C++ front-end / C++ compiler

C++ front-end / C++ compiler

[Christian Baxmann]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 887 bytes --]

Dear Ladies and Gentlemen,
we are trying to write a C++ compiler back-end for the 
GUPRO-Project (graphbased program understanding)
at the University of Koblenz,Germany. We try to avoid 
writing the front-end
We'd like to know if your C++ front-end fulfils our needs, or 
can be modified to fulfil them. Here's a description of what 
we need: 
A C++ front-end that
- generates AST and symbol table
- parses ANSI C++ (and dialects: MSVC++, GNU C++, Borland 
C++)
- generates full source code information (line and column 
numbers, type names etc.) that would make it possible to 
  reconstruct the source code with our back-end and 
your front-end
- runs on NT and Solaris
 
Further information about your front-end (price, 
support, known bugs, how to access the AST, available manuals and 
docs)
would be nice.
 
Yours
Christian Baxmann
 
baxmann@uni-koblenz.de

Re: C++ front-end / C++ compiler

[Martin v. Loewis]

> we are trying to write a C++ compiler back-end for the GUPRO-Project
> (graphbased program understanding) at the University of
> Koblenz,Germany.

Christian,

I don't know what the requirements for your back-end are, but why
can't you use g++ as-is?

> A C++ front-end that
> - generates AST and symbol table

Depends on what your requirements on the AST are. g++ generates tree
structures (see gcc/tree.def, and gcc/cp/cp-tree.def). Also, gcc 2.95
normally generates them on a per-statement basis, and then throws them
away (unless it is a template). The mainline compiler operates in
function-at-once mode, so it produces an AST for the entire function,
and compiles that.

> - parses ANSI C++ (and dialects: MSVC++, GNU C++, Borland C++)

With restrictions, yes.

> - generates full source code information (line and column numbers,
> type names etc.) that would make it possible to reconstruct the
> source code with our back-end and your front-end

g++ currently does not have column numbers, as that would require
integration with the preprocessor (otherwise, column numbers are
meaningless). Integration with cpplib (thus providing column numbers)
is in progress, but not complete.

> - runs on NT and Solaris

Yes.

> Further information about your front-end (price, support, known bugs,
> how to access the AST, available manuals and docs)
> would be nice.

Hmm. This is a boilerplate request, right? gcc is free software, and
free of charge. In the mainline compiler, gcc/cp/ir.texi explains the
internal representation.

Regards,
Martin

Problems trying to compile g++ under NT

[Christian Baxmann]

Hi there!
I'm trying to compile g++ 2.95.2 under Windows NT with VisualC++ 8.0.
Documentation says its possible, but
I got lots of problems.
All I really need is the g++-Front-End, 'cause I'm writing my own Back-End.

I tried to compile the whole g++, but it didn't work, 'cause I got lots of
compiler-error
(syntax-error for PVPROTO and PROTO in system.h and machmode.h) and a
header-file "tree-check.h" that
doesn't exist in the gcc-package.
Those PVPROTO and PROTO-stuff is from ansidecl.h and gansidecl.h, but those
say it's obsolete, 'cause
PARAMs are better/supported.
Is there a chance to avoid those PVPROTO,VPROTO an PROTO-stuff?

I ran those configure-Files under cygwin B20, ran config-nt.bat, tried to
use gcc2.95.2 itself to compile ... that
all didn't work

Is there any further information available about how to compile under NT,
especially what steps does it take, what compiler-options to use, which
files are really needed, compiling with g++-include-dir or the
M$-include-dir ...

Christian

Re: Problems trying to compile g++ under NT

[Martin v. Loewis]

> I'm trying to compile g++ 2.95.2 under Windows NT with VisualC++
> 8.0.  Documentation says its possible, but I got lots of problems.

Where did you read its possible? Compiling the C compiler might be
supported; compiling the C++ compiler is not - that is supposed to be
build with the C compiler generated in stage 1.

> Is there a chance to avoid those PVPROTO,VPROTO an PROTO-stuff?

What do you mean, avoid? You can certainly replace them using global
search-and-replace in your editor, provided you end up with ANSI C
signatures afterwards, and provided your bootstrap compiler supports
ANSI C. It would be better if you get them defined, though.

Alternatively, you could start with a gcc 2.96 snapshot, where these
are replaced with the PARAMS macros.

> Is there any further information available about how to compile under NT,
> especially what steps does it take, what compiler-options to use, which
> files are really needed, compiling with g++-include-dir or the
> M$-include-dir ...

Please have a look at

http://www.xraylith.wisc.edu/~khan/software/gnu-win32/

Regards,
Martin