Digital Mars - D - D alpha compiler

↑ ↓ ← → Russ Lewis <spamhole-2001-07-16 deming-os.org> writes:

Walter wrote:

 This is mainly to help along discussion of language features and to see what
 fundamental issues I've overlooked!

 ftp://ftp.digitalmars.com/dmdalpha.zip

 Enjoy.

 -Walter


Would you like us to post bugs (other than not-implemented-features listed in
the docs)?

--
The Villagers are Online! http://villagersonline.com

.[ (the fox.(quick,brown)) jumped.over(the dog.lazy) ]
.[ (a version.of(English).(precise.more)) is(possible) ]
?[ you want.to(help(develop(it))) ]

Dec 08 2001

↑ ↓ ← → "Walter" <walter digitalmars.com> writes:

"Russ Lewis" <spamhole-2001-07-16 deming-os.org> wrote in message
news:3C12991D.67140D22 deming-os.org...
 Walter wrote:

 This is mainly to help along discussion of language features and to see




 fundamental issues I've overlooked!

 ftp://ftp.digitalmars.com/dmdalpha.zip

 Enjoy.

 -Walter


 Would you like us to post bugs (other than not-implemented-features listed


 the docs)?


I'm absolutely interested in any and all bug reports. Post away, there's
nothing secret here! As for the bugs listed in the doc, I'm interested in
what priorities people have for them. I need to get it whipped into shape
before my DDJ article on D gets published!

-Walter

Dec 08 2001

↑ ↓ ← → Russ Lewis <spamhole-2001-07-16 deming-os.org> writes:

Bug 0001: Trailing (superfluous) semicolons cause syntax errors.

I know, I know, all the style purists hate me for this...but somewhere, early
when I learned C++, I got into the habit of adding superfluous semicolons after
my functions and other blocks:

void foo()
{
    ....
};



It's ugly (to most people's eyes), but legal (I think)

--
The Villagers are Online! http://villagersonline.com

.[ (the fox.(quick,brown)) jumped.over(the dog.lazy) ]
.[ (a version.of(English).(precise.more)) is(possible) ]
?[ you want.to(help(develop(it))) ]

Dec 08 2001

↑ ↓ ← → "Pavel Minayev" <evilone omen.ru> writes:

"Russ Lewis" <spamhole-2001-07-16 deming-os.org> wrote in message
news:3C12B1C5.47154D71 deming-os.org...
 Bug 0001: Trailing (superfluous) semicolons cause syntax errors.


 I know, I know, all the style purists hate me for this...but somewhere,


 when I learned C++, I got into the habit of adding superfluous semicolons


 my functions and other blocks:

 void foo()
 {
     ....
 };



 It's ugly (to most people's eyes), but legal (I think)


Not sure. I might be wrong, but AFAIK VC++ doesn't allow such
things. Borland compilers do. Is it in ANSI C++ or not, I
wonder?

Not sure I'd like to see it be legal in D, anyhow.

Dec 09 2001

↑ ↓ ← → Russ Lewis <spamhole-2001-07-16 deming-os.org> writes:

I use it all over my MSVC code.

Frankly, any superfluous semicolons should be treated by the compiler as null
statements.  I know that that is true in code, I can't swear that it's accurate
in function body declarations...but I think that it's true there, too.

--
The Villagers are Online! http://villagersonline.com

.[ (the fox.(quick,brown)) jumped.over(the dog.lazy) ]
.[ (a version.of(English).(precise.more)) is(possible) ]
?[ you want.to(help(develop(it))) ]

Dec 09 2001

↑ ↓ ← → "Walter" <walter digitalmars.com> writes:

Ok, I'll fix it to allow empty declarations, mainly to be symmetric with
allowing empty statements. -Walter

"Russ Lewis" <spamhole-2001-07-16 deming-os.org> wrote in message
news:3C12B1C5.47154D71 deming-os.org...
 Bug 0001: Trailing (superfluous) semicolons cause syntax errors.

