Re: egcs-1.0.1 on Solaris, HP, and OSF/1

[Jeffrey A Law <law at cygnus dot com>]

  In message <199802050146.RAA27716@paradigm.webvision.com>you write:
  > Hi.
  > 
  > I've recently upgraded to gcc-2.8.0 and egcs-1.0.1 on my various
  > systems, but a large project I've been maintaining with gcc-2.7.2.2 no
  > longer compiles except on Solaris.  (HP-UX 10.20 and OSF/1 4.0B
  > complain in similar -- or perhaps identical -- areas; I'm still trying
  > to figure out what's going on.)
  > 
  > Anyway, I'm getting warnings detected in stl_list.h, stl_alloc.h,
  > stl_construct.h, and stl_tree.h about variables that might be used
  > uninitialized, or clobbered by vfork or longjmp:
  > 
  >     /usr/local/packages/egcs-1.0.1/include/g++/stl_list.h:155: warning: `st
  > ruct __list_node<const char *> * p' might be used uninitialized in this fun
  > ction
Do you get this with egcs?  It is the case that p might actually be uninitialized?

We spent a fair amount of time before the egcs-1.0 release fixing all the
problems with the flow graph to avoid false positive warnings of this nature.

As to the specific questions you asked -- egcs should be a superset of what's
available in gcc-2.8 as far as C++ conformance is concerned.

jeff