Java process does not exit when r.js is run with Rhino with config file and optimize="closure"

[akatkere]

Simple test case:

build.js: ({ appDir: "./src/", baseUrl: "./", dir: "./build/", modules: [ { name: "test", } ], removeCombined: true, optimize: "closure", })

src/test.js: { }

Run r.js with:

java -classpath ../lib/js.jar:../lib/compiler.jar org.mozilla.javascript.tools.shell.Main ../lib/r.js -o build.js

Tracing dependencies for: test Minifying file: /Users/akatkere/tmp/rjs-poc/test/build/test.js

test.js

test.js

[ does not exit]

Thread dump shows a couple of closure compiler threads that are not doing anything. Even if I optimize one module explicitly via command line, it still hangs:

java -classpath ../lib/js.jar:../lib/compiler.jar org.mozilla.javascript.tools.shell.Main ../lib/r.js -o baseUrl=./src/ name=test out=build/test-built.js optimize=closure

Tracing dependencies for: test Minifying file: /Users/akatkere/tmp/rjs-poc/test/build/test-built.js

/Users/akatkere/tmp/rjs-poc/test/build/test-built.js

/Users/akatkere/tmp/rjs-poc/test/src/test.js

[ does not exit]

I was able to get past the issue by explicitly passing a quit function to build:

diff --git a/dist/r.js b/dist/r.js
index 309ceaf..95ec849 100644
--- a/dist/r.js
+++ b/dist/r.js
@@ -24115,7 +24115,7 @@ define('build', function (require) {
      * This function does not return a status, it should throw an error if
      * there is a problem completing the build.
      */
-    build = function (args) {
+    build = function (args, quit) {
         var buildFile, cmdConfig, errorMsg, errorStack, stackMatch, errorTree,
             i, j, errorMod,
             stackRegExp = /( {4}at[^\n]+)\n/,
@@ -24145,7 +24145,7 @@ define('build', function (require) {
             }

             return build._run(cmdConfig);
-        }).then(null, function (e) {
+        }).then(function() { quit(0); }, function (e) {
             var err;

             errorMsg = e.toString();
@@ -25920,7 +25920,7 @@ define('build', function (require) {
                     quit(1);
                 };

-                build(config).then(done, done).then(callback, errback);
+                build(config, quit).then(done, done).then(callback, errback);
             };

             requirejs({
@@ -26006,7 +26006,7 @@ require({
     }
 },       ['env!env/args', 'env!env/quit', 'logger', 'build'],
 function (args, quit, logger, build) {
-    build(args).then(function () {}, function (err) {
+    build(args, quit).then(function () {}, function (err) {
         logger.error(err);
         quit(1);
     });

[jrburke]

I am not sure what is going on here, as you note, it only happens if closure compiler is used. If I use uglifyjs or set optimize: 'none', it is fine. With closure, it does not hang indefinitely, it eventually quits for me, but it does take a long time for it to happen. So I expect it to be an issue with closure compiler in some way.

Unfortunately I cannot use that diff as-is, as that function is designed to be called by code that may want to continue past that point, and not just quit.

So I am not sure how best to go about making this better. Unfortunately I do not have the java skills any more to dig into what is going under the covers with closure compiler. Maybe the interface changed and if I were to call it differently it would work better. I will be looking to someone in the community with Java experience to dig into the root cause.

As a data point for anyone looking into this: I tried on Mac OS X with the closure compiler in the r.js repo, with java -version of:

java version "1.6.0_65"
Java(TM) SE Runtime Environment (build 1.6.0_65-b14-462-11M4609)
Java HotSpot(TM) 64-Bit Server VM (build 20.65-b04-462, mixed mode)

[akatkere]

@jrburke, isn't there any other place where we know we are done optimizing where quit can be called? As far as I can see from the thread dump, closure compiler has a couple of non daemon threads running, and the process will not exit until they exit. I was looking into filing a defect against closure compiler, but the difficulty there is calling Java API is not a documented way of using closure compiler (as far as I can see). So a workaround on r.js since would be much simpler than trying to get the closure compiler side "fixed".

[jrburke]

The problem is that other code, like embedded use inside of other .js programs, requires that the code not exit at that point -- the build() API is just an API to do the build, other code decides when to quit.

It is also disconcerting that this is a specific issue with a specific minifier: uglifyjs and not minifying does not cause a hang. I am unsure of what fix would help, but quitting at the point in the diff above is not an option -- it would break existing code and uses of r.js.