 I know, I know, all the style purists hate me for this...but somewhere,


 when I learned C++, I got into the habit of adding superfluous semicolons


 my functions and other blocks:

 void foo()
 {
     ....
 };
 It's ugly (to most people's eyes), but legal (I think)

Dec 09 2001

↑ ↓ ← → Russ Lewis <spamhole-2001-07-16 deming-os.org> writes:

Bug 0002: Link problems

I compiled the following source file, wanting to test out how the GC works:

        import stdio;

        class Foo
        {
          static this() { printf("%p ",this); }
          ~this () { printf("~%p ",this); }
        }

        int main()
        {
          {
            Foo ptr = new Foo;
            printf("x");
          }
          printf("x");
          long i;
          for(i=0; i<1000*1000*1000; i++);
          printf("x");
          return 0;
        }

My makefile was as follows:

        russ.exe: russ.obj
                link ..\lib\phobos.lib ..\lib\snn.lib russ.obj

        russ.obj: russ.d
                mars russ.d -I..\src\phobos

I get the following errors on the link:

        link ..\lib\phobos.lib ..\lib\snn.lib russ.obj
        OPTLINK (R) for Win32  Release 7.50B1
        Copyright (C) Digital Mars 1989 - 2001  All Rights Reserved

        ..\lib\snn.lib(stream)  Offset 3CFB1H Record Type 00C3
         Error 1: Previous Definition Different : ?WS  YAAAVistream  AAV1  Z
(istream &c
        decl WS(istream &))
        ..\lib\snn.lib(_wfopen)  Offset 51C1DH Record Type 00C3
         Error 1: Previous Definition Different : __WFOPEN
        ..\lib\snn.lib(wconstar)
         Error 42: Symbol Undefined _wmain
        ..\lib\snn.lib(wwinstar)
         Error 42: Symbol Undefined _wWinMain 16
        ..\lib\snn.lib(winstart)
         Error 42: Symbol Undefined _WinMain 16

        --- errorlevel 5

I have never used DM compilers before, so I downloaded dm823c.zip, extracted it,
dmdalpha.zip and extracted it, and then copied all the files in the dmd/
directory over into the corresponding ones in dm/

--
The Villagers are Online! http://villagersonline.com

.[ (the fox.(quick,brown)) jumped.over(the dog.lazy) ]
.[ (a version.of(English).(precise.more)) is(possible) ]
?[ you want.to(help(develop(it))) ]

Dec 08 2001

↑ ↓ ← → "Pavel Minayev" <evilone omen.ru> writes:

"Russ Lewis" <spamhole-2001-07-16 deming-os.org> wrote in message
news:3C12B2C8.949F65F3 deming-os.org...

 I get the following errors on the link:

         link ..\lib\phobos.lib ..\lib\snn.lib russ.obj
         OPTLINK (R) for Win32  Release 7.50B1
         Copyright (C) Digital Mars 1989 - 2001  All Rights Reserved

         ..\lib\snn.lib(stream)  Offset 3CFB1H Record Type 00C3
          Error 1: Previous Definition Different :




Use a different command to link OBJs:

    sc russ.obj

Works fine for me. I had modified the program a little:

    import stdio;

    class Foo
    {
     this() { printf("%p created\n", this); }
     ~this() { printf("%p destroyed\n", this); }
    }

    int main()
    {
     {
      Foo ptr = new Foo;
      printf("ptr %p out of scope\n", ptr);
      ptr = cast(Foo) null; // no effect
     }
     printf("waiting for gc...\n");
     for (long i = 0; i < 1000*1000; i++);
     printf("done\n");
     return 0;
    }

Now when I run it, Foo gets created but is never destroyed,
even with 1000000000 iterations of loop! Shouldn't the GC
run at the end of the program anyhow?

Also note one weird thing. null is a void*, and is incompatible
with any other pointers, requiring explicit cast! This should
definitely be changed...

Dec 09 2001

↑ ↓ ← → "Walter" <walter digitalmars.com> writes:

"Pavel Minayev" <evilone omen.ru> wrote in message
news:9uvcm8$2dng$1 digitaldaemon.com...
 Now when I run it, Foo gets created but is never destroyed,
 even with 1000000000 iterations of loop! Shouldn't the GC
 run at the end of the program anyhow?


It does now. <g>

 Also note one weird thing. null is a void*, and is incompatible
 with any other pointers, requiring explicit cast! This should
 definitely be changed...


Hmm. I've seen different positions on this!

Dec 09 2001

↑ ↓ ← → "Walter" <walter digitalmars.com> writes:

"Russ Lewis" <spamhole-2001-07-16 deming-os.org> wrote in message
news:3C12B2C8.949F65F3 deming-os.org...
 Bug 0002: Link problems

 I compiled the following source file, wanting to test out how the GC


         import stdio;

         class Foo
         {
           static this() { printf("%p ",this); }


The static is an error, it is now diagnosed.

