I just tested latest nightly build 1735
Now I have another issue with linker:
That time some required dependencies are omitted in index.html. But that
classes are declared with dependencies on themselves. Also, no circular
dependencies reported while using "-diagnostics=32768".
goog.addDependency('../../../oasis/gifts/WishGiftAddCommand.js',
['oasis.gifts.WishGiftAddCommand'],
['org.puremvc.as3.patterns.command.SimpleCommand',
'oasis.gifts.WishGiftAddCommand', 'oasis.gifts.WishGiftAddCommand',
'oasis.gifts.WishGiftAddCommand']);
Build output and warnings are available at the old location (with prefix
1735-):
https://drive.google.com/open?id=17IRbQ1xAfpAirjbW6YjQPFEYHmL1HnhC
On Thu, Dec 20, 2018 at 8:16 AM Alex Harui <[email protected]> wrote:
> FWIW, I just pushed changes that should allow this sort of configuration
> (even though, IMO, referencing subclasses from base classes is not a
> recommended practice).
>
>
>
> -Alex
>
>
>
> *From: *mrchnk <[email protected]>
> *Reply-To: *"[email protected]" <[email protected]>
> *Date: *Tuesday, December 18, 2018 at 1:26 PM
>
>
> *To: *"[email protected]" <[email protected]>
> *Subject: *Re: Linkage problem for asjs compiler
>
>
>
> I appreciate your help, Alex!
>
> Thanks. I pass linkage problems with that investigations.
>
>
>
> On Wed, Dec 19, 2018 at 12:21 AM Alex Harui <[email protected]> wrote:
>
> Hi,
>
>
>
> I saw that, but I think the compiler should be allowing that, so that’s
> what I’m going to look into next.
>
>
>
> JavaScript should be able to set up these classes and constructors without
> problem.
>
>
>
> -Alex
>
>
>
> *From: *mrchnk <[email protected]>
> *Reply-To: *"[email protected]" <[email protected]>
> *Date: *Tuesday, December 18, 2018 at 12:24 PM
>
>
> *To: *"[email protected]" <[email protected]>
> *Subject: *Re: Linkage problem for asjs compiler
>
>
>
> I found the issue for that case. The problem is in Animation constructor:
>
>
>
> ...
>
> if (_rootTimeline == null) {
>
> if (_rootFrame == -1) {
>
> _rootFrame = 0;
>
> _rootFramesTimeline = new SimpleTimeline();
>
> _rootTimeline = new SimpleTimeline();
>
> _rootTimeline._startTime = getTimer() / 1000;
>
> _rootFramesTimeline._startTime = 0;
>
> _rootTimeline._active = _rootFramesTimeline._active = true;
>
> ticker.addEventListener("enterFrame", _updateRoot, false, 0,
> true);
>
> } else {
>
> return;
>
> }
>
> }
>
> ...
>
>
>
> It tries to instantiate SimpleTimeout which is derived from Animation
> itself. That's tricky.
>
> After dealing with that I faced other circular dependencies:
>
>
>
> - using a derived class in base's method:
>
>
>
> class Base {
>
> function doSomething() {
>
> if (this is Derived) {
>
> ...
>
> }
>
> }
>
> }
>
> class Derived extends Base { ... }
>
>
>
> - one more setup with static initialized const
>
>
>
> class Facade {
>
> static public var api: Service = new Service();
>
> }
>
> class Service {
>
> function doSomething(forClient:Client) {
>
> forClient.done();
>
> }
>
> }
>
> class Client {
>
> function doSomething() {
>
> Facade.api.doSomething(this);
>
> }
>
> function done() {
>
> trace("done");
>
> }
>
> }
>
>
>
>
>
> On Tue, Dec 18, 2018 at 10:00 PM mrchnk <[email protected]> wrote:
>
> I think the problem with that classes that SimpleTimeline extends
> Animation, but Animation has fields (static too) with type SimpleTimeline.
>
>
>
> On Tue, Dec 18, 2018 at 9:54 PM mrchnk <[email protected]> wrote:
>
> Hi, Alex.
>
>
>
> I tried build #1713. Log output says that following circular dependency is
> found:
>
>
>
> ...
>
> Circular Dependency Found
>
> [com.greensock.core.Animation]
>
> [com.greensock.core.SimpleTimeline]
>
> [com.greensock.core.Animation]
>
> End of Circular Dependency
>
> ...
