Porting code, need guidance, Thx

[Kenton Groombridge]

Hi,
I previous posted this in a gcc and g++ groups since that is the
compiler I am working with, but I didn't get any response.

Hopefully these are the right groups for this question. I am working
(actually playing) on porting Alien vs Predator to Linux. I am using
the code that is available via CVS from icculus.org.

I recently upgraded my gcc to 3.4.1 and now a portion of the code
doesn't compile. Nothing in the code has changed.

I think I figured it out, but want to be sure before I spend a bunch of
time working around it only to find that I was wrong.

My guess is the class ConstIterator makes class Iterator a friend, but
class Iterator is a derived class of the bass class ConstIterator (hope
I said that right). From what I understand and have researched is that
in order for the class ConstIterator to use the class Iterator as a
friend, then class Iterator has to be defined first, and for class
Iterator to be derived from class ConstIterator, then the class
ConstIterator has to be defined first. Essentially a catch 22.

Am I on the right track? If not please point me in the right directly,
but I would like to fix the code. I would like to know how you would
approach the fix for this.

Thanks,
Ken

Here is the error:

g++ -g -Wall -pipe -O2 -DLINUX -Dengine=1 -I. -Iinclude -Iwin95 -Iavp
-Iavp/win95 -Iavp/support -Iavp/win95/frontend -Iavp/win95/gadgets
-I/usr/include/SDL -D_REENTRANT -c -o win95/awtexld.o win95/awtexld.cpp
In file included from win95/awtexld.cpp:10:
win95/hash_tem.hpp: In member function `void _base_HashTable<TYPE,
ARG_TYPE, CMP_ARG_TYPE>::Iterator::Remove()':
win95/hash_tem.hpp:435: error: `nEntriesRemaining' undeclared (first use
this function)
win95/hash_tem.hpp:435: error: (Each undeclared identifier is reported
only once for each function it appears in.)
win95/hash_tem.hpp:439: error: `nodePP' undeclared (first use this function)
win95/hash_tem.hpp:446: error: `chainPP' undeclared (first use this
function)
win95/hash_tem.hpp:447: error: `nChainsRemaining' undeclared (first use
this function)
win95/hash_tem.hpp: In copy constructor `_base_HashTable<TYPE, ARG_TYPE,
CMP_ARG_TYPE>::_base_HashTable(const _base_HashTable<TYPE, ARG_TYPE,
CMP_ARG_TYPE>&)':
win95/hash_tem.hpp:705: error: expected `;' before "it"
win95/hash_tem.hpp:705: error: `it' undeclared (first use this function)
make: *** [win95/awtexld.o] Error 1

Here is the code in question (not very big, but put in this format so
you can find the lines in question in relation to the errors above):

http://webpages.charter.net/kgroombr/hash_tem.hpp

 

[John Harrison]

Hi,
I previous posted this in a gcc and g++ groups since that is the
compiler I am working with, but I didn't get any response.

Hopefully these are the right groups for this question. I am working
(actually playing) on porting Alien vs Predator to Linux. I am using
the code that is available via CVS from icculus.org.

I recently upgraded my gcc to 3.4.1 and now a portion of the code
doesn't compile. Nothing in the code has changed.

I think I figured it out, but want to be sure before I spend a bunch of
time working around it only to find that I was wrong.

My guess is the class ConstIterator makes class Iterator a friend, but
class Iterator is a derived class of the bass class ConstIterator (hope
I said that right). From what I understand and have researched is that
in order for the class ConstIterator to use the class Iterator as a
friend, then class Iterator has to be defined first, and for class
Iterator to be derived from class ConstIterator, then the class
ConstIterator has to be defined first. Essentially a catch 22.

Not correct. One class can be declared as a friend of another without
being declared itself.

Am I on the right track? If not please point me in the right directly,
but I would like to fix the code. I would like to know how you would
approach the fix for this.

This is a change in the rules for dependent name lookup. It's catching a
lot of people out but g++ 3.4 is doing the right thing, whereas older
compilers where not.

For instance in Iterator prefix all references to names in ConstIterator
with this->, so

void Remove()
{
if (!this->nEntriesRemaining)
{
HT_FAIL("HTT: Tried to Remove() via an iterator which was Done()");
}
Node * oldP = *this->nodePP;
*this->nodePP = oldP->nextP;
delete oldP;
if (!*this->nodePP)
{
do
{
++ this->chainPP;
-- this->nChainsRemaining;
}
while (this->nChainsRemaining && !*this->chainPP);
this->nodePP = this->chainPP;
}
-- this->nEntriesRemaining;
-- *tableNEntriesP;
}

This is explained in the FAQ

http://www.parashift.com/c++-faq-lite/containers-and-templates.html#faq-34.17

john

 

[Ben Hutchings]

Kenton Groombridge wrote:

I recently upgraded my gcc to 3.4.1 and now a portion of the code
doesn't compile. Nothing in the code has changed.

I think I figured it out, but want to be sure before I spend a bunch of
time working around it only to find that I was wrong.

My guess is the class ConstIterator makes class Iterator a friend, but
class Iterator is a derived class of the bass class ConstIterator (hope
I said that right). From what I understand and have researched is that
in order for the class ConstIterator to use the class Iterator as a
friend, then class Iterator has to be defined first, and for class
Iterator to be derived from class ConstIterator, then the class
ConstIterator has to be defined first. Essentially a catch 22.