           ~this () { printf("~%p ",this); }
         }

         int main()
         {
           {
             Foo ptr = new Foo;
             printf("x");
           }
           printf("x");
           long i;
           for(i=0; i<1000*1000*1000; i++);


Empty loops should be {}, this is now diagnosed.

           printf("x");
           return 0;
         }

 My makefile was as follows:

         russ.exe: russ.obj
                 link ..\lib\phobos.lib ..\lib\snn.lib russ.obj


This command will link in the entire libraries, instead of just searching
them. Try:
    sc russ.obj
to link.

Dec 09 2001

↑ ↓ ← → Russ Lewis <spamhole-2001-07-16 deming-os.org> writes:

Walter wrote:

 The static is an error, it is now diagnosed.


Yeah, that seemed odd, but I found that syntax in the sample a.d.  So I guess
it's an error there :)

 Empty loops should be {}, this is now diagnosed.


(slap forehead) Yeah, that was a requirement, wasn't it?

 This command will link in the entire libraries, instead of just searching
 them. Try:
     sc russ.obj
 to link.


Thanks!

--
The Villagers are Online! http://villagersonline.com

.[ (the fox.(quick,brown)) jumped.over(the dog.lazy) ]
.[ (a version.of(English).(precise.more)) is(possible) ]
?[ you want.to(help(develop(it))) ]

Dec 09 2001

↑ ↓ ← → Russ Lewis <spamhole-2001-07-16 deming-os.org> writes:

I tried to use the compiler to link in a .lib file with something I was
compiling, since I've seen that done on gcc.  I don't know if the DM d compiler
supports that sort of thing, but instead of reporting an error I get an "Illegal
Operation" window with the following line:

    mars russ.d -I..\src\phobos ..\lib\phobos.lib

--
The Villagers are Online! http://villagersonline.com

.[ (the fox.(quick,brown)) jumped.over(the dog.lazy) ]
.[ (a version.of(English).(precise.more)) is(possible) ]
?[ you want.to(help(develop(it))) ]

Dec 08 2001

↑ ↓ ← → "Pavel Minayev" <evilone omen.ru> writes:

"Russ Lewis" <spamhole-2001-07-16 deming-os.org> wrote in message
news:3C12B418.BF0C42F3 deming-os.org...

 I tried to use the compiler to link in a .lib file with something I was
 compiling, since I've seen that done on gcc.  I don't know if the DM d


 supports that sort of thing, but instead of reporting an error I get an


 Operation" window with the following line:

     mars russ.d -I..\src\phobos ..\lib\phobos.lib


It might have just treated the LIB file as a D program... probably
not the best idea. Anyhow, you should do it when linking:

    mars russ.d -I..\src\phobos
    sc russ.obj phobos.lib

Dec 09 2001

↑ ↓ ← → "Walter" <walter digitalmars.com> writes:

"Russ Lewis" <spamhole-2001-07-16 deming-os.org> wrote in message
news:3C12B418.BF0C42F3 deming-os.org...
 I tried to use the compiler to link in a .lib file with something I was
 compiling, since I've seen that done on gcc.  I don't know if the DM d


 supports that sort of thing, but instead of reporting an error I get an


 Operation" window with the following line:

     mars russ.d -I..\src\phobos ..\lib\phobos.lib


The mars executable doesn't handle the link step, you need to link
separately using sc:

    sc russ.obj

should do it.

Dec 09 2001

↑ ↓ ← → "Pavel Minayev" <evilone omen.ru> writes:

Program consists of two files:

    // mod1.d
    import mod2;

    char[] name()
    {
     return "EvilOne";
    }

    int main(char[][] args)
    {
     greet();
    }


    // mod2.d
    import mod1;
    import stdio;

    void greet()
    {
     printf(name());
    }


First, I've compiled it as follows (repeating -I stuff omitted):

    > mars mod1.d mod2.d
    mod1.d(10): undefined identifier greet

OK, now vice-versa:

    > mars mod2.d mod1.d
    mod2.d(7): undefined identifier name

Then I decided to compile them separately:

    > mars mod1.d
!!!Works!!!
    > mars mod2.d
    Internal error: ..\ztc\cgcs.c 348

Here, the order doesn't matter. mod2 always fails to
compile. Wassup?

