Update Emscripten jobs to add llvm 21 #776

mcbarton · 2026-01-14T20:04:32Z

During the llvm 21 upgrade PR we couldn't upgrade the Emscripten jobs due to an issue where using wasm exceptions broke between llvm 20 and 21. This PR https://github.com/llvm/llvm-project/issues/175717contains the changes which will enable an llvm 21 Emscripten build with wasm exceptions, once the following llvm issue llvm/llvm-project#175717 opened by Anutosh has been resolved .

github-actions · 2026-01-14T20:15:45Z

clang-tidy review says "All clean, LGTM! 👍"

github-actions · 2026-01-19T13:32:22Z

clang-tidy review says "All clean, LGTM! 👍"

codecov · 2026-01-19T13:50:27Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.32%. Comparing base (dea9b71) to head (8277990).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #776   +/-   ##
=======================================
  Coverage   79.32%   79.32%           
=======================================
  Files          11       11           
  Lines        3966     3966           
=======================================
  Hits         3146     3146           
  Misses        820      820

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

patches/llvm/emscripten-clang21-2-enable_exception_handling.patch

github-actions · 2026-01-23T10:18:22Z

clang-tidy review says "All clean, LGTM! 👍"

github-actions · 2026-01-23T10:42:50Z

clang-tidy review says "All clean, LGTM! 👍"

github-actions · 2026-01-23T10:56:56Z

clang-tidy review says "All clean, LGTM! 👍"

anutosh491

Hi @vgvassilev @mcbarton,

All the changes we want have been applied as patches here and I just expect everything to work and the tests to pass here.

But something that concerns me heavily is providing an option to switch between EH model (wasm exceptions vs emscripten exceptions) that is being done through
CPPINTEROP_ENABLE_WASM_EXCEPTIONS

It is just not worth it. The reward we get out of it vs the maintanance cost doesn't match at all !

A) Firstly as a product, we shouldn't be giving our users options for anything & everything. It's just an exception handling model, we can just go with 1

B) We've switched to emscripten 4 on cppinterop and using packages from https://prefix.dev/emscripten-forge-4x for building the deployment. All the packages on this channel (around 400) have been built using wasm exceptions, just cause we realized that emscripten 4 works well with wasm exceptions (this was mainly done for scipy and other R packages but has its effect on everything on the channel like C++/Rust packages)

C) The reward for all this maintaince is just 1 single cell (cell 3) in our demo notebook.

We can obviously solve this cell through different EH models but that should not be the goto. We should have more confidence in one Exception model we pick and report issues on llvm as and when we find them in the future.

Sure, we can do it for both ... but then both model might have different errors in the future that we might have to fix on LLVM's side which will be painful and introduces lot of waiting.

I suggest let's go ahead with one model and build our confidence in it. That should be enough !

github-actions · 2026-01-23T18:51:44Z

clang-tidy review says "All clean, LGTM! 👍"

.github/workflows/emscripten.yml

anutosh491 · 2026-01-23T20:31:41Z

Wait, I don't get why the wasm exceptions related flags/changes have been dropped in the previous commit.

We are moving away from the emscripten EH model and towards the wasm EH model !!

That's what the patches are about and all the discussion & effort Vassil and I put in with the Prs for the past 1.5 weeks !

github-actions · 2026-01-24T08:14:01Z

clang-tidy review says "All clean, LGTM! 👍"

vgvassilev · 2026-01-24T09:23:20Z

.github/workflows/deploy-pages.yml

                -DCMAKE_INSTALL_PREFIX=$PREFIX         \
                -DCMAKE_FIND_ROOT_PATH_MODE_PACKAGE=ON            \
                -DSYSROOT_PATH=$SYSROOT_PATH               \
+                -DCPPINTEROP_ENABLE_WASM_EXCEPTIONS=ON        \


