About clear(). Please help!

I recently learnt from C++ Primer 4th Edition that

Quote:

---

8.2. Condition States
... ...
s.clear(flag): Set specified condition state(s) in s to valid state. Type of flag is strm::iostate.
... ...
cin.clear(istream::failbit); // clear the failbit state
... ...

---

But it turns out that clear sets the condition states according to flag, that's, after clear is called, the condition states is equal to flag.

Does it mean that the author is wrong?
Or is it in the new standard, if any, that clear is used to set specified condition states to valid?

When I refered to C++ Primer 3rd Edition, I found that

Quote:

---

20.7 Condition States
... ...
Using the clear() member function, we reset the condition state to an explicit value.
... ...

---

I'm really very confused now!

As far as I know, clear() resets failed states to normal. Say you try to open a non-existant file, now that stream will evaluate to a fail state, but if you call clear() on it, the state gets restored and you can use it again.

I'm sure the author is wrong. I think that code will clear all flags then set failbit. Have you checked the errata?

Yes, clear() resets failed states to normal. I read the prototype of clear, and found that it's prototyped as "clear(state=goodbit)". So if it's to call clear(), then it resets the states to normal. But here the author says that by callig clear(istream::failbit) clear the failbit state. But in my compiler, it clears all other states and sets the failbit instead.

as others have said, the author was wrong in the 3rd edition... that's probably one of the reasons there's a 4th edition ;)

As i said clear(goodbit) would be a SETTING of goodbit so obviously clear(failbit) isnt going to clear the failbit state as the comment suggested.

check this definition: http://www.cppreference.com/cppio/clear.html

in that definition, clear defaults to goodbit, in which case it clears everything... but if you pass it the failbit, it'll only clear that...

i reckon this says it all

Quote:

---

void clear(iostate state = goodbit);
4 Postcondition: If rdbuf()!=0 then state == rdstate(); otherwise
rdstate()==state|ios_base::badbit.
5 Effects: If (rdstate() & exceptions()) == 0, returns. Otherwise, the function throws an
object fail of class basic_ios::failure (27.4.2.1.1), constructed with implementationdefined
argument values.
__________________
270) This suggests an infinite amount of copying, but the implementation can keep track of the maximum element of the arrays that is
nonzero.
616
© ISO/IEC ISO/IEC 14882:1998(E)
27 Input/output library 27.4.4.3 basic_ios iostate flags functions
void setstate(iostate state);
6 Effects: Calls clear(rdstate() | state) (which may throw basic_ios::failure
(27.4.2.1.1)).

---

look at how setstate is implemented. Kinda says it all for me.

Later in the 4th edition, the author says

Quote:

---

8.2. Condition States
... ...
// remember current state of cin
istream::iostate old_state = cin.rdstate();
cin.clear();
process_input(); // use cin
cin.clear(old_state); // now reset cin to old state
... ...

---

I think this means that clear is used to set the states to a given states, isn't it?
If clear is used to set specified condition states to valid states, isn't the author self-contradiction here?

Is it "C++ Primer Plus" by Stephen Prata?

gg

NO. It's C++ Primer 4th Edition by Stanley B. Lippman, Josée Lajoie, Barbara E. Moo

Well, the 3rd edition is highly recommended by two reviewers on the ACCU website:
http://www.accu.org/cgi-bin/accu/rvo...file=cp001565a

Hopefully, the 4th edition has improved further. But the authors did leave in a "typo" in this case.

gg

I guess the author is wrong at this point because the author is self-contradiction.
And when I refered to C++ Programming Language By Bjarne Stroustrup, I found

Quote:

---

... ...
s.clear(ios_base::badbit); // set state
... ...

---