| From: | Rudi Chiarito |
| Date: | 31 Aug 2000 at 16:23:13 |
| Subject: | Re: AMIOPEN: vpcc as a C++ compiler |
On Thu, Aug 31, 2000 at 10:38:00AM +0100, amimjf@connectfree.co.uk wrote:
> Well i ported jikes which is written totaly in c++, its not simple code
> either but seems (i am not c++ expert), to use all c++ facilities even
> templates, and it built ok appart from some stuff missing from stdlibc++.
I actually ported all of it, managing to make it all to compile (I
started with the latest release of the sources, 1.12, which date to Aug
1st). The compiler itself caused no problems. The only difficulties came
from stdlibc++, which was not completely tested, as Tao and their
customers use mainly VP and C. That and some problems with some wchar
library functions (which can be easily worked around).
I haven't uploaded the sources to CVS yet because a) I had no Internet
access during most of yesterday and I needed some archives for testing
b) the wchar problems were being studied c) I am creating an automated
procedure for the changes I am making, so that when a new release of the
sources appears, the changes don't need to be performed all manually
(this are actually changes in the Makefiles and in the configure
scripts/files, not in the actual .cpp files, which have to be done that
way because they depend on a large number of Unixisms which can't be
easily replicated under Elate).
> It did fall over on lclint with an internal error, which seems to have
> been some kind of interal stack overflow (or simmilar) because it
> happended on a really large 200k+ bison generated c++ file. But appart
> from that it seems quite stable.
Could you mail me that file?
As I said, the compiler itself is pretty much stable. Any problems
occurring with C++ are likely caused by libstdc++.