Merge pull request #25084 from EDVTAZ:emscripten-3.1.54-compat

Add compatibility with latest (3.1.54) emsdk version #25084

### Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [ ] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
      Patch to opencv_extra has the same branch name.
- [ ] The feature is well documented and sample code can be built with the project CMake

### Details

I was following [this tutorial](https://docs.opencv.org/4.9.0/d4/da1/tutorial_js_setup.html) for building opencv with wasm target. The tutorial mentions that the last verified version of emscripten that is tested with opencv is 2.0.10, but I was curious if I could get it to work with more recent versions. I've run into a few issues with the latest version, for which fixes are included in this PR. I've found a few issues that have the same problems I encountered:

- https://github.com/opencv/opencv/issues/24620
- https://github.com/opencv/opencv/issues/20313
- https://stackoverflow.com/questions/77469603/custom-opencv-js-wasm-using-cv-matfromarray-results-in-cv-mat-is-not-a-co
- https://github.com/emscripten-core/emscripten/issues/14803
- https://github.com/opencv/opencv/issues/24572
- https://github.com/opencv/opencv/issues/19493#issuecomment-857167996

I used the docker image for building and comparing results with different emsdk versions. I tested by building with `--build_wasm` and `--build-test` flags and ran the tests in the browser. I addressed the following issues with newer versions of emscripten:

- In newer versions `EMSCRIPTEN` environemnt variable was stopped being set. I added support for deriving location based on the `EMSDK` environment variable, as suggested [here](https://github.com/emscripten-core/emscripten/issues/14803)
- In newer versions emcmake started passing `-DCMAKE...` arguments, however the opencv python script didn't know how to handle them. I added processing to the args that will forward all arguments to `cmake` that start with `-D`. I opted for this in hopes of being more futureproof, but another approach could be just ignoreing them, or explicitly forwarding them instead of matching anything starting with `-D`. These approches were suggested [here](https://github.com/opencv/opencv/issues/19493#issuecomment-855529448)
- With [version 3.1.31](https://github.com/emscripten-core/emscripten/blob/main/ChangeLog.md#3131---012623) some previously exported functions stopped being automatically exported. Because of this, `_free` and `_malloc` were no longer available and had to be explicitly exported because of breaking tests.
- With [version 3.1.42](https://github.com/emscripten-core/emscripten/compare/3.1.41...3.1.42#diff-e505aa80b2764c0197acfc9afd8179b3600f0ab5dd00ff77db01879a84515cdbL3875) the `post-js` code doesn't receive the module named as `EXPORT_NAME` anymore, but only as `moduleArg`/`Module`. This broke existing code in `helpers.js`, which was referencing exported functions through `cv.Mat`, etc. I changed all of these references to use `Module.Mat`, etc. If it is preferred, alternatively the `cv` variable could be reintroduced in `helper.js` as suggested [here](https://github.com/opencv/opencv/issues/24620)

With the above changes in place, I can successfully build and run tests with the latest emscripten/emsdk docker image (also with 2.0.10 and most of the other older tags, except for a few that contain transient issues like [this](https://github.com/emscripten-core/emscripten/issues/17700)).

This is my first time contributing to opencv, so I hope I got everything correct in this PR, but please let me know if I should change anything!

modules/js/src/helpers.js

@@ -42,6 +42,10 @@ if (typeof Module.FS === 'undefined' && typeof FS !== 'undefined') {
     Module.FS = FS;
 }
 
+if (typeof cv === 'undefined') {
+    var cv = Module;
+}
+
 Module['imread'] = function(imageSource) {
     var img = null;
     if (typeof imageSource === 'string') {

platforms/js/build_js.py

@@ -84,7 +84,6 @@ class Builder:
             "-DPYTHON_DEFAULT_EXECUTABLE=%s" % sys.executable,
                "-DENABLE_PIC=FALSE", # To workaround emscripten upstream backend issue https://github.com/emscripten-core/emscripten/issues/8761
                "-DCMAKE_BUILD_TYPE=Release",
-               "-DCMAKE_TOOLCHAIN_FILE='%s'" % self.get_toolchain_file(),
                "-DCPU_BASELINE=''",
                "-DCMAKE_INSTALL_PREFIX=/usr/local",
                "-DCPU_DISPATCH=''",
@@ -145,6 +144,8 @@ class Builder:
                "-DBUILD_PERF_TESTS=ON"]
         if self.options.cmake_option:
             cmd += self.options.cmake_option
+        if not self.options.cmake_option or all(["-DCMAKE_TOOLCHAIN_FILE" not in opt for opt in self.options.cmake_option]):
+            cmd.append("-DCMAKE_TOOLCHAIN_FILE='%s'" % self.get_toolchain_file())
         if self.options.build_doc:
             cmd.append("-DBUILD_DOCS=ON")
         else:
@@ -194,6 +195,7 @@ class Builder:
             flags += self.options.build_flags
         if self.options.webnn:
             flags += "-s USE_WEBNN=1 "
+        flags += "-s EXPORTED_FUNCTIONS=\"['_malloc', '_free']\""
         return flags
 
     def config(self):
@@ -224,10 +226,12 @@ if __name__ == "__main__":
 
     opencv_dir = os.path.abspath(os.path.join(SCRIPT_DIR, '../..'))
     emscripten_dir = None
-    if "EMSCRIPTEN" in os.environ:
+    if "EMSDK" in os.environ:
+        emscripten_dir = os.path.join(os.environ["EMSDK"], "upstream", "emscripten")
+    elif "EMSCRIPTEN" in os.environ:
         emscripten_dir = os.environ["EMSCRIPTEN"]
     else:
-        log.warning("EMSCRIPTEN environment variable is not available. Please properly activate Emscripten SDK and consider using 'emcmake' launcher")
+        log.warning("EMSCRIPTEN/EMSDK environment variable is not available. Please properly activate Emscripten SDK and consider using 'emcmake' launcher")
 
     parser = argparse.ArgumentParser(description='Build OpenCV.js by Emscripten')
     parser.add_argument("build_dir", help="Building directory (and output)")
@@ -256,7 +260,8 @@ if __name__ == "__main__":
                         help="Specify configuration file with own list of exported into JS functions")
     parser.add_argument('--webnn', action="store_true", help="Enable WebNN Backend")
 
-    args = parser.parse_args()
+    transformed_args = ["--cmake_option=%s".format(arg) if arg[:2] == "-D" else arg for arg in sys.argv[1:]]
+    args = parser.parse_args(transformed_args)
 
     log.debug("Args: %s", args)
 
@@ -266,7 +271,7 @@ if __name__ == "__main__":
         del os.environ['EMMAKEN_JUST_CONFIGURE']  # avoid linker errors with NODERAWFS message then using 'emcmake' launcher
 
     if args.emscripten_dir is None:
-        log.error("Cannot get Emscripten path, please use 'emcmake' launcher or specify it either by EMSCRIPTEN environment variable or --emscripten_dir option.")
+        log.error("Cannot get Emscripten path, please use 'emcmake' launcher or specify it either by EMSCRIPTEN/EMSDK environment variable or --emscripten_dir option.")
         sys.exit(-1)
 
     builder = Builder(args)