Dec 09 2001

↑ ↓ ← → "Walter" <walter digitalmars.com> writes:

Ok, these are fixed. -Walter

"Pavel Minayev" <evilone omen.ru> wrote in message
news:9uvct6$2dq0$1 digitaldaemon.com...
 Now even more funny. I've made all identifiers in mod1
 fully qualified:

     // mod1.d, line 10
     mod2.greet();

 Compiling:

     > mars mod1.d mod2.d
     mod1.d(10): undefined identifier
 Expression::toCBuffer('import').greet.greet

 ?WHAT THE?

Dec 09 2001

↑ ↓ ← → Axel Kittenberger <axel dtone.org> writes:

Walter wrote:

 This is mainly to help along discussion of language features and to see
 what fundamental issues I've overlooked!
 
 ftp://ftp.digitalmars.com/dmdalpha.zip
 
 Enjoy.


Buck! The license in the download that seems to cover now this 
project doesn't look like something I like .... 

- Axel

Dec 09 2001

↑ ↓ ← → "Walter" <walter digitalmars.com> writes:

While I intend to do some sort of open source license in the future, for the
moment I think it's best that it be given some time to gel before multiple
incompatible versions appear <g>. A colleague of mine is investigating what
it would take to hook it up to GCC's back end.

"Axel Kittenberger" <axel dtone.org> wrote in message
news:9v0gf1$qt$1 digitaldaemon.com...
 Walter wrote:

 This is mainly to help along discussion of language features and to see
 what fundamental issues I've overlooked!

 ftp://ftp.digitalmars.com/dmdalpha.zip

 Enjoy.


 Buck! The license in the download that seems to cover now this
 project doesn't look like something I like ....

 - Axel

Dec 09 2001

↑ ↓ ← → a <a b.c> writes:

Walter wrote:
 
 While I intend to do some sort of open source license in the future, for the
 moment I think it's best that it be given some time to gel before multiple
 incompatible versions appear <g>. A colleague of mine is investigating what
 it would take to hook it up to GCC's back end.


	You'll have to let us know how that goes.  Frankly, I'm not in the mood
to buy, install and setup windows in a semi-comfortable manner just to
test an alpha compiler.

Dan

Dec 09 2001

↑ ↓ ← → Russ Lewis <spamhole-2001-07-16 deming-os.org> writes:

a wrote:

 You'll have to let us know how that goes.  Frankly, I'm not in the mood
 to buy, install and setup windows in a semi-comfortable manner just to
 test an alpha compiler.


I've got it copied over to my Linux box to try it out with Wine.  Trouble is,
I've
never used Wine before...and I don't have it configured properly yet.

Speaking of which...is mars.exe specifically a Windows executable, or might it
be
worth trying in the Linux DOS Emulator (dosemu).

Ofc, Walter, I don't expect you to support running your compiler under
Wine...but I
am very curious about whether it might work or not :)

--
The Villagers are Online! http://villagersonline.com

.[ (the fox.(quick,brown)) jumped.over(the dog.lazy) ]
.[ (a version.of(English).(precise.more)) is(possible) ]
?[ you want.to(help(develop(it))) ]

Dec 09 2001

↑ ↓ ← → "Walter" <walter digitalmars.com> writes:

It's specifically a Win32 executable, not a DOS one. It doesn't push the API
very hard, so if Wine works at all, then mars.exe should <g>.


"Russ Lewis" <spamhole-2001-07-16 deming-os.org> wrote in message
news:3C1429BB.ED2E6D27 deming-os.org...
 Speaking of which...is mars.exe specifically a Windows executable, or


 worth trying in the Linux DOS Emulator (dosemu).

 Ofc, Walter, I don't expect you to support running your compiler under


 am very curious about whether it might work or not :)

Dec 09 2001

↑ ↓ ← → Christof Meerwald <cmeerw web.de> writes:

On Sun, 09 Dec 2001 20:19:23 -0700, Russ Lewis wrote:
 Ofc, Walter, I don't expect you to support running your compiler under
Wine...but I
 am very curious about whether it might work or not :)


It does work with Wine (at least with Wine-20011108)

You only have to make sure that all filenames are in lowercase only (if you
want to install DM on a case-sensitive filesystem like ext2)

And you have to use "--" on the wine command line to make wine stop parsing
command line options, e.g. use

  wine -- c:/dmd/bin/mars -o sieve.d

instead of 

  wine c:/dmd/bin/mars -o sieve.d


bye, Christof

-- 
http://cmeerw.cjb.net                             JID: cmeerw jabber.at
mailto cmeerw at web.de

...and what have you contributed to the Net?

