Compile SWT natives for Windows on Arm64 (WoA). #1045

chirontt · 2024-02-11T19:22:20Z

This PR contains changes to the build.bat file, to support compiling the SWT natives for the WoA platform, and associated files to produce the SWT binaries module's jar/zip files for WoA.

On a WoA box, run the following commands to produce the SWT natives (swt*.dll) for WoA:

cd binaries\org.eclipse.swt.win32.win32.aarch64
mvn clean process-resources -Dnative=win32.win32.aarch64

and the swt*.dll files for WoA will be created in the current directory.

Cross-compiling between the x64 and Arm64 platforms, where either of them can be host or target, is also possible, after setting up the TARGET_ARCH environment variable with correct "target architecture" value, and with the SWT_JAVA_HOME property pointing to an available JDK of the target architecture.

For example, to cross-compile on an x64 host and produce SWT natives for Arm64 target, run the following commands:

set TARGET_ARCH=arm64
cd binaries\org.eclipse.swt.win32.win32.aarch64
mvn clean process-resources -Dnative=win32.win32.aarch64 -DSWT_JAVA_HOME=\path\to\arm64_jdk

and the swt*.dll files for WoA will be created in the current directory. Note the above SWT_JAVA_HOME property in the command line which points to an available Arm64 JDK of the target architecture. If needed, an Arm64 JDK for WoA can be downloaded from Microsoft, Liberica, or Zulu.

Similarly, to cross-compile on an Arm64 host and produce SWT natives for x64 target, run the following commands:

set TARGET_ARCH=x64
cd binaries\org.eclipse.swt.win32.win32.x86_64
mvn clean process-resources -Dnative=win32.win32.x86_64 -DSWT_JAVA_HOME=\path\to\x64_jdk

and the swt*.dll files for x64 will be created in the current directory.

Note that for cross-compiling between x64 and Arm64 to work, install the MSVC compiler version 2022, with mandatory build tools for both platforms included in the installation. Further more, there must exist a JDK installation of the target platform available on the host machine, so that the SWT_JAVA_HOME variable can point to it during cross-compiling and linking.

Once the SWT natives for Arm64 are generated, the SWT binaries module's files can be produced with the following build commands in current directory:

mvn clean verify -DskipTests=true

and the module's jar/zip files will be produced in the target directory:

org.eclipse.swt.win32.win32.aarch64-<version>.jar
org.eclipse.swt.win32.win32.aarch64-<version>-sources.jar
swt-<version>-win32-win32-aarch64.zip

hmartinez82 · 2024-02-11T22:00:27Z

bundles/org.eclipse.swt/Eclipse SWT PI/win32/library/os_custom.c

+#elif defined(_M_ARM64)
+	/* Not implemented yet */
+	functionPtr; /* to prevent: warning C4100: 'functionPtr': unreferenced formal parameter */
+	return FALSE;


You should return TRUE here and SWT on Windows on ARM should require Windows 10 1903 or higher.

Or we can implement the disassembly hack that is done for x64 with ARM64 code.

For more on the matter, see:
mintty/mintty#983 (comment)
and
https://stackoverflow.com/a/58547831/827532

Yes, return TRUE solves the problem of scrollbars not in dark mode, on my WoA box! The scrollbars now show dark color correctly:

I will fix the code later, thanks. Does that mean we don't need any disassembly hack for WoA then?

If we state that SWT on Windows on ARM require Windows 1903+ then we can return TRUE there and not do the assembly hack.

Also. I personally would've not done that assembly hack. If you look at one of the links above, this is way less hacky:

auto ord135 = GetProcAddress(hUxtheme, MAKEINTRESOURCEA(135)); if (g_buildNumber < 18334) _AllowDarkModeForApp = reinterpret_cast<fnAllowDarkModeForApp>(ord135); else _SetPreferredAppMode = reinterpret_cast<fnSetPreferredAppMode>(ord135);

In the Validate_AllowDarkModeForWindowWithTelemetryId() method all they would have to do was to just query the build version of Windows to decide if TRUE or FALSE stating that uxtheme.dll has the expected function at ordinal 135.

I've updated the PR with new commit returning TRUE for Arm64, so no need for any further implementation. Windows on Arm64 is really only Windows 11 going forward, so there's no need for any checking for Windows 10 1903+ in Arm64 because Windows 10 is no longer supported in Arm64 by Microsoft.

I think that what @SyntevoAlex is saying is that since these three functions are undocumented (they are not even exported by name, only by ordinal) they could be moved to another ordinal completely or even removed in the future builds of Windows 11 , thus it's unsafe to just return TRUE without checking first if the function is indeed the one expected at a given ordinal.

Yes, the functions are still NOT documented, so they still need to be checked.

@SyntevoAlex @hmartinez82 I don't understand any of the proposed changes to the os_custom.c file.

I'm thinking of removing this os_custom.c file from this PR, so that you guys can submit it in a separate new PR with your proposed changes for Arm64. And my PR will wait for your new PR to be submitted first.

