I read the Firefox OS is gaining support by cell phone manufactures and distributors (e.g. "Deutsche Telekom").

While Android support seems to be a hot topic here, what about the upcoming "Firefox OS".

I suppose this is Java-based in a similar way as Android, and maybe a unified support for both can be provided.

-Michael

As far as I know, it is Javascript. Not Java.

I really should get my Pascal-to-Javascript translator project on track :(

Michael.

Michael Van Canneyt wrote

As far as I know, it is Javascript. Not Java.

As Firefox is a very fast Javascript interpreter this does make sense.

But it supposedly makes it completely incompatible with Android from ground up (this might be on purpose).

But can an "OS" completely done in Jacascript be a valid alternate foundation for cell phone and Tablet designs ?

-Michael

Michael Van Canneyt wrote

I really should get my Pascal-to-Javascript translator project on track :(

How far has this advanced yet ?

-Michael

Michael Schnell wrote

On 07/15/2013 10:12 AM, Michael Van Canneyt wrote: > I really should get my Pascal-to-Javascript translator project on track :( How far has this advanced yet ?

I have the AST trees ready, it's just a matter of writing the conversion step between the 2.

Nothing hard, but needs time. As all things.

Michael.

Michael Van Canneyt wrote

but needs time. As all things.

Understood;-) .

(Great project nonetheless...)

-Michael

Michael Schnell wrote

I read the Firefox OS is gaining support by cell phone manufactures and distributors (e.g. "Deutsche Telekom").

There is also Bada and WebOS ... in my opinion there are too many mobile OS, for me personally I think it is best to simply forget that mobile OS's with less than 5% market share is not worth the time to study them.

A lot of people who should know better think that the GUI is the OS. As far as I know, Firefox OS is Linux with all GUI interfacing done by HTML5 hence JavaScript etc., from my limited reading I don't know whether there's a native-code capability which might mean that as a platform it's even more restrictive than Android.

Felipe Monteiro de Carvalho wrote

There is also Bada and WebOS ...

and Windows 8 phone, and ...

But besides iOS and Android, I only see Firefox OS as a candidate for decent growth.

Lets wait and see...

-Michael

Why would this make it incompatible? If you're interested, I'd suggest doing some research before posting this kind of stuff. Firefox OS used to run on top of Android and AFAIK they intend to have Android compatibility as well...

That is why I put quotes around "OS". In fact iOS is BSD Unix in a similar disguise.

But besides the GUI API these "OS"es provide API definitions for phone stuff handling and other hardware/system resources.

With Android these API definitions are done by Java objects, No idea how this is done with iOS and Firefox OS.

As far as I know, Firefox OS is Linux

As is Android.

with all GUI interfacing done by HTML5 hence JavaScript etc.,

I recon that you are right here. I gather that HTML5 shows a rather nice (and ubiquitous) ways to do a GUI API for user programs. Maybe this in fact is an advantage over Android.

from my limited reading I don't know whether there's a native-code capability which might mean that as a platform it's even more restrictive than Android.

I read that any propriety stuff is supposed to be avoided in that project. So it should be rather easy to open the system for this.

OTOH, Michael's Pascal-to-Javascript translator project comes handy in that case.

-Michael

And Sailfish and Tizen...
(Is Bada still alive BTW?)

Agreed.
However, a lot of these smaller players all focus on Javascript+HTML5+OS-specific APIs/extensions.
As the first 2 parts of that stack are always the same, targeting that part may be a bit more worthwile.

However, as long as there is overwhelming market share for iOS+Android, supporting those does seem to be the easiest way.

I think we're all writing from a position of limited practical experience, but I think the major issue is whether Firefox OS allows arbitrary binaries "backdoor" access to GUI facilities.

If a binary app can't get to the GUI, if there's no provision for binary (or Java) applets, and if there's no way of injecting even a transient bytecode binary then it might be that translating to Javascript plus Android-style custom-drawn controls is the only option.

OTOH, Michael's Pascal-to-Javascript translator project comes handy in that case.

But do we really want to go there? The FPC and Lazarus projects are stretched enough as it is, without having to get involved in Javascript language issues.

Reinier Olislagers wrote

However, as long as there is overwhelming market share for iOS+Android, supporting those does seem to be the easiest way.

Yep, but - as always - being prepared is not a bad thing.

In fact with Android having a market share of some 70 %, it was a bad decision of Embarcadero's to support only iOS in the mobile world.

-Michael

Delphi XE5 - scheduled for release in september is supposed to contain Android support.

Michael.

Michael Van Canneyt wrote

Delphi XE5 - scheduled for release in september is supposed to contain Android support.

They already delayed this several times, so I don't hold my breath.

In fact this would be the first support for Linux after Kylix death. (Of course without VCL compatibility.) In fact, this done, they rather easily should be able to support PC Linux and non-mobile ARM systems.

