Migrating dmd to D?

[Daniel Murphy]

"dennis luehring" <dl.soluz at gmx.net> wrote in message 
news:khku3t$15ja$1 at digitalmars.com...
> Am 11.03.2013 16:23, schrieb Daniel Murphy:
>> "dennis luehring" <dl.soluz at gmx.net> wrote in message
>> news:khkqug$v57$1 at digitalmars.com...
>>> Am 11.03.2013 15:20, schrieb Daniel Murphy:
>>>> "Daniel Murphy" <yebblies at nospamgmail.com> wrote in message
>>>> news:khfoa6$fm7$1 at digitalmars.com...
>>>>> "Daniel Murphy" <yebblies at nospamgmail.com> wrote in message
>>>>> news:kgumek$2tp4$1 at digitalmars.com...
>>>>>> "Zach the Mystic" <reachBUTMINUSTHISzach at gOOGLYmail.com> wrote in
>>>>>> message
>>>>>> news:pabfuaorrjbljxzrglbv at forum.dlang.org...
>>>>>>>>
>>>>>>>> Something like this https://github.com/yebblies/magicport2 ?
>>>>>>>
>>>>>>> Yes! I need to look it over more thoroughly, but I couldn't ask for 
>>>>>>> a
>>>>>>> better beginning. Can I trust that you'll be a willing part of 
>>>>>>> future
>>>>>>> discussions on this matter, even if only to play Devil's Advocate?
>>>>>>
>>>>>> More like a full-blown attempt than a beginning.  I started this a 
>>>>>> long
>>>>>> time ago.
>>>>>>
>>>>>> There are three parts to it:
>>>>>> - c++ parser/d printer, with lots of cheating and special cases
>>>>>> - patches to the c++ source
>>>>>> - patched version of dmd to build the result (no error on variable
>>>>>> shadowing etc)
>>>>>>
>>>>>> It produces a 70000 line d file which appears to get through 3/7ths 
>>>>>> of
>>>>>> semantic1.  Root needs to be ported, and a cleaner interface to the
>>>>>> backend is needed to compile the glue layer.
>>>>>>
>>>>>
>>>>> Update: With the bulk of root converting or ported, and the glue layer
>>>>> stubbed out, I can build dmd from the converted source then lex and
>>>>> parse
>>>>> the source (+druntime headers) again.
>>>>>
>>>>> The highlight was the dynamically resized struct in root/stringtable.
>>>>> Something went horribly wrong there.
>>>>
>>>> Update: I can now generate the source, then build a frontend from that,
>>>> then
>>>> process the frontend's source again with the built compiler.  It also
>>>> works
>>>> on the conversion tool, and pulls in a sizeable chunk of druntime and
>>>> phobos.
>>>
>>> do i get it right - you've converted the dmd C++ code with it?
>>>
>>
>> Umm...
>>
>> C++ compiler source -> my tool -> D source
>> D source -> normal dmd -> self-host dmd
>> D source -> self-host dmd -> no problems, but only the frontend so no 
>> code
>> generation
>> tool source -> self-host dmd -> same thing
>
> but interesting enough to get its own root newsgroup post i think - or
> it the "quality"(converted source etc. whatever) too bad
>
>

I'm planning to when it can do the entire test suite, and all of phobos and 
druntime.

The code generated is very close to what you would get running it though a 
(bad) formatter, with comments removed.  I will eventually preserve the 
comments and improve the formatting.

Performance wise the code is pretty nasty because I'm allocating all 
OutBuffers on the heap and inserting tracing code.  This will need to be 
fixed eventually but is fine for checking correctness.

Here's an example conversion: (from PrettyFuncInitExp)

Almost all of the differences are from the primitive pretty-printer.

--------------------------------------------------------
C++ version
--------------------------------------------------------

Expression *PrettyFuncInitExp::resolveLoc(Loc loc, Scope *sc)
{
    FuncDeclaration *fd;
    if (sc->callsc && sc->callsc->func)
        fd = sc->callsc->func;
    else
        fd = sc->func;

    const char *s;
    if (fd)
    {
        const char *funcStr = fd->Dsymbol::toPrettyChars();
        HdrGenState hgs;
        OutBuffer buf;
        functionToCBuffer2((TypeFunction *)fd->type, &buf, &hgs, 0, 
funcStr);
        buf.writebyte(0);
        s = (const char *)buf.extractData();
    }
    else
    {
        s = "";
    }

    Expression *e = new StringExp(loc, (char *)s);
    e = e->semantic(sc);
    e = e->castTo(sc, type);
    return e;
}


--------------------------------------------------------
D version
--------------------------------------------------------

    Expression resolveLoc(Loc loc, Scope sc)
    {
    tracein("resolveLoc");
    scope(success) traceout("resolveLoc");
    scope(failure) traceerr("resolveLoc");
    {
        FuncDeclaration fd;
        if ((sc.callsc && sc.callsc.func))
        (fd = sc.callsc.func);
         else (fd = sc.func);
        const(char)* s;
        if (fd)
        {
            const(char)* funcStr = fd.Dsymbol.toPrettyChars();
            HdrGenState hgs;
            OutBuffer buf = new OutBuffer();
            functionToCBuffer2((cast(TypeFunction)fd.type), buf, (&hgs), 0, 
funcStr);
            buf.writebyte(0);
            (s = (cast(const(char)*)buf.extractData()));
        }
         else {
            (s = "");
        }
        Expression e = (new StringExp(loc, (cast(char*)s)));
        (e = e.semantic(sc));
        (e = e.castTo(sc, type));
        return e;
    }
    }