Obvious error, which actually compiled

[Chris Theis]

Hi all,

yesterday I did some porting of my code from the MSVC 7.1 compiler to GNU
3.3 and found the following typo flaw. Interestingly enough the code
compiled and ran fine under MSVC but GCC complained, as I would have
expected. However, I did not yet figure out why it worked under MSVC and
probably somebody can shed some light on this:

if( !pNode->IsValid() ) {
pNode->SetDirty( pNode->GetFather()->IsDirty ); // IMPORTANT: This
should read ->IsDirty() as it is a function returning a bool!
}

The declaration of SetDirty looks like this:

SetDirty( bool bFlag );

The declaration of IsDirty() looks like:

bool IsDirty() const;

Is there anybody who has an idea?

Thanks
Chris

[marcwentink]

> pNode->GetFather()->IsDirty

This gets the function pointer of "IsDirty", which could be NULL which
is 'false' read as a bool or a function and then "an address" = 'true'.

Marcus Wentink

[Jakob Bieling]

Chris Theis <ch*************@nospam.cern.ch> wrote:

Hi all,

yesterday I did some porting of my code from the MSVC 7.1 compiler to
GNU 3.3 and found the following typo flaw. Interestingly enough the
code
compiled and ran fine under MSVC but GCC complained, as I would have
expected. However, I did not yet figure out why it worked under MSVC
and probably somebody can shed some light on this:

if( !pNode->IsValid() ) {
pNode->SetDirty( pNode->GetFather()->IsDirty ); // IMPORTANT:
This should read ->IsDirty() as it is a function returning a bool!
}

The declaration of SetDirty looks like this:

SetDirty( bool bFlag );

The declaration of IsDirty() looks like:

bool IsDirty() const;

Is there anybody who has an idea?

According to 4.12/1 pointers to members can be converted to bool,
which seems to be happening implicitly above. So I would assume VC++ 7.1
to be correct on this one.

hth
--
jb

(reply address in rot13, unscramble first)

[Rolf Magnus]

Jakob Bieling wrote:

Chris Theis <ch*************@nospam.cern.ch> wrote:

Hi all,

yesterday I did some porting of my code from the MSVC 7.1 compiler to
GNU 3.3 and found the following typo flaw. Interestingly enough the
code
compiled and ran fine under MSVC but GCC complained, as I would have
expected. However, I did not yet figure out why it worked under MSVC
and probably somebody can shed some light on this:

if( !pNode->IsValid() ) {
pNode->SetDirty( pNode->GetFather()->IsDirty ); // IMPORTANT:
This should read ->IsDirty() as it is a function returning a bool!
}

The declaration of SetDirty looks like this:

SetDirty( bool bFlag );

The declaration of IsDirty() looks like:

bool IsDirty() const;

Is there anybody who has an idea?
According to 4.12/1 pointers to members can be converted to bool,

But IsDirty is not a pointer to member. Member functions are not converted
into pointers to them implicitly, like nonmember functions. You have to use
the & operator to get the address.
which seems to be happening implicitly above. So I would assume VC++ 7.1
to be correct on this one.

I'd say it isn't.

[Ben Pope]

Rolf Magnus wrote:

Jakob Bieling wrote:

Chris Theis <ch*************@nospam.cern.ch> wrote:

Hi all,

yesterday I did some porting of my code from the MSVC 7.1 compiler to
GNU 3.3 and found the following typo flaw. Interestingly enough the
code
compiled and ran fine under MSVC but GCC complained, as I would have
expected. However, I did not yet figure out why it worked under MSVC
and probably somebody can shed some light on this:

if( !pNode->IsValid() ) {
pNode->SetDirty( pNode->GetFather()->IsDirty ); // IMPORTANT:
This should read ->IsDirty() as it is a function returning a bool!
}

The declaration of SetDirty looks like this:

SetDirty( bool bFlag );

The declaration of IsDirty() looks like:

bool IsDirty() const;

Is there anybody who has an idea?

According to 4.12/1 pointers to members can be converted to bool,

But IsDirty is not a pointer to member. Member functions are not converted
into pointers to them implicitly, like nonmember functions. You have to use
the & operator to get the address.

I think that a default compiler extension in 7.1 is to not require the
&, whether that applies to lookup of the conversion as well, I'm not sure.

Ben Pope
--
I'm not just a number. To many, I'm known as a string...

[Pete Becker]

ma*********@hotmail.com wrote:

pNode->GetFather()->IsDirty

This gets the function pointer of "IsDirty", which could be NULL which
is 'false' read as a bool or a function and then "an address" = 'true'.

Well, yes, but that's a Microsoft extension. The only standard syntax
for taking the address of a member function is with a qualified name,
i.e. classname::functionname.

--

Pete Becker
Roundhouse Consulting, Ltd.

[Pete Becker]

Pete Becker wrote:

Well, yes, but that's a Microsoft extension. The only standard syntax
for taking the address of a member function is with a qualified name,
i.e. classname::functionname.

Whoops, that's also a Microsoft extension. The standard syntax for
taking the address of a member function requires an & and a qualified
name, i.e. &classname::functionname.

--

Pete Becker
Roundhouse Consulting, Ltd.

[Chris Theis]

"Rolf Magnus" <ra******@t-online.de> wrote in message
news:du*************@news.t-online.com...

Jakob Bieling wrote:

Chris Theis <ch*************@nospam.cern.ch> wrote:

Hi all,

yesterday I did some porting of my code from the MSVC 7.1 compiler to
GNU 3.3 and found the following typo flaw. Interestingly enough the
code
compiled and ran fine under MSVC but GCC complained, as I would have
expected. However, I did not yet figure out why it worked under MSVC
and probably somebody can shed some light on this:

if( !pNode->IsValid() ) {
pNode->SetDirty( pNode->GetFather()->IsDirty ); // IMPORTANT:
This should read ->IsDirty() as it is a function returning a bool!
}

The declaration of SetDirty looks like this:

SetDirty( bool bFlag );

The declaration of IsDirty() looks like:

bool IsDirty() const;

Is there anybody who has an idea?

According to 4.12/1 pointers to members can be converted to bool,

But IsDirty is not a pointer to member. Member functions are not converted
into pointers to them implicitly, like nonmember functions. You have to
use
the & operator to get the address.

which seems to be happening implicitly above. So I would assume VC++ 7.1
to be correct on this one.

I'd say it isn't.

Yes, that's what I also thought. But it seems that this is really a
"feature" of a MS extension, as Pete pointed out.

Thanks guys
Chris