@chirontt I can do that, but I don't know how to compile this. I have no Maven or Java experience :(
That's why I copied the code here. I can try to explain to you were those numbers 0x18 and 0x1C are coming from:

The validation check for x64, was done by looking at the machine code of those three functions, AllowDarkModeForWindow(), AllowDarkModeForWindowWithTelemetryId(), SetPreferredAppMode().
This was done by someone one Visual Studio, in an x64 machine.
With the disassembly of the code, one can see which bytes are at each address after the start of the function (the start of the function is the value returned by GetProcAddress())

Here's SetPreferredAppMode() for instance on x64:

As you can see, you have to open the disassembly view to be able to Step Into something you don't have the source code for. You do that by stepping in the Disassembly View until you get to the call and then you hit Step Into and finally see the first instruction at the beginning of the function.

The numbers circled in blue in the picture above are being tested in this line for instance https://github.com/eclipse-platform/eclipse.platform.swt/pull/1045/files#diff-d6068e15e4497052c1129ba671192ebce3920a189d39a48d0cb66e151019f9b8R227 in Validate_SetPreferredAppMode()

I then did all this dance, for the three functions in ARM64, got the disassembly of them all (see my pasted ASM code further down) and came up with three possible validations. Those three opcodes that I used for checking are 0x18 bytes from the beginning of theAllowDarkModeForWindow(), AllowDarkModeForWindowWithTelemetryId() and 0x1C bytes from the beginning of SetPreferredAppMode(). Since in ARM64 all instructions are fixed length (unlike x86 and x86_64) it was easy to just convert them to an uint32_t and compare directly to the expected opcodes.

@SyntevoAlex @chirontt Please review #1048 :)

SyntevoAlex · 2024-02-12T09:24:00Z

All of the above comments are wrong.

The problem is that API is not documented. Ordinal numbers are unreliable and may change any time, that's why APIs are additionally validated. Otherwise it's possible that one day, there would be entirely different API on ordinal 135, and Eclipse would just immediately crash on every start.

By returning TRUE, you just disable safety checks and say "it's fine", regardless of whether it's fine or not.

SyntevoAlex · 2024-02-12T09:28:54Z

This is (in short way) explained in code with code comment:

	/*
	 * In next Windows version, some other function can end up having this ordinal.
	 * Compare function's code to known signature to make sure.
	 */

hmartinez82 · 2024-02-12T09:34:47Z

So we need an ARM64 equivalent of the hack? I can try looking into it.

SyntevoAlex · 2024-02-12T09:36:48Z

Right. One would need to dissasemble ARM windows DLLs and decide which part of it uniquely identifies desired functions.

SyntevoAlex · 2024-02-12T09:37:25Z

Or you can just return FALSE for now. It can be supported later.

SyntevoAlex · 2024-02-12T09:48:42Z

If you choose to return FALSE, please add another code comment explaining why.

hmartinez82 · 2024-02-12T10:27:07Z

AllowDarkModeForWindow():

00007FFB16C32450 D503237F             pacibsp  
00007FFB16C32454 A9BE53F3             stp         x19,x20,[sp,#-0x20]!  
00007FFB16C32458 F9000BF5             str         x21,[sp,#0x10]  
00007FFB16C3245C A9BF7BFD             stp         fp,lr,[sp,#-0x10]!  
00007FFB16C32460 910003FD             mov         fp,sp  
00007FFB16C32464 53001C33             uxtb        w19,w1  
00007FFB16C32468 D29523C1             mov         x1,#0xA91E  
00007FFB16C3246C AA0003F5             mov         x21,x0  
00007FFB16C32470 9400F226             bl          #GetPropW (07FFB16C6ED08h)
...

AllowDarkModeForWindowWithTelemetryId():

00007FFB16C33210 D503237F             pacibsp  
00007FFB16C33214 A9BE53F3             stp         x19,x20,[sp,#-0x20]!  
00007FFB16C33218 F9000BF5             str         x21,[sp,#0x10]  
00007FFB16C3321C A9BF7BFD             stp         fp,lr,[sp,#-0x10]!  
00007FFB16C33220 910003FD             mov         fp,sp  
00007FFB16C33224 53001C34             uxtb        w20,w1  
00007FFB16C33228 D29523C1             mov         x1,#0xA91E  
00007FFB16C3322C AA0003F5             mov         x21,x0  
00007FFB16C33230 9400EEB6             bl          #GetPropW (07FFB16C6ED08h)
...