-Michael

Since Firefox (and i assume Firefox OS) uses the new "asm.js" subset of javascript (a js subset that looks like assembly which can be converted to native code, something like a js-based portable LLVM bitcode), couldn't FPC generate asm.js code instead of x86 assembly/machine code?

Obviously that wouldn't solve the GUI issue (one will still need to implement a HTML-based (or whatever) LCL backend, but i think with canvas nowadays that is possible).

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20130715/6c83053f/attachment.html>

Or perhaps even FPC=>LLVM backend=>emscripten=>javascript but perhaps that's a couple of conversions too many...

Reinier Olislagers wrote

Firefox OS used to run on top of Android

I don't think that this is the way it is meant to be installed on devices by - say - Deutsche Telekom. Why would they wont to install anything "on top of Android", when they have Android anyway ?

and AFAIK they intend to have Android compatibility as well...

I don' see how an API done in Java can be compatible with an API done in JavaScript. (But it was not me who suggested that Firefox OS is completely JavaScript, I did not check the facts here...)

-Michael

Interesting stuff, indeed.

Obviously that wouldn't solve the GUI issue (one will still need to implement a HTML-based (or whatever) LCL backend, but i think with canvas nowadays that is possible).

This in fact is a completely different issue.

In fact, having support for a HTML5 backend (aka "Widget Type"), with the normal Lazarus LCL form designer for designing the the user interface, would allow for nice stuff in the non-mobile world as well.

-Michael

No, it's for testing beta releases only. The complete release will run on a phone without android/Lunix underneath, of course.

That's another testbed, running the browser as an emulator on android. Early access, attracting programmers and user, ...

Regarding news about "Intelligence Agencies" around the world Firefox OS will for sure have it's cusomers. It completely open source, in contrast to Android where parts are closed and Windwos as usual.

And it would give the opportunity to test apps that will run on Firefox OS later on ... although the non-HTML5-stuff, system apis and the like, would have to be emulated from the outside (of the widget, in pascal).

Marc Santhoff wrote

Regarding news about "Intelligence Agencies" around the world Firefox OS will for sure have it's cusomers.

Especially in Europe and even more in Germany.

(awaiting the news that the said Agency succeeded in turning down Mozilla org. )

-Michael

Yes, very. Perhaps comparable with the JVM backend.

> Obviously that wouldn't solve the GUI issue (one will still need to > implement a HTML-based (or whatever) LCL backend, but i think with > canvas nowadays that is possible). This in fact is a completely different issue.

Perhaps comparable with the custom-drawn Android LCL.

Mark Morgan Lloyd wrote

Perhaps comparable with the custom-drawn Android LCL.

That is what I was thinking of.

-Michael

Michael Schnell wrote

While Android support seems to be a hot topic here,

If that is *hot*, I think the level for firefox OS is proportionately already.

A JS backend would be something that some developers would indeed like (me included).

Regards,
Sven

I would keep out LLVM and emscripten and just directly generate asm.js code which is then simply compatible to the libraries provided by the emscripten project.

Regards,
Sven

I would keep out LLVM and emscripten and just directly generate asm.js code which is then simply compatible to the libraries provided by the emscripten project.

I do agree with Sven,
the way to go is http://goldparser.org/
the same technique has been used for project llvm_pascal http://code.google.com/p/llvm-pascal/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20130715/5d0e8280/attachment.html>

Why would I want to use a parser generator if we have a fully working multi-backend compiler available?!

Regards,
Sven

asm.js does seem to have a lot going for it. So I suppose a fair question is to what extent custom-drawn LCLs for Android and Firefox OS could share code, and whether this abstraction could be useful for other targets e.g. framebuffer or "naked" X.

this is fair :-) and I get your answer as a confirmation that there is a chance one day to discover with pleasure a target for asm.js -------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20130716/e20eca45/attachment.html>

No one has started support for it yet, but who knows what the future will bring...

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20130716/b59a640c/attachment.html>

http://asmjs.org/spec/latest/#introduction looks really interesting as a target arch for fpc.

AFAIU (but I might be wrong) the Android GUI API is based on HTML4, while the FirefoxOS GUI API is based on HTML5. I understand that HTML5 is a lot more versatile when it you want do display LCL visual objects in a Browser including providing GUI events that trigger Pascal events via an event queue (see EXTPASCAL / EXTJS for HTML4 plus Java script based stuff, that might be used via Comet based HTTP server functionality and often seems to ask for some trouble).

Moreover a future move of Android's towards HTML5 would be rather logical (keep you crystal ball at hand).

-Michael

Sven Barth wrote

I would keep out LLVM and emscripten and just directly generate asm.js code which is then simply compatible to the libraries provided by the emscripten project.

That of course is advantageous, as it can be done in Pascal, as the fpc is used to and happy with.