That is a problem, but not the problem that's being reported.

Am I on the right track? If not please point me in the right directly,
but I would like to fix the code. I would like to know how you would
approach the fix for this.

<snip>

I would be inclined to make ConstIterator's member variables
protected. This is not generally a good idea, but I wouldn't want to
go making more drastic changes to the existing code. See
<http://www.parashift.com/c++-faq-lite/basics-of-inheritance.html#faq-19.8>.

I think the errors reported by the compiler are the result of the
change in binding of names in templates that was made in version 3.4
of GNU C++. (The change was made in the standard long ago.) See
<http://www.decadentplace.org.uk/womble/cplusplus/template-faq.html#base-lookup>.
Given this, they can be fixed by adding "this->" before the names of
the inherited member variables in the Iterator class. The FAQ I
linked to explains why.

 

[Kenton Groombridge]

Thanks a bunch,

This fixed this issue, but now I have another compile problem. I have
been tinkering with it for several hours but am unsure what it is. The
error is:

win95/hash_tem.hpp: In copy constructor `_base_HashTable<TYPE, ARG_TYPE,
CMP_ARG_TYPE>::_base_HashTable(const _base_HashTable<TYPE, ARG_TYPE,
CMP_ARG_TYPE>&)':
win95/hash_tem.hpp:705: error: expected `;' before "it"
win95/hash_tem.hpp:705: error: `it' undeclared (first use this function)
win95/hash_tem.hpp:705: error: (Each undeclared identifier is reported
only once for each function it appears in.)
win95/hash_tem.hpp:705: error: expected primary-expression before ';' token
win95/hash_tem.hpp:705: error: expected `)' before ';' token
win95/hash_tem.hpp:705: error: expected `;' before ')' token
make: *** [win95/awtexld.o] Error 1

On line 705 I see where it is calling the copy constructor
_base_HashTable::ConstIterator it(ht), but all looks good to me. At
least from what I learned. Any guess what could cause this one? My
guess is that line 705 is good, but the error lies somewhere in the copy
constructor, but it also looks good to me.

I have three C++ books and the most recent one is about seven years old.
Have a recommendation on one that will bring me up to speed. Most of
the programs in my books don't compile with gcc 3.4.1 until I do some of
the code fixes that I found with google searching.

Thanks again,
Ken

Kenton, I never post to moderated groups, in case you missed my reply

in comp.lang.c++ here it is direct to your inbox.

 

[llewelly]

Hi,
I previous posted this in a gcc and g++ groups since that is the
compiler I am working with, but I didn't get any response.

Hopefully these are the right groups for this question. I am working
(actually playing) on porting Alien vs Predator to Linux. I am using
the code that is available via CVS from icculus.org.

I recently upgraded my gcc to 3.4.1 and now a portion of the code
doesn't compile. Nothing in the code has changed.

g++ 3.4 implements two-phase lookup as required by the 1998
standard. That makes it in some degree source-code incompatible
with g++ 3.x .

You can find examples on the gcc website; see the 'C++' section of:
http://gcc.gnu.org/gcc-3.4/changes.html

I think I figured it out, but want to be sure before I spend a bunch of
time working around it only to find that I was wrong.

My guess is the class ConstIterator makes class Iterator a friend,

This is true but irrevelant. Friendship affects access control, not
namelookup. No declaration for 'nEntriesRemaining' is found,
therefor the issue is namelookup. This is in fact akin to one of
the examples on the gcc 3.4 changes page I linked to above.

but
class Iterator is a derived class of the bass class ConstIterator (hope
I said that right).
Correct.

From what I understand and have researched is that
in order for the class ConstIterator to use the class Iterator as a
friend, then class Iterator has to be defined first,

A definition is *not* necessary. Only a forward declaration is
necessary - and in some cases the friend declaration itself can
fulfill this purpose. But the whole friend issue is irrevelant.

and for class
Iterator to be derived from class ConstIterator, then the class
ConstIterator has to be defined first.
True.

Essentially a catch 22.

No, becuase declaring something friend does not require a definition
- but, as I said before, the whole friend issue is irrevelant.

[snip]

I would like to know how you would
approach the fix for this.

[snip]

Replace 'nEntriesRemaining' whith 'this->nEntriesRemaining'.

The Vandervoode and Josuttis _C++ Templates_ book has a section on
dependent names which I think explains why this works.

 

[Kenton Groombridge]

I finally figured this one out. I got to get used to the stricter rules
of the C++ standard. I also need some new books which stick to these
standards since most of the programs in my books no longer compile with
gcc 3.1. Been programming the wrong way for a long time. Anybody that
has book recommendations that stick with the strict C++ standard, please
respond. All my books are at least 7 to 12 years old.

Here is the fix for anybody that has issue with anything like this:

The error is on line 705 which is:

for (_base_HashTable::ConstIterator it(ht); !it.Done(); it.Next() )


The fix is:

for (typename _base_HashTable::ConstIterator it(ht); !it.Done();
it.Next() )

as found on:

http://gcc.gnu.org/gcc-3.4/changes.html

which states:

You must now use the typename and template keywords to disambiguate
dependent names, as required by the C++ standard.