SetPreferredAppMode():

00007FFB16C33E10 D503237F             pacibsp  
00007FFB16C33E14 F81F0FF3             str         x19,[sp,#-0x10]!  
00007FFB16C33E18 A9BF7BFD             stp         fp,lr,[sp,#-0x10]!  
00007FFB16C33E1C 910003FD             mov         fp,sp  
00007FFB16C33E20 D00007E8             adrp        x8,wil::details::g_enabledStateManager+40h (07FFB16D31000h)  
00007FFB16C33E24 B94BD913             ldr         w19,[x8,#0xBD8]  
00007FFB16C33E28 2A0003E1             mov         w1,w0  
00007FFB16C33E2C 912F6100             add         x0,x8,#0xBD8  
00007FFB16C33E30 97FF7910             bl          _InterlockedExchange (07FFB16C12270h)  
00007FFB16C33E34 2A1303E0             mov         w0,w19  
00007FFB16C33E38 A8C17BFD             ldp         fp,lr,[sp],#0x10  
00007FFB16C33E3C F84107F3             ldr         x19,[sp],#0x10  
00007FFB16C33E40 D50323FF             autibsp  
00007FFB16C33E44 D65F03C0             ret  
00007FFB16C33E48 B0000090             adrp        xip0,GetBufferedPaintBitsEx+7320h (07FFB16C44000h)
...

SyntevoAlex

I didn't review Java specific build files because I have no experience with these.

SyntevoAlex · 2024-02-12T21:55:53Z

bundles/org.eclipse.swt/Eclipse SWT PI/win32/library/build.bat

+  IF "x.%OUTPUT_DIR%"=="x." (
+    CALL :ECHO "ERROR: OUTPUT_DIR environment variable must be set for manual cross-compile."
+    EXIT /B 1
+  )


I think that this code can be simplified.

Realistically I don't expect any new archs in upcoming years, so there will be just two - amd64 and arm64. So I suggest to have just one new variable that would override %PROCESSOR_ARCHITECTURE%. For example, COMPILE_ARCH could be the only new variable. If it's empty, it falls back to current processor arch:

IF "%COMPILE_ARCH%"=="" SET COMPILE_ARCH=%PROCESSOR_ARCHITECTURE%

this way, you only use one var instead of two, and there's less code.

Doing it this way means the cross-compiling idea is abandoned. Is this what you really want?

The required OUTPUT_DIR variable must be set, either internally within this build.bat file, or externally set by the caller of this build.bat script. Perhaps we can make this build.bat script smarter, in working out the required OUTPUT_DIR value internally from the COMPILE_ARCH (or XC_HOST_TARGET) value, by parsing either x64_arm64 or arm64_x64 and working out the needed OUTPUT_DIR value from there.

Cross-compiled arm64 is the same as native-compiled arm64. Therefore, if arm64 is compiled, output dirs will be for arm64. Even if it's compiled on amd64.

Nope, on an x64 box, passing 'arm64' to initialize and compile the C code for Arm64 will certainly fail to work (I've tried it). For cross-compile, MSVC must know from which host to which target explicitly, hence it requires 'x64_arm64' on an x64 box, or 'arm64_x64' on an Arm64 box, for cross-compiling.

ok, so the expected value for BUILD_ARCH is for example x64_arm64, I didn't catch that initially. I think that it still makes sense to compose BUILD_ARCH from %PROCESSOR_ARCHITECTURE% and requested target arch, because the code would be easier to understand. What do you think?

REM Compose host architecture string for MSVC IF "%PROCESSOR_ARCHITECTURE%"=="AMD64" ( SET HOST_ARCH=x64 ) ELSE IF "%PROCESSOR_ARCHITECTURE%"=="ARM64" ( SET HOST_ARCH=arm64 ) ELSE ( CALL :ECHO "ERROR: Unknown host architecture: %PROCESSOR_ARCHITECTURE%." EXIT /B 1 ) REM %TARGET_ARCH% may be specified by the caller for cross-compiling. REM If not, build for builder machine's architecture IF "%TARGET_ARCH%"=="" ( SET TARGET_ARCH=%HOST_ARCH% ) REM Compose build argument for MSVC IF "%TARGET_ARCH%"=="%HOST_ARCH%" ( SET BUILD_ARCH=%TARGET_ARCH% ) ELSE ( SET BUILD_ARCH=%HOST_ARCH%_%TARGET_ARCH% ) REM Select build directory based on target arch IF "%TARGET_ARCH%"=="x64" ( IF "x.%OUTPUT_DIR%"=="x." SET OUTPUT_DIR=..\..\..\org.eclipse.swt.win32.win32.x86_64 ) ELSE IF "%TARGET_ARCH%"=="arm64" ( IF "x.%OUTPUT_DIR%"=="x." SET OUTPUT_DIR=..\..\..\org.eclipse.swt.win32.win32.aarch64 ) ELSE ( CALL :ECHO "ERROR: Unknown target architecture: %TARGET_ARCH%." EXIT /B 1 )

Thinking about it some more, I figured that my approach has an even bigger advantage: user only specifies what he wants, and doesn't have to care about how to make this happen. Script will automatically figure the need to use cross-compiling.

This, in turn, will simplify builder scripts: builder will just say "build that and that", regardless of what they're running on.

OK, your approach is a bit verbose, but it's quite clear in intention, so I have no problem with it. And current Maven build scripts would still work with it without change (this was my intention all along.) And doing cross-compiling has to be a manual operation anyway, that the user need to explicitly specify an extra environment variable in order to trigger it, so I won't expect the current automated build scripts to do this cross-compiling, but the facility is there for automated cross-compile by the build scripts if required.