>
>
>
> Here are both .as and .js files mentioned in that dependency
>
> https://gist.github.com/mrchnk/caba399919690a785761242c35e5a01a
> <https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgist.github.com%2Fmrchnk%2Fcaba399919690a785761242c35e5a01a&data=02%7C01%7Caharui%40adobe.com%7C87fb5378894e44e98ffe08d6652f895f%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636807652177799218&sdata=1uFJ6ddxUr9ohlGz%2FQbdPtG%2FFtv18BhyP5QgaoiwKe4%3D&reserved=0>
>
>
>
> Build output and warnings are available at the old location (with prefix
> 1713-):
>
> https://drive.google.com/open?id=17IRbQ1xAfpAirjbW6YjQPFEYHmL1HnhC
> <https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdrive.google.com%2Fopen%3Fid%3D17IRbQ1xAfpAirjbW6YjQPFEYHmL1HnhC&data=02%7C01%7Caharui%40adobe.com%7C87fb5378894e44e98ffe08d6652f895f%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636807652177809223&sdata=93r73QVuxkIZF6k7jYjLMH1mxx3iAsunorjxmaiGv3k%3D&reserved=0>
>
>
>
> On Tue, Dec 18, 2018 at 11:43 AM Alex Harui <[email protected]> wrote:
>
> Hi,
>
>
>
> That log showed that lots of files are not in the sorted order, not just
> Client. In a test of the Tour De Flex running in Royale, all files 700+
> files end up in the sorted order. In trying to reproduce the issue, I
> discovered that circularities in the graph are not reported, so I’ve added
> that to the output. So wait a few hours for the new nightly (build #1712
> or later) and try again and search the output for
>
>
>
> Circular Dependency Found
>
>
>
> Hopefully there will be one which would explain the differences in the
> lists. It should be followed by a list of classes, then:
>
>
>
>
>
> End of Circular Dependency
>
>
>
> Post the list and the ActionScript sources and JavaScript from the output
> folder for the classes. The JS output may not handle certain kinds of
> static initializers and think there are static dependencies when there
> doesn’t need to be which could result in circularities that would work fine
> in SWF. We’ll see what it takes to make the compiler smarter, but you
> might make more progress by modifying the static initializers.
>
>
>
> One pattern for a workaround is to test if the class has been initialized
> and if not, initialize it. So instead of:
>
>
>
> public class SomeClass {
>
> public static var foo:int = SomeOtherClass.foo;
>
> }
>
>
>
> You could do:
>
> public class SomeClass {
>
> public static var initialized:Boolean;
>
> public static var foo:int;;
>
> public static function initialize() {
>
> SomeClass.foo = SomeOtherClass.foo;
>
> SomeClass.initialized = true;
>
> }
>
> }
>
>
>
> And then before you actually read SomeClass.foo, add
>
>
>
> if (!SomeClass.initialized) SomeClass.initialize();
>
>
>
> But first, we have to see if there is a circularity being reported.
>
>
>
> HTH,
>
> -Alex
>
>
>
> *From: *mrchnk <[email protected]>
> *Reply-To: *"[email protected]" <[email protected]>
> *Date: *Monday, December 17, 2018 at 9:48 AM
>
>
> *To: *"[email protected]" <[email protected]>
> *Subject: *Re: Linkage problem for asjs compiler
>
>
>
> Hi, Alex!
>
>
>
> I downloaded the latest nightly build of royale (which is 1707) and build
> my project using it with "-diagnostics=32768" argument.
>
> Here is build output (too large to be posted as a gist):
>
> https://drive.google.com/open?id=17IRbQ1xAfpAirjbW6YjQPFEYHmL1HnhC
> <https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdrive.google.com%2Fopen%3Fid%3D17IRbQ1xAfpAirjbW6YjQPFEYHmL1HnhC&data=02%7C01%7Caharui%40adobe.com%7C87fb5378894e44e98ffe08d6652f895f%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636807652177819233&sdata=5uGo4YmE83PWjJz5EG1EeCCK63wSiRD7eMtsTsW9OrU%3D&reserved=0>
>
>
>
> I also enabled warnings for that build and save error output as a
> separated log file.
>
>
>
> That project is large and has a long history, so it could be really bad at
> dependencies with spaghetti code.
>
> But that seems to be ok for a mxmlc compiler from AIR SDK.