Suggested change

-DCPPINTEROP_ENABLE_WASM_EXCEPTIONS=ON \

-DCPPINTEROP_ENABLE_WASM_EXCEPTIONS=ON \

We make CPPINTEROP_ENABLE_WASM_EXCEPTIONS=ON default and we implement a new option CPPINTEROP_EXTRA_WASM_FLAGS=<...> where people can specify -sSUPPORT_LONGJMP=wasm -fwasm-exceptions which will be appended and overrule the default wasm exception model...

github-actions · 2026-01-24T17:07:20Z

clang-tidy review says "All clean, LGTM! 👍"

vgvassilev · 2026-01-24T17:09:41Z

CMakeLists.txt

+if(EMSCRIPTEN)
+  set(
+    CPPINTEROP_EXCEPTION_HANDLING_FLAGS
+    "-sSUPPORT_LONGJMP=wasm -fwasm-exceptions"
+    CACHE STRING
+    "Compiler flags for CPP interop exception handling"
+  )
+endif()


Suggested change

if(EMSCRIPTEN)

set(

CPPINTEROP_EXCEPTION_HANDLING_FLAGS

"-sSUPPORT_LONGJMP=wasm -fwasm-exceptions"

CACHE STRING

"Compiler flags for CPP interop exception handling"

)

endif()

option(CPPINTEROP_EXTRA_WASM_FLAGS "Enable custom flags for wasm" "-sSUPPORT_LONGJMP=wasm -fwasm-exceptions")

vgvassilev · 2026-01-24T17:14:32Z

docs/Emscripten-build-instructions.rst

+   EMCC_CFLAGS="-sSUPPORT_LONGJMP=wasm -fwasm-exceptions" emmake make libclang -j $(nproc --all)
+   EMCC_CFLAGS="-sSUPPORT_LONGJMP=wasm -fwasm-exceptions" emmake make clangInterpreter clangStaticAnalyzerCore -j $(nproc --all)
+   EMCC_CFLAGS="-sSUPPORT_LONGJMP=wasm -fwasm-exceptions" emmake make lldWasm -j $(nproc --all)


We can revert back these things as now we have them in the variables by default.

github-actions · 2026-01-24T17:24:55Z

clang-tidy review says "All clean, LGTM! 👍"

github-actions · 2026-01-24T17:44:55Z

clang-tidy review says "All clean, LGTM! 👍"

mcbarton · 2026-01-24T17:58:26Z

CMakeLists.txt

 option(CPPINTEROP_USE_REPL "Use clang-repl as backend" ON)
 option(CPPINTEROP_ENABLE_TESTING "Enable the CppInterOp testing infrastructure." ON)
+if(EMSCRIPTEN)
+  set(CPPINTEROP_EXTRA_WASM_FLAGS "-sSUPPORT_LONGJMP=wasm -fwasm-exceptions" CACHE STRING "Extra flags for wasm")


option(CPPINTEROP_EXTRA_WASM_FLAGS "Enable custom flags for wasm" "-sSUPPORT_LONGJMP=wasm -fwasm-exceptions") which was suggested didn't work. This works though.

The error you get using option can be found here https://github.com/compiler-research/CppInterOp/actions/runs/21318907457/job/61366057216?pr=776#step:9:324

github-actions · 2026-01-24T18:02:55Z

clang-tidy review says "All clean, LGTM! 👍"

github-actions · 2026-01-24T18:24:53Z

clang-tidy review says "All clean, LGTM! 👍"

anutosh491 · 2026-01-25T06:18:27Z

Emscripten-build-instructions.md

-emmake make lldWasm -j $(nproc --all)
+EMCC_CFLAGS="-sSUPPORT_LONGJMP=wasm -fwasm-exceptions" emmake make libclang -j $(nproc --all)
+EMCC_CFLAGS="-sSUPPORT_LONGJMP=wasm -fwasm-exceptions" emmake make clangInterpreter clangStaticAnalyzerCore -j $(nproc --all)
+EMCC_CFLAGS="-sSUPPORT_LONGJMP=wasm -fwasm-exceptions" emmake make lldWasm -j $(nproc --all)


This is probably not needed right ?

