.

Suggested change

if (bufferSize < 0)

if (bufferSize < 0)

Copilot · 2026-02-03T20:47:02Z

src/native/libs/System.Native/pal_console_wasm.c

+
+int32_t SystemNative_GetSignalForBreak(void)
+{
+    return false;


The function returns 'false' which is a boolean literal, but the return type is int32_t. This should return 0 instead of false for type consistency. The standard implementation in pal_console.c returns the value of g_signalForBreak (which is a bool but implicitly converted to int), while the WASI implementation also incorrectly returns 'false'. For clarity and correctness, this should return 0.

Suggested change

return false;

return 0;

Copilot · 2026-02-03T20:47:02Z

src/native/libs/System.Native/pal_console_wasm.c

+int32_t SystemNative_InitializeTerminalAndSignalHandling(void)
+{
+    return true;


The function returns 'true' which is a boolean literal, but the return type is int32_t. This should return 1 instead of true for type consistency.

Copilot · 2026-02-03T20:47:02Z

src/mono/mono/mini/cfgdump.c

 #include <netinet/in.h>
 #include <netdb.h>


This file includes netinet/in.h and netdb.h without header guards, but the PR disables HAVE_SYS_SOCKET_H for browser in tryrun.browser.cmake. The file uses socket APIs (socket, connect, sockaddr_in, etc.) throughout and is conditionally compiled based on !HOST_WIN32. Since browser doesn't support sockets, this file will fail to compile for browser targets. The compile-time guard on line 11 should likely also exclude HOST_BROWSER, or these header includes need proper HAVE_* guards.

Copilot · 2026-02-03T20:47:02Z

src/coreclr/pal/src/debug/debug.cpp

+#if defined(__EMSCRIPTEN__)
+    if ((uintptr_t)((PBYTE)pBuffer + cbBuffer) <= emscripten_get_heap_size())


The addition (PBYTE)pBuffer + cbBuffer could potentially overflow if pBuffer is near the end of the address space. Consider checking for overflow before performing the comparison, for example by checking if cbBuffer > (SIZE_MAX - (uintptr_t)pBuffer) first.

jkotas · 2026-02-03T21:40:06Z

src/coreclr/pal/src/thread/thread.cpp

    if (pThread->GetCreateSuspended())
    {
+#ifdef __wasm__
+        // CREATE_SUSPENDED is not supported in single-threaded WASM because it uses


Creating threads is not supported in single threaded wasm. Should the scope of these ifdefs be much larger?

jkotas · 2026-02-03T21:41:59Z

src/coreclr/pal/src/synchmgr/synchmanager.cpp

            goto TNW_exit;
        }

+#ifdef __wasm__


Should this be FEATURE_SINGLE_THREADED?

jkotas · 2026-02-03T21:45:28Z

src/coreclr/pal/src/debug/debug.cpp

@@ -751,6 +755,13 @@ PAL_ProbeMemory(
    DWORD cbBuffer,


This whole method should be ifdefed out for Wasm.

pavelsavara · 2026-02-04T00:10:11Z

Whole this PR makes emscripten emulator smaller only by 20KB of JavaScript - from 154KB to 134KB.
And only 4KB of .wasm with -O3. Those are CoreCLR numbers, it will have similar impact for Mono but I have not measured it yet.

I hoped for more, I'm not sure I want to merge this...

pavelsavara · 2026-02-04T00:47:28Z

These are JS emulators of mmap that are slow and terrible, we should not use them in CoreCLR

adding _mmap_js (referenced by root reference (e.g. compiled C/C++ code))
adding _msync_js (referenced by root reference (e.g. compiled C/C++ code))
adding _munmap_js (referenced by root reference (e.g. compiled C/C++ code))

The rest of the JS support is OK. See https://gist.github.com/pavelsavara/f11dd1078548a16fc923de444f1c9d9b

All the symbols in the wasm files are https://gist.github.com/pavelsavara/b7653066a83554168739558630d34847

Among them these caught my eye

GetNativeSwiftPhysicalLowering

GetThreadStaticsByIndex
SwitchStackAndExecuteHandler
Thread::HandleThreadAbort
ThreadNative_Abort
ThreadNative_ResetAbort
ThreadNative_Start
ThreadSuspend::LockThreadStore

WKS::gc_heap and SVR::gc_heap

COMDouble
COMSingle
COMDelegate
OleVariant
RealCOMPlusThrow

jkotas · 2026-02-04T00:57:04Z

These are JS emulators of mmap that are slow and terrible, we should not use them in CoreCLR

#117813 is the main offender.

pavelsavara added 2 commits February 3, 2026 19:12

first pass

75b667b

more

0dffd94

pavelsavara added this to the 11.0.0 milestone Feb 3, 2026

pavelsavara self-assigned this Feb 3, 2026

pavelsavara added arch-wasm WebAssembly architecture area-System.Net.Sockets os-browser Browser variant of arch-wasm labels Feb 3, 2026

Copilot AI review requested due to automatic review settings February 3, 2026 20:38

Copilot started reviewing on behalf of pavelsavara February 3, 2026 20:38 View session

Copilot AI reviewed Feb 3, 2026

View reviewed changes

jkotas reviewed Feb 3, 2026

View reviewed changes

more

881e744

		#if defined(__EMSCRIPTEN__)
		if ((uintptr_t)((PBYTE)pBuffer + cbBuffer) <= emscripten_get_heap_size())

[browser] native - no sockets, no threads, no blocking #123962

Are you sure you want to change the base?

[browser] native - no sockets, no threads, no blocking #123962

Conversation

pavelsavara commented Feb 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dotnet-policy-service bot commented Feb 3, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

jkotas Feb 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jkotas Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

jkotas Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

pavelsavara commented Feb 4, 2026

Uh oh!

pavelsavara commented Feb 4, 2026

Uh oh!

jkotas commented Feb 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

pavelsavara commented Feb 3, 2026 •

edited

Loading

jkotas Feb 3, 2026 •

edited

Loading