>
>
>
> On Mon, Dec 17, 2018 at 10:49 AM Alex Harui <[email protected]> wrote:
>
> Hi,,
>
>
>
> After more investigation, I’m not sure what is going wrong. I tested some
> scenarios with all dependencies removed and they worked. I added more
> diagnostics to the compiler. It can take several hours for the nightly
> build to update, but when it does, please add -diagnostics=32768 to your
> build and post a link to the output.
>
>
>
> Check the nightly build status here:
> http://apacheroyaleci.westus2.cloudapp.azure.com:8080/job/royale-asjs/
> <https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fapacheroyaleci.westus2.cloudapp.azure.com%3A8080%2Fjob%2Froyale-asjs%2F&data=02%7C01%7Caharui%40adobe.com%7C87fb5378894e44e98ffe08d6652f895f%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636807652177819233&sdata=E%2BCSA2mLffOZrFfs2AzuhCt7fvNgX7tgRTPRMB91%2Bzo%3D&reserved=0>
> The updated compiler should be in build #1705 or later.
>
>
>
> Thanks,
>
> -Alex
>
>
>
> *From: *Alex Harui <[email protected]>
> *Reply-To: *"[email protected]" <[email protected]>
> *Date: *Friday, December 14, 2018 at 11:32 PM
>
>
> *To: *"[email protected]" <[email protected]>
> *Subject: *Re: Linkage problem for asjs compiler
>
>
>
> Hi,
>
>
>
> I think I found the problem. A class that had all of its dependencies
> removed in the circulars calculation (except its base class) wasn’t being
> output to the addDependency list because the base class dependency wasn’t
> being added into the addDependency list. I won’t have much time to work on
> it until Sunday evening.
>
>
>
> I did notice that openfl.swc does not contain externs/typedefs. That
> might be an issue for minification some day.
>
>
>
> -Alex
>
>
>
>
>
> *From: *mrchnk <[email protected]>
> *Reply-To: *"[email protected]" <[email protected]>
> *Date: *Friday, December 14, 2018 at 11:06 AM
> *To: *"[email protected]" <[email protected]>
> *Subject: *Re: Linkage problem for asjs compiler
>
>
>
> Alex, here are files you requested:
>
>
>
> https://gist.github.com/mrchnk/c0965fdfef6caf0632acc78d1138940c
> <https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgist.github.com%2Fmrchnk%2Fc0965fdfef6caf0632acc78d1138940c&data=02%7C01%7Caharui%40adobe.com%7C87fb5378894e44e98ffe08d6652f895f%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636807652177829238&sdata=nmV%2FjDB8orGH3mc27%2BWTLPtHmO5LRvlC%2FyamGZk%2F9%2BQ%3D&reserved=0>
>
>
>
> I used npm distribution of openfl version 8.7.0. You can download it into
> the current directory using npm or yarn package manager:
>
>
>
> npm install [email protected]
>
>
>
> After that command, you can find openfl.swc, openfl.min.js as well as
> non-minified version openfl.js at path node_modules/openfl/dist of the
> current directory.
>
>
>
> Also I used npm script "asconfigc --debug=true" to build the code
>
>
>
> I tried to disable "remove-circulars" option (thanks to Harbs for pointing
> me to that). It still compiles, and linkage problems seem to goes away. But
> I figured out that circular dependencies cannot be resolved properly by the
> closure library after all. Is there any common solution to that problem?
>
>
>
> On Fri, Dec 14, 2018 at 8:56 PM Alex Harui <[email protected]> wrote:
>
> In addition to the data I asked for earlier, please also post link to
> openfl.swc and openfl.min.js.
>
>
>
> Thanks,
>
> -Alex
>
>
>
> *From: *mrchnk <[email protected]>
> *Reply-To: *"[email protected]" <[email protected]>
> *Date: *Friday, December 14, 2018 at 8:22 AM
> *To: *"[email protected]" <[email protected]>
> *Subject: *Re: Linkage problem for asjs compiler
>
>
>
> I don't know anything about the option "remove-circulars", so I think I
> use the default value here, which is enabled I guess.
>
> Actually, I am using asconfigc wrapper over the compiler, but I have
> nothing about that option in my asconfig.json neither.
>
>
>
> On Fri, Dec 14, 2018 at 2:39 PM Harbs <[email protected]> wrote:
>
> Are you using the remove-circulars compiler option?
>
>
>
> On Dec 14, 2018, at 10:24 AM, Alex Harui <[email protected]> wrote:
>
>
>
> OK.