I'll merge this with the PR and squash it. And the PR's description need be changed too, to reflect the updated contents.

Another thing: I plan to apply these changes of build.bat, to the build.bat of the Equinox project, to generate the Equinox launcher native for Arm64. If you have different idea (for the Equinox project), let me know in advance, thanks.

hmartinez82 · 2024-02-13T05:44:57Z

AllowDarkModeForWindow():

00007FFB16C32450 D503237F             pacibsp  
00007FFB16C32454 A9BE53F3             stp         x19,x20,[sp,#-0x20]!  
00007FFB16C32458 F9000BF5             str         x21,[sp,#0x10]  
00007FFB16C3245C A9BF7BFD             stp         fp,lr,[sp,#-0x10]!  
00007FFB16C32460 910003FD             mov         fp,sp  
00007FFB16C32464 53001C33             uxtb        w19,w1  
00007FFB16C32468 D29523C1             mov         x1,#0xA91E  
00007FFB16C3246C AA0003F5             mov         x21,x0  
00007FFB16C32470 9400F226             bl          #GetPropW (07FFB16C6ED08h)
...

AllowDarkModeForWindowWithTelemetryId():

00007FFB16C33210 D503237F             pacibsp  
00007FFB16C33214 A9BE53F3             stp         x19,x20,[sp,#-0x20]!  
00007FFB16C33218 F9000BF5             str         x21,[sp,#0x10]  
00007FFB16C3321C A9BF7BFD             stp         fp,lr,[sp,#-0x10]!  
00007FFB16C33220 910003FD             mov         fp,sp  
00007FFB16C33224 53001C34             uxtb        w20,w1  
00007FFB16C33228 D29523C1             mov         x1,#0xA91E  
00007FFB16C3322C AA0003F5             mov         x21,x0  
00007FFB16C33230 9400EEB6             bl          #GetPropW (07FFB16C6ED08h)
...

SetPreferredAppMode():

00007FFB16C33E10 D503237F             pacibsp  
00007FFB16C33E14 F81F0FF3             str         x19,[sp,#-0x10]!  
00007FFB16C33E18 A9BF7BFD             stp         fp,lr,[sp,#-0x10]!  
00007FFB16C33E1C 910003FD             mov         fp,sp  
00007FFB16C33E20 D00007E8             adrp        x8,wil::details::g_enabledStateManager+40h (07FFB16D31000h)  
00007FFB16C33E24 B94BD913             ldr         w19,[x8,#0xBD8]  
00007FFB16C33E28 2A0003E1             mov         w1,w0  
00007FFB16C33E2C 912F6100             add         x0,x8,#0xBD8  
00007FFB16C33E30 97FF7910             bl          _InterlockedExchange (07FFB16C12270h)  
00007FFB16C33E34 2A1303E0             mov         w0,w19  
00007FFB16C33E38 A8C17BFD             ldp         fp,lr,[sp],#0x10  
00007FFB16C33E3C F84107F3             ldr         x19,[sp],#0x10  
00007FFB16C33E40 D50323FF             autibsp  
00007FFB16C33E44 D65F03C0             ret  
00007FFB16C33E48 B0000090             adrp        xip0,GetBufferedPaintBitsEx+7320h (07FFB16C44000h)
...

@SyntevoAlex any recommendations here? Since the instructions have fixed width, I was going to just compare the expected opcodes at a certain offset from the start of the functions.
For AllowDarkModeForWindow() check for D29523C1 mov x1,#0xA91E
For AllowDarkModeForWindowWithTelemetryId() check for D29523C1 mov x1,#0xA91E
For SetPreferredAppMode() check for 912F6100 add x0,x8,#0xBD8

Any other suggestion?

SyntevoAlex · 2024-02-13T11:21:12Z

I think that suggested checks are reasonable.

hmartinez82 · 2024-02-14T01:38:34Z

@chirontt here are the verifications to be used instead of just returning TRUE in the _M_ARM64 sections:
For Validate_AllowDarkModeForWindow():

#ifdef _M_ARM64
	if (*reinterpret_cast<const uint32_t*>(&functionPtr[0x18]) == 0xD29523C1) // mov x1,#0xA91E
	{
		return TRUE;
	}

	return FALSE;

For AllowDarkModeForWindowWithTelemetryId():

#ifdef _M_ARM64
	if (*reinterpret_cast<const uint32_t*>(&functionPtr[0x18]) == 0xD29523C1) // mov x1,#0xA91E
	{
		return TRUE;
	}

	return FALSE;

For SetPreferredAppMode():

#ifdef _M_ARM64
        if (*reinterpret_cast<const uint32_t*>(&functionPtr[0x1C]) == 0x912F6100) // add x0,x8,#0xBD8
	{
		return TRUE;
	}

    return FALSE;

@niraj-modi Are these satisfactory?

chirontt · 2024-02-17T05:05:32Z

I've restored the os_custom.c file to its base version before this PR change, effectively removing it from this PR, as the file is now handled by PR #1048

SyntevoAlex · 2024-02-17T19:40:27Z

#1048 was merged.
Please squash branch (thus removing any changes in os_ciustom.c) and rebase it.
I'm still waiting for your reply regarding script improvement.

chirontt · 2024-02-18T01:21:28Z

Branch squashed and rebased to latest master; PR's description updated also, to correctly describe the updated contents.

