I've already sent Watlter the testcase that replicates the bug in simplest terms. Currently, it is not possible to compile STLport or use STLport in your applications, esp. if you use <string>.

Please do not barrage Walter with bug reports.

Details: DMC complains about _DEFAULT_SIZE not being defined. The test case looks like:

--------------------
template<class Tp, int>
class Parent {
public:
  enum { DefaultSize = 32 };

private:
  Tp  array[DefaultSize];

public:
  Parent(int size);
};

typedef Parent<char, DefaultSize> NewBug;
--------------------

What happened is DMC++ now implements dependent/nondependent name lookup. DefaultSize is not in scope at the declaration of NewBug according to my understanding of the Standard.

There were a couple of problems compiling STLPort with this new lookup, but a corrected STLPort has been posted.

I believe the compiler is operating correctly.

Walter Bright wrote...
> What happened is DMC++ now implements dependent/nondependent name lookup. DefaultSize is not in scope at the declaration of NewBug according to my understanding of the Standard.
> 
> There were a couple of problems compiling STLPort with this new lookup, but a corrected STLPort has been posted.
> 
> I believe the compiler is operating correctly.

The actual gimpel PC-Lint doesn't accept the code either:
                     _
  typedef Parent<char, DefaultSize> NewBug;
  bug.cpp  14  Error 40: Undeclared identifier 'DefaultSize'
  bug.cpp  14  Error 1048: expected a constant expression
  bug.cpp  14  Error 40: Undeclared identifier 'DefaultSize'
  bug.cpp  14  Error 1048: expected a constant expression

It's also my understanding that DefaultSize is not in scope here.


- Heinz

In article <dl1nqt$5fi$1@digitaldaemon.com>, Walter Bright says...
>
>There were a couple of problems compiling STLPort with this new lookup, but a corrected STLPort has been posted.

Where is the corrected STLPort? Is the corrected version on the STLPort website, the STLPort shipped with DMC++, or the STLPort 4.6.2 from Scott Michel's site?

Doug

"tink" <tink_member@pathlink.com> wrote in message news:dl7u2k$2fjl$1@digitaldaemon.com...
> In article <dl1nqt$5fi$1@digitaldaemon.com>, Walter Bright says...
> >
> >There were a couple of problems compiling STLPort with this new lookup,
but
> >a corrected STLPort has been posted.
>
> Where is the corrected STLPort? Is the corrected version on the STLPort
website,

No.

> the STLPort shipped with DMC++,

Yes.

Walter Bright wrote:
> What happened is DMC++ now implements dependent/nondependent name lookup.
> DefaultSize is not in scope at the declaration of NewBug according to my
> understanding of the Standard.

Imagine my surprise when suddenly the code no longer compiled. Not exactly "The Principle of Least Surprise".

I understood what was going on when I looked at the code, and posting a fair warning seemed prudent.

> There were a couple of problems compiling STLPort with this new lookup, but
> a corrected STLPort has been posted.

Is there a diff somewhere so I can update 4.6.2 on my site?

> I believe the compiler is operating correctly.

It's your compiler. Of course it is! :-)


-scooter

tink wrote:
> In article <dl1nqt$5fi$1@digitaldaemon.com>, Walter Bright says...
> 
>>There were a couple of problems compiling STLPort with this new lookup, but
>>a corrected STLPort has been posted.
> 
> 
> Where is the corrected STLPort? Is the corrected version on the STLPort website,
> the STLPort shipped with DMC++, or the STLPort 4.6.2 from Scott Michel's site?

Took me most of the day to recompile STLport. Updated ZIP file and patch files for 4.6.2:

http://mordred.cs.ucla.edu/STLport_DMC/STLport_8.46.zip
http://mordred.cs.ucla.edu/STLport_DMC/STLport_8.46.patch


-scooter