std::auto_ptr bugs

std::auto_ptr bugs

[Michael Roth]

Hi,

First, the std::auto_ptr is commented out in "memory" (egcs 1.0.*)
and their is a note that the class is still discussed by the C++
standardization committee. Is this really true ?

Second, if you uncomment std::auto_ptr their are several errors:

1.) 'auto_ptr& operator=(const auto_ptr& a)'  misses a 'return *this'
    after the closing bracket of the 'if' block.

2.) 'auto_ptr& operator=(const auto_ptr<T>& a)' misses the 
    'return *this' after the closing bracket of the 'if' block, too.

3.) IMHO 'X* get() const' is wrong. If I'm not false 'get()' should get
    the ownership of the pointer back ('owns=true;' is missing) !?!?

4.) The line 'X* release const __STL_NOTHROW { owns = false; return ptr }'
    is buggy. It misses a '()' after the word 'release'.



cu
		Michael



      o-------------------------------------------------------o
      ! arago,                          Michael Roth          !
      ! Institut fuer komplexes         Software Engineer     !
      ! Datenmanagement GmbH            eMail: mroth@arago.de !
      ! Fichtestr. 12                                         !  
      ! 60316 Frankfurt am Main         http://www.arago.de   !
      ! Tel: +49-69-40568-0             Fax: +49-69-40568-111 !
      o-------------------------------------------------------o

Re: std::auto_ptr bugs

[Joe Buck]

> First, the std::auto_ptr is commented out in "memory" (egcs 1.0.*)
> and their is a note that the class is still discussed by the C++
> standardization committee. Is this really true ?

It's no longer true ... the standardization committee made a final
decision, at the very end of '97 just in time for the vote (this final
decision is not the same as the CD2 draft that you can find everywhere on
the web, unfortunately).  I haven't checked to see if the commented-out
version matches that decision or not.

But you're right, there are multiple errors in the commented-out version.
That file comes unmodified from SGI's STL, so we need to pass bug reports
on to them.

Re: std::auto_ptr bugs

[Thomas Hoyt]

Joe Buck wrote:
>
> > First, the std::auto_ptr is commented out in "memory" (egcs 1.0.*)
> > and their is a note that the class is still discussed by the C++
> > standardization committee. Is this really true ?
>
> It's no longer true ... the standardization committee made a final
> decision, at the very end of '97 just in time for the vote (this final
> decision is not the same as the CD2 draft that you can find everywhere on
> the web, unfortunately).  I haven't checked to see if the commented-out
> version matches that decision or not.
>
> But you're right, there are multiple errors in the commented-out version.
> That file comes unmodified from SGI's STL, so we need to pass bug reports
> on to them.
 
Does anyone know if SGI is still actively working on their released STL?
Updates were pretty regular until the last one in October.  Is there any
sign of conformance to the FDIS soon?
 
--Thomas Hoyt

-- 
"My God, it's written in COBOL..."