Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!wuarchive!cs.utexas.edu!helios!cs.tamu.edu
From: jeffw@cs.tamu.edu (Jeffrey A Waller)
Newsgroups: comp.lang.c++
Subject: Re vectors and clobbering.  Hashing defect?
Message-ID: <6570@helios.TAMU.EDU>
Date: 14 Jul 90 07:42:25 GMT
Sender: usenet@helios.TAMU.EDU
Organization: Computer Science Department, Texas A&M University
Lines: 82

Three things:

#1

<1427@camex.COM> geoff@circus.camex.com (Geoffrey Knauth) responds to
article <140@emtek.UUCP>  matt@emtek.UUCP (Matt Meola):

Matt::

>>Does anyone know why cfront mangles the name catch_sig(int) into
>>catch_sig__FiPFve_v ?

Geoffrey::

>I believe the extra characters type-encode the function and provide it
>a signature.

How about this:

void catch_sig(int sig,void (*func)(...)) {/* stuff */}

catch_sig__FiPFve_v

catch_sig is a global function that takes an integer as first parameter
and a pointer to a function that takes void,... as parameters and returns
void.  Is there another catch_sig perhaps???  Or a call to catch_sig with
those kind of parameters--type safe linkage yea!

#2

<5039@uwm.edu> peter@csd4.csd.uwm.edu (Peter J Diaz de Leon) writes

Peter::
	
>When I try to compile it under Turbo C++ 1.0 I get the error:
>
> cannot find IntArray::IntArray() to initialize a vector in function main
>
>on the line in the main program which reads:
>
>         pia = new IntArray[1025];

I think I remember in the 1.2 Stroustrup book that to be an element of a 
vector, a class must have a constructor with no arguments.  
IntArray(int sz = ArraySize) should qualify, but apparently TC++ dosen't
recognize it, for a quick fix--maybe not much help, redefine
IntArray(int sz = ArraySize) as two constructors.  IntArray(int sz) and
IntArray(), in which sz is set to ArraySize, maybe that will work.

#3

A long time ago I posted about some trouble I was having with multiple
inheritance.  The crux of the problem arouse from cfront's generation of
of identifiers including __ptbl as part of the name.   These names referred
to functions.  I haven't read this news group much since then, so this
might have already been discussed at length, in which case, sorry.

Error: A linker error complaining about some multiply defined symbols
       each when run back through cfilt indicate a symbol such as
       classname::__ptbl.  No other indication of what this symbol is
       used for.  C code is generated by cfront 2.0.

The various files compile without warnings even with +w and +p switches.
The same code compiles and links fine on one installation of C++, but not
on another.  Both C++ compilers are the same version, in fact they were
derived from the same tape.  The difference in the installations (I think)
is that on one installation, the identifier length limit is set to 32.  On
the other (the one that works), the identifier length limit is set to 
256 (or some other big number).  Looking at the C code generated by cfront,
it looks like the identifier names have been hashed by the cfront generating
32 character identifiers.


MY QUESTION IS:

What is __ptbl used for anyway, and is it possible that a hashing collision
has occured to give the multiply defined symbols error.

						-Jeff Waller
        					jeffw@cssun.tamu.edu

P.S.  When are templates likely to be avaliable?