Dec 15 2001

↑ ↓ ← → "Walter" <walter digitalmars.com> writes:

Cool!

"Christof Meerwald" <cmeerw web.de> wrote in message
news:9vfoo4$ms4$1 digitaldaemon.com...
 On Sun, 09 Dec 2001 20:19:23 -0700, Russ Lewis wrote:
 Ofc, Walter, I don't expect you to support running your compiler under




 am very curious about whether it might work or not :)


 It does work with Wine (at least with Wine-20011108)

 You only have to make sure that all filenames are in lowercase only (if


 want to install DM on a case-sensitive filesystem like ext2)

 And you have to use "--" on the wine command line to make wine stop


 command line options, e.g. use

   wine -- c:/dmd/bin/mars -o sieve.d

 instead of

   wine c:/dmd/bin/mars -o sieve.d


 bye, Christof

 --
 http://cmeerw.cjb.net                             JID: cmeerw jabber.at
 mailto cmeerw at web.de

 ...and what have you contributed to the Net?

Dec 15 2001

↑ ↓ ← → "Walter" <walter digitalmars.com> writes:

"a" <a b.c> wrote in message news:3C14164B.913EF689 b.c...
 Walter wrote:
 While I intend to do some sort of open source license in the future, for




 moment I think it's best that it be given some time to gel before




 incompatible versions appear <g>. A colleague of mine is investigating




 it would take to hook it up to GCC's back end.


 You'll have to let us know how that goes.  Frankly, I'm not in the mood
 to buy, install and setup windows in a semi-comfortable manner just to
 test an alpha compiler.



I wouldn't expect anyone to. -Walter

Dec 09 2001

↑ ↓ ← → Axel Kittenberger <axel dtone.org> writes:

Walter wrote:

 While I intend to do some sort of open source license in the future, for
 the moment I think it's best that it be given some time to gel before
 multiple incompatible versions appear <g>. A


I work now a while in the OpenSource Environment, there might be valid 
concerns I do understand (If you want to make money/finance yourself) but 
this particular one of multiple version is in practise baseless, never 
underestimate the power of the "official" distribution. Beeing on the other 
side I know how few chamces one has at the community maintainting an non 
official version of an OpenSource Package.

I would suggest distiebuting under GPL (also your libraries) or one of the 
very GPL-like licenses as this gives commercial competiton great drawbacks 
as they have to publicate their modifications, or globassy have to move one 
of their applications to GPL if they use your library. Then require from 
patch contributions a copyright assignment like the GPL does (and some 
other projects mainly backed by companies (meaning most of the devs work 
there)), this allows -you- to link the code against your libraries without 
having to OpenSoure them, allows you to sell superceded version parallell 
to the GPL version (ie QT) allows you to sell the GPL libraries with a non 
GPL license for people who want to pay for nat having to transiest to GPL 
(see sleepycat.com and their buisness model)


 colleague of mine is investigating what it would take to hook it up to
 GCC's back end.


GPL, or until your willing to workaround with files. However the FSF is 
doing pretty strong politics with gcc, and it has its reason why the GCC 
companents are interwoven with each other, and do no longer work on file 
interfacing, so they are able to force all frontends to be GPL or having to 
do pretty unpleasing workarounds (I do not yet know of any)

Dec 10 2001

↑ ↓ ← → Axel Kittenberger <axel dtone.org> writes:

 one of their applications to GPL if they use your library. Then require
 from patch contributions a copyright assignment like the GPL does (and


A painfull error, of course I mean ... assignment like the *FSF* does

Generally sorry for the aou mismatches I'm currently playing with Dvorak 
keyboard layout...

- Axel

Dec 10 2001

↑ ↓ ← → Russ Lewis <spamhole-2001-07-16 deming-os.org> writes:

It's kind of a triviality, but gcc's -MM (list includes) is very useful to me.
Having that work on the D compiler would be a high priority for me :)

--
The Villagers are Online! http://villagersonline.com

.[ (the fox.(quick,brown)) jumped.over(the dog.lazy) ]
.[ (a version.of(English).(precise.more)) is(possible) ]
?[ you want.to(help(develop(it))) ]

Dec 09 2001