To summarize the description, if you're on an x64 box and want to cross-compile to produce the SWT natives for Arm64, here are the commands:

set TARGET_ARCH=arm64
cd binaries\org.eclipse.swt.win32.win32.aarch64
mvn clean process-resources -Dnative=win32.win32.aarch64 -DSWT_JAVA_HOME=\path\to\arm64_jdk

If you're authorized, then you could commit the generated swt*.dll binaries to the repo.

SyntevoAlex

Looks good to me.
I guess we'll merge it when master is open again. Probably there's no rush to merge it ASAP?

github-actions · 2024-02-18T21:02:42Z

Test Results

299 files + 1 299 suites +1 6m 35s ⏱️ +12s
4 100 tests +34 4 092 ✅ +32 8 💤 +3 0 ❌ - 1
12 212 runs +34 12 137 ✅ +32 75 💤 +3 0 ❌ - 1

Results for commit 22037ec. ± Comparison against base commit b5e86e3.

♻️ This comment has been updated with latest results.

chirontt · 2024-03-05T00:30:25Z

Branch rebased to latest master; ready to merge.

HannesWell · 2024-03-05T00:54:02Z

Thanks for working on this.

If you're authorized, then you could commit the generated swt*.dll binaries to the repo.

The dll's should not be added manually only the Jenkins build should do that. With the current build pipeline the manually submitted would then probably be deleted on the next change that requires recompilation.
I'm currently on vacation but when back next week I can help you with setting up the build-pipeline for that, given the infrastructure is ready.

laeubi · 2024-03-05T05:17:22Z

Branch rebased to latest master; ready to merge.

I'm currently on vacation but when back next week I can help you with setting up the build-pipeline for that, given the infrastructure is ready.

@jonahgraham can you tell about the state of the hardware here?

binaries/org.eclipse.swt.win32.win32.aarch64/META-INF/MANIFEST.MF

laeubi

This module needs to be mentioned in https://github.com/eclipse-platform/eclipse.platform.swt/blob/master/binaries/pom.xml as well to be included in the build.

HannesWell · 2024-03-19T23:26:18Z

The build of the native binaries for Windows on ARM is now working compiling against a beta jdk-17 win32.aarch64.
On Jenkins the build is also passing (there are just some new warnings that make the build unstable) and the GH-workflows just fail because there are no .dll files in the win32.aarch64 fragment. So it looks like this is shaping up well.

What's the opinion of the other committers about using that to build the SWT natives for WoA for some time until Temurin provides official releases for them? Do you think that would be a ok or should we wait until there is an official GA release for those binaries? I'm torn on this one. It is not the cleanest solution, but it would allow more testing of the WoA jdk17, which would maybe help the temurin maintainers to gain more confidence about it.

I'm not sure if the WebView2Loader.dll file is needed anymore in SWT. I've read somewhere that it's no longer used in SWT, but I can't find the reference anymore.

I recently run the build without the WebView2Loader.dll for the existing win32.x86_64 fragment accidentally and some test-failures happened. I have to admit I don't now why such native library isn't just shipped with the WebView2, but at the moment it seems necessary.
But of course it would be nice to get rid of it. If you are interested you can create a PR where you remove the WebView2Loader.dll and we'll see if we can make it work.

SyntevoAlex · 2024-03-20T06:11:59Z

What's the opinion of the other committers about using that to build the SWT natives for WoA for some time until Temurin provides official releases for them?

I totally support this idea. I understand that JDK only plays minimal role in produced DLLs, mostly it just provides a header and a library to describe the shape of JNI interface. I expect that these files would be mostly identical in any flavor of aarch64 JDK.

SyntevoAlex · 2024-03-20T06:14:41Z

By "mostly identical" I mean that I'm convinced that DLLs compiled with one JDK or another JDK will work the same way.

chirontt · 2024-03-23T20:26:24Z

I'm not sure if the WebView2Loader.dll file is needed anymore in SWT. I've read somewhere that it's no longer used in SWT, but I can't find the reference anymore.

But of course it would be nice to get rid of it. If you are interested you can create a PR where you remove the WebView2Loader.dll and we'll see if we can make it work.

Sorry, I take back my comment about the (no)need for WebView2Loader.dll file; its very requirement was added about 3 years ago to support MS Edge browser in SWT:

eclipse.platform.swt/bundles/org.eclipse.swt/Eclipse SWT Browser/win32/org/eclipse/swt/browser/Edge.java

Line 31 in 103b10e

Library.loadLibrary("WebView2Loader");

So we need to add the WebView2Loader.dll for WoA to this PR, but I see that you've already raised a separate PR #1137 for it.

