Re: bison'd files don't compile under Microsoft VC++

[Hans Aberg]

At 11:20 -0500 2002/02/14, Florian Krohm wrote:
>You missed my point.

No, you missed my point:

> You cannot take a C program, compile it with a C++
>compiler and expect it to compile and/or run.

One can do this with a certain subset of C; bison.simple has been written
in this combined C/C++ subset.

> It was not meant to do so.

Which C++ of course was designed to be. But that was not the problem:

>If you do it nevertheless and encounter problems, the problems are yours not
>theirs.

The problem does not occur when using a C++ compiler to compile C in this
common subset, but when using new C++ (non-C) classes having non-trivial
constructors.

Those that encountered this problem were evidently some of the Bison
developers in their own C++ programs when switching to a better C++
compiler.

I did not encounter the problem, because I wrote my own C++ skeleton files.

>I agree that Bison should generate genuine C++ parsers. But they should have
>their own templates.

This was done instead of genuine C++ support because it requires less effort.

If somebody (you?) would have volunteered writing the C++ skeleton file, it
would not have been needed.

> Raping bison.simple is not The Right Thing IMHO.

I really don't know what you have in your mind here, as the changes did not
alter the performance under C. (The C compiler does not see them.)

  Hans Aberg