bug in iostreams/copy.hpp line 81

[nobody]

in boost 1_33_1 iostreams library

file copy.hpp

line 91 and following...

it should rather be :

streamsize amt;
while ((amt=iostreams::read(src, buf.data(),buffer_size)) != -1)
{
   std::copy(buf.data(), buf.data() + amt, p.first + total);
   total += amt;
}
return total;

otherwise it happens that the std::copy line is called with amt =-1 which is a bug isnt it ?

Regards,

Mathieu

[alexis_wilke]

Logged In: YES 
user_id=554061
Originator: NO

Mathieu,

Actually, the copy operator will test if(_first < _last) <do the copy> and if amt is -1 that condition will be false. Yet I agree that your way we would avoid one call and one variable ('done' in the original loop). I'm not part of boost so I cannot decide whether to apply the change. But this is only an optimization.

Alexis

[anonymous]

There is a patch for this in "official patch format" in: ​http://article.gmane.org/gmane.comp.lib.boost.user/23522

[chris@…]

Alexis,

My copy operator did not perform your test and I saw the bug (5 or 6 levels down the stack a copy for last-first would eventually SIGSEGV in memcpy.)

-

Chris DUrso

[qbowater@…]

Replying to alexis_wilke:

Mathieu,

Actually, the copy operator will test if(_first < _last) <do the copy> and if amt is -1 that condition will be false. Yet I agree that your way we would avoid one call and one variable ('done' in the original loop). I'm not part of boost so I cannot decide whether to apply the change. But this is only an optimization. Alexis

In VC8 DEBUG mode,it will assert in std::copy because "first > last". It make difficult to debug program.I think it should be repaired necessarily.

[anonymous]

This looks like a duplicate of #1070

[Jonathan Turkanis]

Fixed in [42294] in branches/iostreams_dev, to be merged into trunk shortly. The fix is a slight modification of the suggestion above.