If there already is a decent LLVM backend for asm.js (which I doubt because asm.js itself is a work in progress) this might be a trigger for rethinking.

-Michael

Sven Barth wrote

Why would I want to use a parser generator if we have a fully working multi-backend compiler available?!

+1

Either stay with the current (excellent) Parser, or switch to a major standard such as LLVM or gcc !

-Michael

Michael Schnell wrote

http://asmjs.org/spec/latest/#introduction looks really interesting as a target arch for fpc.

Should we switch to the fpc-devel mailing list on that issue ?

-Michael

I think it is a Java GUI. HTML does not enter the picture to my knowledge.

Michael.

Ehm... you do know that the Android GUI is not based on any HTML? It's a Java based system...

Regards,
Sven

I think an important issue is whether asm.js can be translated to something which can either be executed from a shell, or whether there's a runtime environment (e.g. a Firefox extension) that provides its own shell plus standard utilities including make etc.

If asm.js can only be translated to something that will run as a GUI applet in the context of Firefox, it's going to be very difficult to run the FPC test programs that (in principle at least) demonstrate the compiler's correctness.

It's just a subset of JavaScript. In the browser you normally have a HTML page that references it, but the examples of emscripten (which can also generate asm.js compatible code) show how you can execute it on the command line as well.

Regards,
Sven

Sven Barth wrote

show how you can execute it on the command line as well.

Do you start Firefox, a part of same (or whatever) with some command line arguments to do this ?

-Michael

Sven Barth wrote

Ehm... you do know that the Android GUI is not based on any HTML? It's a Java based system...

While I learned this only after I phrased this message, OTOH HTML5 also is a Java based system and thus the thought might not be _completely_ wrong.

-Michael

"Michael Schnell" wrote

On 07/16/2013 12:11 PM, Sven Barth wrote: > > Ehm... you do know that the Android GUI is not based on any HTML? It's a

Java based system...

While I learned this only after I phrased this message, OTOH HTML5 also

is a Java based system and thus the thought might not be _completely_ wrong.

What the?! Michael, please inform yourself before writing such things. HTML5 has *nothing* to do with Java. It uses JavaScript and besides the C-like syntax those two have nothing in common!

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20130716/0fc9a1e1/attachment.html>

No, HTML5 is not Java based.
Sounds like this thread is just a place for "experts" to... etc.

arguments to do this ?

If you want a GUI you'll need to use a browser AFAIK, but otherwise a JS command line interpreter is sufficient.

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20130716/f4da49fb/attachment.html>

Where do you get your information ?

You could not be further off the mark. HTML 5 has nothing to do with Java.

Michael.

Sven Barth wrote

but otherwise a JS command line interpreter is sufficient.

That was what I tried to ask. I never heard of any "JS command line interpreter". Is there one provided by Mozilla ?

-Michael

OK, I did find Rhino. https://developer.mozilla.org/en-US/docs/Rhino

-Michael

Sven Barth wrote

HTML5 has *nothing* to do with Java.

Sorry for again typing faster than thinking. :-(

-Michael

Michael Van Canneyt wrote

Where do you get your information ?

I was just being confused and apologize for wasted bandwidth :-( .

-Michael

So can anybody provide step-by-step instructions for writing a "Hello, World!" in Javascript, compiling it to asm.js, and running that from a (Linux) shell? Even better, converting the asm.js to native code so it can be run without a wrapper?

I Do know that confusing Java and Java-Script is similar to confusing a HedgeHog and a Pig, But things like that happen in my advanced age :-( .

-Michael

So can anybody provide step-by-step instructions for writing a "Hello, World!" in Javascript, compiling it to asm.js, and running that from a (Linux) shell? Even better, converting the asm.js to native code so it can be run without a wrapper?

asm.js is just a subset of JavaScript. It runs on current browsers as is. Specs: http://asmjs.org/spec/latest/

R.

Dunno about Rhino, but AFAIK the Emscripten developers use Google's V8 which has a command-line version.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20130716/014ff781/attachment.html>

Which is no *&%%&$ use to somebody trying to run e.g. the FPC test suite under makefile control.

Hello Mark,

So can anybody provide step-by-step instructions for writing a "Hello,

It looks like you are motivated enough about this feature, as I do have common interest on this feature,
so let's try to discuss about it,

the step required are the following,
pascal source ode converter to javascript source code, this is why I was talking about the Gold Parser which is not really a parser but more a translator,
anyway, since you may have knowledge and skills to use the Code DOM of FPC, you may be able to generate appropriate javascript source code respecting the asm.js constraints,

then you will need http://code.google.com/p/fpcjs/ this wrapper will help you to encapsulate the OdinMonkey library (this is the actual asm.js compiler),

you will need to retrieve the OdinMonkey library from the latest FireFox Beta build,
located on my computer in
C:\Program Files (x86)\Nightly
I guess the main library entry points are located in mozjs.dll so the wrapper fpcjs may need some search and replace replace,

finally you should be able to follow the sample on the main project page: http://code.google.com/p/fpcjs/

if you do obtain any kind of result,
I would like to be updated about it,

Regards, Jerome
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20130716/60a9209e/attachment.html>

>> So can anybody provide step-by-step instructions for writing a "Hello, >> World!" in Javascript, compiling it to asm.js, and running that from a >> (Linux) shell?