HannesWell · 2024-03-24T17:56:33Z

I'm not sure if the WebView2Loader.dll file is needed anymore in SWT. I've read somewhere that it's no longer used in SWT, but I can't find the reference anymore.

But of course it would be nice to get rid of it. If you are interested you can create a PR where you remove the WebView2Loader.dll and we'll see if we can make it work.

Sorry, I take back my comment about the (no)need for WebView2Loader.dll file; its very requirement was added about 3 years ago to support MS Edge browser in SWT:

Yes, exactly. I looked into WebView2Loader topic again yesterday but didn't have the time to answer here, yet.
I also tried to statically link the required code into the swt-win32.dll but didn't succeed (but I'm not experienced in that area and cannot even tell if that is possible). So I found no better solution than just adding the loader-dll for arm64 too.

@HannesWell There's one for ARM64 too. We use it at work. You can download it from https://www.nuget.org/packages/Microsoft.Web.WebView2/1.0.2365.46 . The NuGet package is just a zip file, inside it you'll find build\native\arm64\WebView2Loader.dll

Thank you @hmartinez82 for that hint. I have now I have now added the arm64 variant of the WebView2Loader.dll from https://nuget.info/packages/Microsoft.Web.WebView2/1.0.1150.38 to match #1137.

In order to get IP/License clearance to add the WebView2Loader.dll I created

https://gitlab.eclipse.org/eclipsefdn/emo-team/iplab/-/issues/14072

Once that is approved this should be ready for submission.

chirontt · 2024-03-26T00:18:11Z

@HannesWell Thanks for adding the WebView2Loader.dll to the branch. I've also added the WebView2_LICENSE.txt file, copied from the x86_64 counterpart.

HannesWell · 2024-03-27T18:02:55Z

@HannesWell Thanks for adding the WebView2Loader.dll to the branch. I've also added the WebView2_LICENSE.txt file, copied from the x86_64 counterpart.

That's a good catch. And that gave me the idea that it would be better to share the legal-files for the fragments of a platform in a common location. I therefore created #1144.
Once that is merged, this PR can be further simplified.

chirontt · 2024-03-29T00:27:49Z

With #1144 merged, I've rebased and updated this PR's branch to bring it up to date.

This adds the org.eclipse.swt.win32.aarch64 fragment and contains changes to the build.bat file, to support compiling the SWT natives for the Windows on Arm64 (WoA) platform. On a WoA box, run the following commands to produce the SWT natives (swt*.dll) for WoA: cd binaries\org.eclipse.swt.win32.win32.aarch64 mvn clean process-resources -Dnative=win32.win32.aarch64 and the swt*.dll files for WoA will be created in the current directory. Cross-compiling between the x64 and Arm64 platforms, where either of them can be host or target, is also possible, after setting up the TARGET_ARCH environment variable with correct target architecture value, and with the SWT_JAVA_HOME property pointing to an available JDK of the target architecture. For example, to cross-compile on an x64 host and produce SWT natives for Arm64 target, run the following commands: set TARGET_ARCH=arm64 cd binaries\org.eclipse.swt.win32.win32.aarch64 mvn clean process-resources -Dnative=win32.win32.aarch64 -DSWT_JAVA_HOME=\path\to\arm64_jdk and the swt*.dll files for WoA will be created in the current directory. Note the above SWT_JAVA_HOME property in the command line which points to an available Arm64 JDK of the target architecture. Similarly, to cross-compile on an Arm64 host and produce SWT natives for x64 target, run the following commands: set TARGET_ARCH=x64 cd binaries\org.eclipse.swt.win32.win32.x86_64 mvn clean process-resources -Dnative=win32.win32.x86_64 -DSWT_JAVA_HOME=\path\to\x64_jdk and the swt*.dll files for x64 will be created in the current directory. Note that for cross-compiling between x64 and Arm64 to work, install the MSVC compiler version 2022, with mandatory build tools for both platforms included in the installation. Once the SWT natives for Arm64 are generated, the SWT binaries module's files can be produced with the following build commands in current directory: mvn clean verify -DskipTests=true and the module's jar/zip files will be produced in the target directory: org.eclipse.swt.win32.win32.aarch64-<version>.jar org.eclipse.swt.win32.win32.aarch64-<version>-sources.jar swt-<version>-win32-win32-aarch64.zip Also add the WebView2Loader.dll version 1.0.1150.38 (matching the swt.win32.x86_64 fragment), obtained from the NuGet package microsoft.web.webview2 https://nuget.info/packages/Microsoft.Web.WebView2/1.0.1150.38 located in the sub-folder build/native/arm64

HannesWell

With #1144 merged, I've rebased and updated this PR's branch to bring it up to date.

Thank you.
I have just pushed another small update for the Jenkins file to have the work-around for the different source of the win32.aarch64 JDK located near the URL definition of the JDKs used by default.