Building clangInterpreter above this line would need building lldwasm anyways.

Also once set, there's no need to repeat I'm guessing

export EMCC_CFLAGS="-sSUPPORT_LONGJMP=wasm -fwasm-exceptions"

And then we can just have emmake make ...

Yes, lldWasm is not a target we need and was planning to remove it at some point but haven't got around to it.

I don't plan to apply your export EMCC_CFLAGS suggestion though as that makes the environment variable persistent and if not unset then it would override cppinterops new config flag making the flag pointless. I will put all the make targets on a single line to reduce the diff, but keep the EMCC_CFLAGS to a single command rather than using export

lib/CppInterOp/CMakeLists.txt

anutosh491 · 2026-01-25T06:29:47Z

unittests/CppInterOp/VariableReflectionTest.cpp

  GTEST_SKIP() << "Test fails with Cling on Windows on ARM";
+#endif
+#ifdef __EMSCRIPTEN__
+  GTEST_SKIP() << "This test crashes for Emscripten builds of CppInterOp";


Hmm, could you paste the crash. Let's see if we can fix !

The ci doesn'f give any useful output for the error since we have a minimum size build. I don't have time to debug this locally in the time we have left. If you have time to debug locally though and find a solution i'll try it out in the PR.

vgvassilev · 2026-01-25T07:14:40Z

CMakeLists.txt

  if (APPLE OR EMSCRIPTEN)
    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fno-common -Woverloaded-virtual -Wcast-qual -fno-strict-aliasing -Wno-long-long -Wall -W -Wno-unused-parameter -Wwrite-strings")
+    if (EMSCRIPTEN)
+      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fwasm-exceptions")


Shouldn’t this go with in the option default value?

Yes if should, and is what I had originally. The reason it is like this at the moment is because -sSUPPORT_LONGJMP=wasm causes googletest to have a build warning, and we treat building warnings as errors. I planned on fixing is soon after post release since I don't really have any time go debug the googletest build error in the time we have before release. Would a fixme suffice for now?

Can we filter that one warning to error flag out from gtest? I think we have the logic in clad to copy paste from.

Changed the logic so that -sSUPPORT_LONGJMP=wasm is just filtered out of CMAKE_CXX_FLAGS just for the Emscripten Googletest build. We will see if my quick attempt at a filter works when the ci runs.

github-actions · 2026-01-25T09:43:32Z

clang-tidy review says "All clean, LGTM! 👍"

github-actions · 2026-01-25T10:19:12Z

clang-tidy review says "All clean, LGTM! 👍"

vgvassilev

This looks good to me. Let's wait for @anutosh491 to review it.

anutosh491

I think this is good.

Let's see if the deployment notebook works well after we merge this (the raytracer and mamba magic install should be a good test to confirm wasm exceptions)

Also falling short on time to check the error on this
#776 (comment) but going through the test, I can somewhat sense what the error is going to be (and I know off the fix already 😄 ) ... but I don't mind getting it in through another PR if someone can help me with the error log.

github-actions · 2026-01-26T12:02:09Z

clang-tidy review says "All clean, LGTM! 👍"

mcbarton force-pushed the wasm-exceptions-llvm-21 branch 8 times, most recently from dc36556 to 5d9fc77 Compare January 19, 2026 13:23

anutosh491 reviewed Jan 19, 2026

View reviewed changes

patches/llvm/emscripten-clang21-2-enable_exception_handling.patch Outdated Show resolved Hide resolved

mcbarton force-pushed the wasm-exceptions-llvm-21 branch 2 times, most recently from 718d5cb to 746beec Compare January 23, 2026 10:10

mcbarton force-pushed the wasm-exceptions-llvm-21 branch from 746beec to bf68cf8 Compare January 23, 2026 10:34