Nothing gets compiled from JS to asm.js (do you know what "strict subset" means?). Did you even look at the specs? Do you know at all what asm.js is?

Which is no *&%%&$ use to somebody trying to run e.g. the FPC test suite under makefile control.

That's a high goal for someone who cannot be bothered to inform himself what the subject of his desire (asm.js) is. Btw I don't know what *&%%&$ means but judging from your professional attitude it must be some kind of profanity.

Ever tried to do your work yourself instead of begging for step-by-step instructions? Speaking from experience I can tell you it is very fulfilling.

R.

Yes, and it rather looks as though you don't. asm.js has strictly numeric types, so you can't write a conventional "Hello, World!" in it. If you want to do that, then you have to write it in a suitable high-level language: I suggested Javascript, but I can see how that could be confusing so let's just accept /anything/ here.

Now if something like that can't, ultimately, be run from a shell or a makefile it's going to be problematic for an FPC port. It's going to be particularly problematic for the various tests that are run automatically to test the compiler's correctness: it's all very well saying that in the target environment FPC -> asm.js is used strictly to run apps in the context of Firefox but that's hardly going to be convenient for working through a large corpus of test cases.

Yes. I'll say no more lest I let irritation get the better of diplomacy.

[...]asm.js has strictly numeric types,

yes

so you can't write a conventional "Hello, World!" in it.

There are javascript libraries implementing strings via arrays.

When I google for "asm.js command line tool" then the first hit is the SpiderMonkey engine. Do you mean that?

[...]

Mattias

Yes, but asm.js is usually used as an intermediate language akin to Java bytecodes. cf Emscripten etc.

As I understand it, SpiderMonkey is for embedding Javascript in a (GUI-based) program, rather than for executing a Javascript or asm.js program from the command line (or from a makefile etc.) and getting text output.

For example, this (via Google)
https://github.com/benjcooley/asm-js-firefox-test/blob/master/compile-hello-world.sh shows a "Hello, World" in C++ being compiled using Emscripten, but the target is an HTML file rather than something that could be run from a shell or makefile.

[Later] Something like Rhino Shell might do the job. Possibly others.

Asm.js is 100% compatible with normal JavaScript - the only difference is that it is slower. But for testing this shouldn't be a problem and you can use any JavaScript interpreter. Google V8's source code has a command-line version (which Emscripten uses or used at some point for tests too) that you can build (it is very easy, i did it at the past when i was messing with Emscripten). Here is some information:
http://stackoverflow.com/questions/1802478/running-v8-javascript-engine-standalone

You will obviously need some sort of "interface" between the asm.js function and the outside world for tests (f.e. what "writeln" would do?) but you'd need that interface anyway since even for real asm.js use people will need to provide such interfaces
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20130717/1cd2f400/attachment.html>

Nevertheless you are free to call normal JS code without problems. asm.js is a subset of JS after all. How do you think the emscripten libraries are implemented?
asm.js can run on any JS interpreter that supports a certain JS standard. The only difference is that interpreters with special support for asm.js can optimize that code more than usual JS code. Also CPUs don't know strings either, nevertheless you can see them on your screen. It's all about interpreting the memory in the correct way and writing to the correct location.

Please read the tutorial of emscripten as it explains how to run a JS/asm.js program on the command line:
https://github.com/kripken/emscripten/wiki/Tutorial

Regards,
Sven

Jy V wrote

I guess the main library entry points are located in mozjs.dll so the wrapper fpcjs may need some search and replace replace,

Ah, this is what I guessed: The command-line java script interpreter does use parts of the normal Firefox distribution.

-Michael

Sven Barth wrote

> [Later] Something like Rhino Shell might do the job. Possibly others.

Such as ejscript.

Please read the tutorial of emscripten as it explains how to run a JS/asm.js program on the command line: https://github.com/kripken/emscripten/wiki/Tutorial

Looks pretty good. So it should, in principle, be possible not only to have FPC target asm.js, but to run the test cases etc.

Yes, the most complicated part should be writing the backend at all...

Regards,
Sven

You ruined my fun, Sven. Shame on you.
That's what I wanted the very polite Mark to find out for himself. All hints were there. And now you just told him how the 'magic' works. :( At least this thread can hopefully die now.

R.