Unfortunately I could not yet get more details about an estimation when Temurin will officially support WoA (see adoptium/adoptium-support#616 (comment)), so it looks like we have to continue with that workaround until it is available and justj can adopted it.
But as it was discussed, it shouldn't have an impact.

In order to get IP/License clearance to add the WebView2Loader.dll I created
* https://gitlab.eclipse.org/eclipsefdn/emo-team/iplab/-/issues/14072

The IP clearance was granted so everything is finally ready.

I did a final pass over the changes and all looks good to me (although I'm not an expert in configuring MSVC, but AFAICT it looks good).

Unfortunately there is currently an issue with the Eclipse build infrastructure that prevents the build from succeeding. But once that is resolved and we have a green build I think this is ready for submission.

HannesWell · 2024-03-29T17:42:11Z

Unfortunately there is currently an issue with the Eclipse build infrastructure that prevents the build from succeeding. But once that is resolved and we have a green build I think this is ready for submission.

The infrastructure issue has been resolved and all builds had passed.
After I submitted this PR the new natives finally have been built successfully at the master branch (the Jenkins instance had one or two hick-ups and a out-of-schedule I-build occupied a lot of resources) and SWT now has a win32.aarch64 fragment with native binaries.

Thank you @chirontt for this nice contribution and everyone that helped.

Please note that o.e.swt.win32.aarch64 is not yet included in the org.eclipse.e4.rcp feature and will therefore not yet be included in the Eclipse 4.32 I-builds update-site:
https://github.com/eclipse-platform/eclipse.platform.ui/blob/3f00c8b2d69ac6f06a803c046a61a9a104ba5790/features/org.eclipse.e4.rcp/feature.xml#L227-L271

This should only be done once the Equinox launcher is available and and I-build for WoA is at least prepared.
Further progress can be tracked in eclipse-platform/eclipse.platform.releng.aggregator#577 or the linked more specific issues.

hmartinez82 · 2024-03-29T20:18:18Z

Where can one download the ARM64 binaries?

akurtakov · 2024-03-29T20:33:08Z

At some point they should be visible in I-builds pages e.g. https://download.eclipse.org/eclipse/downloads/drops4/I20240329-0530/#SWT but not yet there.
@HannesWell is an entry point to see what/how/where is used to publish to this page.

chirontt · 2024-03-29T21:32:08Z

Please note that o.e.swt.win32.aarch64 is not yet included in the org.eclipse.e4.rcp feature and will therefore not yet be included in the Eclipse 4.32 I-builds update-site:
https://github.com/eclipse-platform/eclipse.platform.ui/blob/3f00c8b2d69ac6f06a803c046a61a9a104ba5790/features/org.eclipse.e4.rcp/feature.xml#L227-L271

Right, they are contained in my PR (for WoA) for that eclipse.platform.ui repo: eclipse-platform/eclipse.platform.ui#1787

Please review it.

chirontt · 2024-03-29T22:35:57Z

Where can one download the ARM64 binaries?

They are now generated and stored in this repo:
https://github.com/eclipse-platform/eclipse.platform.swt/tree/master/binaries/org.eclipse.swt.win32.win32.aarch64

chirontt requested review from niraj-modi and iloveeclipse as code owners February 11, 2024 19:22

chirontt mentioned this pull request Feb 11, 2024

[help wanted][win32] SWT Support for Windows on ARM #1019

Closed

hmartinez82 reviewed Feb 11, 2024

View reviewed changes

SyntevoAlex reviewed Feb 12, 2024

View reviewed changes

hmartinez82 mentioned this pull request Feb 14, 2024

Handle validation of uxtheme.dll undocumented exported functions in ARM64 builds #1048

Merged

SyntevoAlex linked an issue Feb 17, 2024 that may be closed by this pull request

[help wanted][win32] SWT Support for Windows on ARM #1019

Closed

chirontt force-pushed the swt_windows_arm64_compile branch from f426a5c to 1524c8d Compare February 18, 2024 01:06

SyntevoAlex approved these changes Feb 18, 2024

View reviewed changes

SyntevoAlex self-assigned this Feb 18, 2024

chirontt force-pushed the swt_windows_arm64_compile branch from 1524c8d to be6bd28 Compare March 5, 2024 00:28

laeubi reviewed Mar 5, 2024

View reviewed changes

binaries/org.eclipse.swt.win32.win32.aarch64/META-INF/MANIFEST.MF Show resolved Hide resolved

laeubi reviewed Mar 5, 2024

View reviewed changes

HannesWell force-pushed the swt_windows_arm64_compile branch from 7b4e3d7 to 6b4cd63 Compare March 19, 2024 23:09

HannesWell force-pushed the swt_windows_arm64_compile branch from e0c4bc0 to 6b4cd63 Compare March 19, 2024 23:59

This was referenced Mar 22, 2024

Updated the code related to LoongArch and added the necessary so files. #1135

Open

[help wanted] Native support for Eclipse SDKs on Windows Arm64 eclipse-platform/eclipse.platform.releng.aggregator#577

