[Bug libstdc++/67503] New: String cannot be loaded from binary representation

[Bug libstdc++/67503] New: String cannot be loaded from binary representation

[radventure at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

            Bug ID: 67503
           Summary: String cannot be loaded from binary representation
           Product: gcc
           Version: 5.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libstdc++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: radventure at yandex dot ru
  Target Milestone: ---

#include <iostream>
#include <string>

int main() {
  unsigned char buff1[sizeof(std::string)], buff2[sizeof(std::string)];
  std::string s1("SMAL STRING BUG"), s2;
  new (&buff1) std::string(s1);
  s2 = *(reinterpret_cast<std::string*>(&buff1));
  std::cout << s2 << std::endl;
  std::swap(buff1, buff2);
  s2 = *(reinterpret_cast<std::string*>(&buff2));
  std::cout << s2 << std::endl;
}

After swapping buffers _N_dataplus._M_p pointer points into we buff1 but actual
data stored in small local buffer was coped correctly. If initial string length
will be greater when data will be stored into the heap and everything will be
Ok.

[Bug libstdc++/67503] String cannot be loaded from binary representation

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |INVALID

--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> ---
That's not valid, you can't do that with non-trivial types.

[Bug libstdc++/67503] String cannot be loaded from binary representation

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Also there is an alignment issue with your example too.

[Bug libstdc++/67503] String cannot be loaded from binary representation

[radventure at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

--- Comment #3 from radventure at yandex dot ru ---
I can solve the alignment but prbolem will not be fixed. 
I agree with remark about "non-trivial types" but this code works in previous
gcc versions and works in visual c++ 2015.

[Bug libstdc++/67503] String cannot be loaded from binary representation

[radventure at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

radventure at yandex dot ru changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|INVALID                     |WONTFIX

[Bug libstdc++/67503] String cannot be loaded from binary representation

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|WONTFIX                     |INVALID

--- Comment #4 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to radventure from comment #3)
> I can solve the alignment but prbolem will not be fixed. 
> I agree with remark about "non-trivial types" but this code works in
> previous gcc versions and works in visual c++ 2015.

It was never valid before and it's not valid now. Just because it appeared to
work previously doesn't make it valid.

[Bug libstdc++/67503] String cannot be loaded from binary representation

[radventure at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

--- Comment #5 from radventure at yandex dot ru ---
When you use local buffer for storing string value it not necessary to have
pointer to it. And we can reduce the size of string by the syzeof(pointer).

[Bug libstdc++/67503] String cannot be loaded from binary representation

[radventure at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

--- Comment #6 from radventure at yandex dot ru ---
(In reply to Jonathan Wakely from comment #4)
> (In reply to radventure from comment #3)
> > I can solve the alignment but prbolem will not be fixed. 
> > I agree with remark about "non-trivial types" but this code works in
> > previous gcc versions and works in visual c++ 2015.
> 
> It was never valid before and it's not valid now. Just because it appeared
> to work previously doesn't make it valid.

Code validity it's great. But what about backward compatibility?

[Bug libstdc++/67503] String cannot be loaded from binary representation

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This is just like having a struct like:
Struct
{
  Char a[4];
  Char *b = a;
};

Sorry for char and struct being in the wrong case. Also using a c++11 feature
to show the case easier.  This is a valid thing to do and that is exactly what
the new std::string does. This saves space and allows for short strings not
have to an extra allocation.

[Bug libstdc++/67503] String cannot be loaded from binary representation

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to radventure from comment #5)
> When you use local buffer for storing string value it not necessary to have
> pointer to it. And we can reduce the size of string by the syzeof(pointer).

Well an union would work here except it would make all other code more messy.
There are a few places need to check that it equals the array before doing
something special.

[Bug libstdc++/67503] String cannot be loaded from binary representation

[radventure at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

--- Comment #9 from radventure at yandex dot ru ---
I understand the "small string optimization" idea. I agree about allocation
counting. But I don't see space economy, now sizeof(string) is 28 bytes in
32-bit environment instead of 4 bytes early.

[Bug libstdc++/67503] String cannot be loaded from binary representation

[radventure at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67503

--- Comment #10 from radventure at yandex dot ru ---
(In reply to radventure from comment #9)
> I understand the "small string optimization" idea. I agree about allocation
> counting. But I don't see space economy, now sizeof(string) is 28 bytes in
> 32-bit environment instead of 4 bytes early.

Now sizeof(string) is 24 bytes.