>
>
>
> Please post a link to the entire output of the compiler including, if
> possible, the command-line you used. I won’t have time to look at it right
> away as I am at the end of my work day.
>
>
>
> Also post links to Client.as, Client.js, OasisClient.as, OasisClient.js,
> OasisHTML5Launcher.as, OasisHTML5Launcher.js
>
>
>
> And also verify how the compiler “should” find Client.js. Is Client.as in
> the source-path or is Client.js in a SWC?
>
>
>
> Thanks,
>
> -Alex
>
>
>
> *From: *mrchnk <[email protected]>
> *Reply-To: *"[email protected]" <[email protected]>
> *Date: *Thursday, December 13, 2018 at 10:24 PM
> *To: *"[email protected]" <[email protected]>
> *Subject: *Re: Linkage problem for asjs compiler
>
>
>
> Same problem with the nightly build. I also asked about my issue in openfl
> community.
>
> I think that problem may be related to circular dependencies in my code.
>
> Is there any way that can be handled automatically by royale compiler?
>
>
>
> On Thu, Dec 13, 2018 at 11:23 PM Alex Harui <[email protected]> wrote:
>
> Hi,
>
>
>
> Try a nightly build of Royale. I think some things have been fixed in the
> calculation of dependencies. I think if you are using npm, you can
> uninstall royale and then:
>
>
>
> npm install
> http://apacheroyaleci.westus2.cloudapp.azure.com:8080/job/royale-asjs/lastSuccessfulBuild/artifact/out/apache-royale-0.9.6-bin-js.tar.gz
> <https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fapacheroyaleci.westus2.cloudapp.azure.com%3A8080%2Fjob%2Froyale-asjs%2FlastSuccessfulBuild%2Fartifact%2Fout%2Fapache-royale-0.9.6-bin-js.tar.gz&data=02%7C01%7Caharui%40adobe.com%7C87fb5378894e44e98ffe08d6652f895f%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636807652177829238&sdata=sc7A22ZvevfcYbCJOSxEEzJflgQcr6hGOD2%2F8wgWESY%3D&reserved=0>
>
>
>
> No guarantees that will fix your issue, but at least we can eliminate some
> possibilities.
>
>
>
> HTH,
>
> -Alex
>
>
>
> *From: *mrchnk <[email protected]>
> *Reply-To: *"[email protected]" <[email protected]>
> *Date: *Thursday, December 13, 2018 at 10:34 AM
> *To: *"[email protected]" <[email protected]>
> *Subject: *Linkage problem for asjs compiler
>
>
>
> Greetings!
>
>
>
> I tried to compile a large project using npm distribution of Apache Royale
> asjs compiler (@apache-royale/[email protected]). I changed all flash.*
> classes to openfl.*. After several hotfixes that project starts to compile.
>
>
>
> Here is index.html generated:
> https://gist.github.com/mrchnk/4f468d1abb1b02466a0dad43265c4333
> <https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgist.github.com%2Fmrchnk%2F4f468d1abb1b02466a0dad43265c4333&data=02%7C01%7Caharui%40adobe.com%7C87fb5378894e44e98ffe08d6652f895f%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636807652177839244&sdata=%2B8kkO4RPaCFY%2FNp8d0aoo%2B7dNUoD%2FK%2B%2FCreWkkREaek%3D&reserved=0>
>
>
>
> The page is loaded with javascript errors:
>
> Uncaught Error: Undefined nameToPath for Client
>
> at visitNode (base.js:1356)
>
> at visitNode (base.js:1354)
>
> at Object.goog.writeScripts_ (base.js:1368)
>
> at Object.goog.require (base.js:705)
>
> at (index):1681
>
>
>
> (index):1687 Uncaught ReferenceError: OasisHTML5Launcher is not defined
>
> at (index):1687
>
>
>
> So Client.as is the superclass of OasisClient.as and is used in main
> OasisHTML5Launcher.
>
> It is successfully compiled to Client.js and declared as a dependency at
> index.html
>
> But a link to js file and declaration for that class is missing (I mean
> "goog.addDependency('../../../Client.as' ... )")
>
>
>
> So how to deal with that? Is it a Royale bug? Are there some guidelines I
> missed about my code?
>
>
>
> Thanks in advance.
>
>
>
> PS: I'm kinda new to mailing lists. So don't hesitate to correct me if I
> do something wrong.
>
>
>
>
