home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: gnu.g++.bug
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!cis.ohio-state.edu!cygnus.com!mrs
- From: mrs@cygnus.com (Mike Stump)
- Subject: Re: [dave@hiauly1.hia.nrc.ca: gcc-2.1]
- Message-ID: <9212212148.AA15079@cygnus.com>
- Sender: gnulists@ai.mit.edu
- Organization: GNUs Not Usenet
- Distribution: gnu
- Date: Mon, 21 Dec 1992 21:48:58 GMT
- Approved: bug-g++@prep.ai.mit.edu
- Lines: 54
-
- These problems has been fixed, and should not exist in the next
- release of gcc.
-
- > Date: Sun, 3 May 92 17:19:23 -0400
- > From: John David Anglin <dave@hiauly1.hia.nrc.ca>
-
- > I have been trying to get groff-1.05 and libg++-2.0 going on hp9000s730
- > running hpux 8.07. I am enclosing a context diff for gcc-2.1 which resolve
- > several problems:
-
- > 1) Changes need to be made to collect2.c to make vprintf argument types
- > correct for hpux. Further, collect2 must be compiled with CC=cc so
- > that system varargs is used rather than gcc varargs. The va_list types
- > are incompatible. This incompatibility prevents use of hpux varargs
- > routines with gcc. It also causes problems in building libg++ because
- > the struct type used with gcc can't be cast to a pointer type.
-
- Has been fixed (varargs are compatable now).
-
- > 2) ON_EXIT defined for hpux runs destructors on program termination.
-
- Fixed.
-
- > 3) Bug fix to assemble_external fixes undefined reference problem in
- > #pragma interface code (inline routines) and makes the virtual function
- > table global. This is needed for filebuf initialization in libg++.
-
- Fixed.
-
- > There are several other problems in the c++ mode which I am aware of but
- > have no solution:
-
- > 4) -O generates bad code and should not be used at the moment.
-
- Fixed.
-
- > 5) There are problems with const. E.g., "extern const struct a;"
- > generates a code space reference. Also, don't try to multiply a
- > long const with a long.
-
- Fixed.
-
- > 6) Function arguments which in turn call other functions are not
- > always handled properly. In one dump I examined, the primary
- > function was called before its arguments were evaluated. The
- > argument evaluation then trashed the stack. In other cases,
- > improperly aligned offsets are generated for storage of
- > temporary variables on the stack. Several cases of this problem
- > exist in the gtroff program. I eventually got gtroff to work by
- > explicitly defining intermediate variables wherever I
- > encountered a problem.
-
- Fixed.
-
-