mcbarton force-pushed the wasm-exceptions-llvm-21 branch from bf68cf8 to f22dd9a Compare January 23, 2026 10:49

anutosh491 requested changes Jan 23, 2026

View reviewed changes

mcbarton changed the title ~~[wip] Build llvm 21 with wasm exceptions~~ Build llvm 21 with wasm exceptions Jan 23, 2026

mcbarton marked this pull request as ready for review January 23, 2026 12:55

mcbarton changed the title ~~Build llvm 21 with wasm exceptions~~ Update Emscripten jobs to add llvm 21 Jan 23, 2026

mcbarton requested a review from vgvassilev January 23, 2026 19:47

vgvassilev reviewed Jan 23, 2026

View reviewed changes

.github/workflows/emscripten.yml Outdated Show resolved Hide resolved

vgvassilev requested a review from anutosh491 January 23, 2026 20:23

Build llvm 21 with wasm exceptions

247ff63

mcbarton force-pushed the wasm-exceptions-llvm-21 branch from 20098ea to 247ff63 Compare January 23, 2026 21:33

Update emscripten.yml

00df377

vgvassilev reviewed Jan 24, 2026

View reviewed changes

Apply outcome of Zoom call discussion

8382e80

vgvassilev reviewed Jan 24, 2026

View reviewed changes

Apply Vassil suggestion

2aab862

vgvassilev reviewed Jan 24, 2026

View reviewed changes

Fix mistake in PR

26081a2

Try to fix ci

ff5807a

mcbarton commented Jan 24, 2026

View reviewed changes

Fix ci

a61c3c6

anutosh491 reviewed Jan 25, 2026

View reviewed changes

lib/CppInterOp/CMakeLists.txt Show resolved Hide resolved

anutosh491 reviewed Jan 25, 2026

View reviewed changes

vgvassilev reviewed Jan 25, 2026

View reviewed changes

mcbarton added 2 commits January 25, 2026 09:26

Remove references to lldWasm

4198238

Add FIXME to cmakelist.txt

c66c09d

Filter -sSUPPORT_LONGJMP=wasm from Emscripten Googletest build

4de3374

vgvassilev approved these changes Jan 25, 2026

View reviewed changes

anutosh491 approved these changes Jan 26, 2026

View reviewed changes

Fix mamba magic command in deployment

8277990

vgvassilev merged commit f0938a8 into compiler-research:main Jan 26, 2026
23 of 49 checks passed

	-DCPPINTEROP_ENABLE_WASM_EXCEPTIONS=ON \
	-DCPPINTEROP_ENABLE_WASM_EXCEPTIONS=ON \

Update Emscripten jobs to add llvm 21 #776

Update Emscripten jobs to add llvm 21 #776

Uh oh!

Conversation

mcbarton commented Jan 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Jan 14, 2026

Uh oh!

github-actions bot commented Jan 19, 2026

Uh oh!

codecov bot commented Jan 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Uh oh!

github-actions bot commented Jan 23, 2026

Uh oh!

github-actions bot commented Jan 23, 2026

Uh oh!

github-actions bot commented Jan 23, 2026

Uh oh!

anutosh491 left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 23, 2026

Uh oh!

Uh oh!

anutosh491 commented Jan 23, 2026

Uh oh!

github-actions bot commented Jan 24, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 24, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 24, 2026

Uh oh!

github-actions bot commented Jan 24, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 24, 2026

Uh oh!

github-actions bot commented Jan 24, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mcbarton Jan 25, 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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 25, 2026

Uh oh!

github-actions bot commented Jan 25, 2026

Uh oh!

vgvassilev left a comment

Choose a reason for hiding this comment

Uh oh!

anutosh491 left a comment

Choose a reason for hiding this comment

mcbarton commented Jan 14, 2026 •

edited

Loading

codecov bot commented Jan 19, 2026 •

edited

Loading

mcbarton Jan 25, 2026 •

edited

Loading