Open

HannesWell force-pushed the swt_windows_arm64_compile branch from 6b4cd63 to 44c37c5 Compare March 23, 2024 12:40

HannesWell mentioned this pull request Mar 23, 2024

Update WebView2Loader.dll to WebView2 1.0.1150.38 #1137

Merged

HannesWell force-pushed the swt_windows_arm64_compile branch from 44c37c5 to 9976ce7 Compare March 24, 2024 17:49

chirontt force-pushed the swt_windows_arm64_compile branch from 9976ce7 to 9b90d36 Compare March 26, 2024 00:13

chirontt force-pushed the swt_windows_arm64_compile branch from 9b90d36 to d4b1768 Compare March 29, 2024 00:21

HannesWell force-pushed the swt_windows_arm64_compile branch from d4b1768 to 4513c52 Compare March 29, 2024 07:50

HannesWell force-pushed the swt_windows_arm64_compile branch from 4513c52 to 22037ec Compare March 29, 2024 07:51

HannesWell approved these changes Mar 29, 2024

View reviewed changes

HannesWell merged commit 9ccc8e7 into eclipse-platform:master Mar 29, 2024
10 of 13 checks passed

chirontt deleted the swt_windows_arm64_compile branch March 29, 2024 18:28

HannesWell mentioned this pull request Mar 31, 2024

fix missing win32.aarch64 fragment in various p2.inf #1151

Merged

hmartinez82 mentioned this pull request Apr 13, 2024

Validate SetPreferredAppMode() in ARM64 with ADRP instruction instead #1172

Open

Compile SWT natives for Windows on Arm64 (WoA). #1045

Compile SWT natives for Windows on Arm64 (WoA). #1045

Conversation

chirontt commented Feb 11, 2024 • edited

hmartinez82 Feb 11, 2024 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hmartinez82 Feb 12, 2024 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hmartinez82 Feb 14, 2024 • edited

Choose a reason for hiding this comment

hmartinez82 Feb 14, 2024 • edited

Choose a reason for hiding this comment

SyntevoAlex commented Feb 12, 2024 • edited

SyntevoAlex commented Feb 12, 2024

hmartinez82 commented Feb 12, 2024

SyntevoAlex commented Feb 12, 2024

SyntevoAlex commented Feb 12, 2024

SyntevoAlex commented Feb 12, 2024

hmartinez82 commented Feb 12, 2024 • edited

SyntevoAlex left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chirontt Feb 13, 2024 • edited

Choose a reason for hiding this comment

SyntevoAlex Feb 13, 2024 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

SyntevoAlex Feb 14, 2024 • edited

Choose a reason for hiding this comment

SyntevoAlex Feb 15, 2024 • edited

Choose a reason for hiding this comment

SyntevoAlex Feb 15, 2024 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hmartinez82 commented Feb 13, 2024 • edited

SyntevoAlex commented Feb 13, 2024

hmartinez82 commented Feb 14, 2024 • edited

chirontt commented Feb 17, 2024

SyntevoAlex commented Feb 17, 2024

chirontt commented Feb 18, 2024

SyntevoAlex left a comment

Choose a reason for hiding this comment

github-actions bot commented Feb 18, 2024 • edited

Test Results

chirontt commented Mar 5, 2024

HannesWell commented Mar 5, 2024 • edited

laeubi commented Mar 5, 2024

laeubi left a comment

Choose a reason for hiding this comment

HannesWell commented Mar 19, 2024

SyntevoAlex commented Mar 20, 2024 • edited

SyntevoAlex commented Mar 20, 2024

chirontt commented Mar 23, 2024

HannesWell commented Mar 24, 2024 • edited

chirontt commented Mar 26, 2024

HannesWell commented Mar 27, 2024

chirontt commented Mar 29, 2024

HannesWell left a comment

Choose a reason for hiding this comment

HannesWell commented Mar 29, 2024

hmartinez82 commented Mar 29, 2024

akurtakov commented Mar 29, 2024

chirontt commented Mar 29, 2024

chirontt commented Mar 29, 2024

chirontt commented Feb 11, 2024 •

edited

hmartinez82 Feb 11, 2024 •

edited

hmartinez82 Feb 12, 2024 •

edited

hmartinez82 Feb 14, 2024 •

edited

hmartinez82 Feb 14, 2024 •

edited

SyntevoAlex commented Feb 12, 2024 •

edited

hmartinez82 commented Feb 12, 2024 •

edited

chirontt Feb 13, 2024 •

edited

SyntevoAlex Feb 13, 2024 •

edited

SyntevoAlex Feb 14, 2024 •

edited

SyntevoAlex Feb 15, 2024 •

edited

SyntevoAlex Feb 15, 2024 •

edited

hmartinez82 commented Feb 13, 2024 •

edited

hmartinez82 commented Feb 14, 2024 •

edited

github-actions bot commented Feb 18, 2024 •

edited

HannesWell commented Mar 5, 2024 •

edited

SyntevoAlex commented Mar 20, 2024 •

edited

HannesWell commented Mar 24, 2024 •

edited