egcs-971016/libstc++/stl/hashtable patch (revised)

egcs-971016/libstc++/stl/hashtable patch (revised)

[Greg Galloway]

Fri Oct 17 11:58:18 EDT 1997  Gregory L. Galloway <gregg@eoeml@gtri.gatech.edu>

	* hashtable.h(__stl_prime_list): Added literal constant suffixes to 
	suppress warnings about integer constant which is too large.
	(new_node): Moved return stmt outside try block to suppress warning.

*** libstdc++/stl/hashtable.h.orig	Fri Oct 17 11:51:14 1997
--- libstdc++/stl/hashtable.h	Fri Oct 17 11:52:42 1997
***************
*** 182,193 ****
  static const int __stl_num_primes = 28;
  static const unsigned long __stl_prime_list[__stl_num_primes] =
  {
!   53,         97,         193,       389,       769,
!   1543,       3079,       6151,      12289,     24593,
!   49157,      98317,      196613,    393241,    786433,
!   1572869,    3145739,    6291469,   12582917,  25165843,
!   50331653,   100663319,  201326611, 402653189, 805306457, 
!   1610612741, 3221225473, 4294967291
  };
  
  inline unsigned long __stl_next_prime(unsigned long n)
--- 182,193 ----
  static const int __stl_num_primes = 28;
  static const unsigned long __stl_prime_list[__stl_num_primes] =
  {
!   53ul,         97ul,         193ul,       389ul,       769ul,
!   1543ul,       3079ul,       6151ul,      12289ul,     24593ul,
!   49157ul,      98317ul,      196613ul,    393241ul,    786433ul,
!   1572869ul,    3145739ul,    6291469ul,   12582917ul,  25165843ul,
!   50331653ul,   100663319ul,  201326611ul, 402653189ul, 805306457ul, 
!   1610612741ul, 3221225473ul, 4294967291ul
  };
  
  inline unsigned long __stl_next_prime(unsigned long n)
***************
*** 526,532 ****
      try {
  #       endif /* __STL_USE_EXCEPTIONS */
        construct(&n->val, obj);
-       return n;
  #       ifdef __STL_USE_EXCEPTIONS
      }
      catch(...) {
--- 526,531 ----
***************
*** 534,539 ****
--- 533,539 ----
        throw;
      }
  #       endif /* __STL_USE_EXCEPTIONS */
+     return n;
    }
    
    void delete_node(node* n)

Re: egcs-971016/libstc++/stl/hashtable patch (revised)

[Joe Buck]

> 	* hashtable.h(__stl_prime_list): Added literal constant suffixes to 
> 	suppress warnings about integer constant which is too large.
> 	(new_node): Moved return stmt outside try block to suppress warning.

We aren't maintaining the STL files; SGI is.  I think we'll have to send
bugs and patches to them, unless we want to fork (do we?).  They've
generally been fairly cooperative.

Re: egcs-971016/libstc++/stl/hashtable patch (revised)

[Jeffrey A Law]

  In message < 199710172339.QAA22589@atrus.synopsys.com >you write:
  > 
  > > 	* hashtable.h(__stl_prime_list): Added literal constant suffixes to 
  > > 	suppress warnings about integer constant which is too large.
  > > 	(new_node): Moved return stmt outside try block to suppress warning.
  > 
  > We aren't maintaining the STL files; SGI is.  I think we'll have to send
  > bugs and patches to them, unless we want to fork (do we?).  They've
  > generally been fairly cooperative.
We don't want to diverge unless there's a compelling reason.

I don't know C++ well enough to know if anything Greg sent
represents anything serious enough to diverge.

I'm going to forward Greg's stuff to Ulrich just in case Ulrich has
any opinions (unfortunately he doesn't read egcs :(

jeff

Re: egcs-971016/libstc++/stl/hashtable patch (revised)

[Joe Buck]

>   > We aren't maintaining the STL files; SGI is.  I think we'll have to send
>   > bugs and patches to them, unless we want to fork (do we?).  They've
>   > generally been fairly cooperative.
> We don't want to diverge unless there's a compelling reason.
> 
> I don't know C++ well enough to know if anything Greg sent
> represents anything serious enough to diverge.

At least some of Greg's patches seem to be papering over the flow analysis
problems in exception code (moving return statements out of try blocks
and the like).  The existing code is valid.  So in that sense those
patches wouldn't be needed if the compiler were fixed.

Some of the other patches represent missing forwarding headers (<iostream>
and the like), headers that Ulrich said he was working on but that I
haven't seen yet (does he have a private copy of new work somewhere?).

Re: egcs-971016/libstc++/stl/hashtable patch (revised)

[Oleg Krivosheev]

hi,

On Fri, 17 Oct 1997, Jeffrey A Law wrote:

> We don't want to diverge unless there's a compelling reason.
> 
> I don't know C++ well enough to know if anything Greg sent
> represents anything serious enough to diverge.
> 
> I'm going to forward Greg's stuff to Ulrich just in case Ulrich has
> any opinions (unfortunately he doesn't read egcs :(

well, if SGI stopped working on STL, it might be 
compeling enough reason to move forward...

As far as i understand Greg patches mostly prevents
anoying warnings so it's probably not enough to split.

From the other hand i was bitten again by 
new lookup rules in egcs - it rightfully refises 
to instantiate some template friends. IMHO this is
SGI STL bug(s). I had some patches but the question remains
who is going to fix STL.

regards

OK