[M3devel] Think we need a new release. C target

[Daniel Alejandro Benavides D.]

Hi all:
Interestingly Java had (along ago) UNSAFE extensions into the language:

http://groups.google.com/group/microsoft.public.dotnet.general/browse_thread/thread/bdbaaafa49579931/d9a25710e7e1cdc4%3Fq%3D%2522David%2BChase%2522%23d9a25710e7e1cdc4&ei=iGwTS6eaOpW8Qpmqic0O&sa=t&ct=res&cd=49&source=groups&usg=AFQjCNGAFKXr0Lh9Zhp47J8i8nnVUMePyw

I insist we should get back the Original test suite of SUN, this would help so much the Modula-3 RT (remember this guys worked hard in Modula-3 until the compiler got unsupported or at least until they thought it wouldn't get that much support).

Thanks in advance


--- El vie, 17/2/12, Daniel Alejandro Benavides D. <dabenavidesd at yahoo.es> escribió:

De: Daniel Alejandro Benavides D. <dabenavidesd at yahoo.es>
Asunto: Re: [M3devel] Think we need a new release. C target
Para: "Dragiša Durić" <dragisha at m3w.org>
CC: "m3devel" <m3devel at elegosoft.com>, "Jay" <jay.krell at cornell.edu>
Fecha: viernes, 17 de febrero, 2012 07:26

Hi all:
I don't think that's the case, Java has a JNI and it can work, but CM JVM is just a lot safer and easier to use, direct access to C code of the RT and you know what is better than that out there, in case of an attack? Modula-3 + Java seems the combination to win for me, ESC for both, I mean and what else do you need?

The point of what is Modula-3 a Systems Programming Language doesn't change that it is easier to deal with UNSAFE code, still we could add a keyword for say PROVED to allow the back-end check only for correct optimizations on it or to check UNSAFE code allowed to do so in either Java or Modula-3 or C. I don't care too much about PROVED or UNSAFE modules but more normal code and that's where Modula-3 could be helped by JVM dynamic verification (that said, it is not the best thing to do because this requires a lot of smart checking,
 I can give you an example of a program in Java that still manages to corrupt any class file in the cd, fooling the JVM) without JVM doing that much. That said compiler verification is much harder than normal checking, still there is research to allow one to do so.
Thanks in advance

--- El vie, 17/2/12, Dragiša Durić <dragisha at m3w.org> escribió:

De: Dragiša Durić <dragisha at m3w.org>
Asunto: Re: [M3devel] Think we need a new release. C target
Para: "Daniel Alejandro Benavides D." <dabenavidesd at yahoo.es>
CC: "Jay" <jay.krell at cornell.edu>, "m3devel" <m3devel at elegosoft.com>
Fecha: viernes, 17 de febrero, 2012 06:29

To port to JVM or Javascript, you have to throw through the window a lot of what Modula-3 is. You will get, in best case, part of
 Modula-3.
On the other side, targeting to C (or C++) and losing object model from sight (while debugging), ie losing or distorting, also looks like an horrible side effect to me.
It looks like the best direction to concentrate effort is current GCC (a lot of platforms) and LLVM ((almost) new kid on the block with many good promises). The best thing about LLVM target is - IM is standardized and fully documented. Since we all know what pain is tagging along behind GCC IM (thanks to RMS losing licensing battle to SRC), LLVM looks like a promise of future freedom for Modula-3. Maye some day we will not be traumatized by every major (and most minor) GCC releases.
BTW, freepascal has it's own backend infrastructure. Maybe worth a try.
dd

On Feb 15, 2012, at 8:57 PM, Daniel Alejandro Benavides D. wrote:
Hi all:
The point is whether we want to migrate our current RT to C or JavaScript, my question is why not (Java/) JVM or Obliq.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://m3lists.elegosoft.com/pipermail/m3devel/attachments/20120218/4a4c9d5c/attachment-0002.html>