Bug in datatest.cpp and workaround

[trungantran]

Compiler: g++
STL Platform: STLPort(with debug mode turned on)

I got these error messages when running "cryptest v":

c:/mingw/include/stlport/stl/debug/_string.h(226): STL error : Index
out of bounds
c:/mingw/include/stlport/stl/debug/_string.h(226): STL assertion
failure:     __n < this->size()

The source of this problem is a bug in the implementation of
"PutDecodedDatumInto":

void PutDecodedDatumInto(const TestData &data, const char *name,
BufferedTransformation &target)
{
        std::string s1 = GetRequiredDatum(data, name), s2;

        while (!s1.empty())
        {
                while (s1[0] == ' ') //<-- Bug is here
                        s1 = s1.substr(1);
                ...
        }
}

It's obvious that if s1 is empty, "s1[0]" will be an invalid read
because 0 == s1.size() (According to C++ standard, index is valid only
if its value is less than the length of string).

The workaround is to change:
        while (s1[0] == ' ')
                s1 = s1.substr(1);

to:
        while (s1[0] == ' ') {
                s1 = s1.substr(1);
                if (s1.empty())
                        return; //avoid invalid read if s1 is empty
        }

Note:
1. This bug makes many tests failed because datum of some fields ends
with a space character ("MAC" field in authenticated-symmetric cipher
test for e.g).

2. To reproduce this bug with other compiler/STL platform (MSVC/
Dinkumware for e.g), just replace "s1[0]" with "s1.at(0)" (because
std::string::operator[ ]  doesnt check whether index is valid) and you
will get the "std::out_of_range" exception (MSVC/Dinkumware will
produce "invalid string position" error messages).

Regards,
        An


--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the "Crypto++ Users" Google Group.
To unsubscribe, send an email to cryptopp-users-unsubscribe@....
More information about Crypto++ and this group is available at http://www.cryptopp.com.
-~----------~----~----~----~------~----~------~--~---

[Wei Dai]

Thanks, I've checked in your fix.

--------------------------------------------------
From: "trungantran" <trantrungan@...>
Sent: Thursday, June 18, 2009 11:11 AM
To: "Crypto++ Users" <cryptopp-users@...>
Subject: Bug in datatest.cpp and workaround

>
> Compiler: g++
> STL Platform: STLPort(with debug mode turned on)
>
> I got these error messages when running "cryptest v":
>
> c:/mingw/include/stlport/stl/debug/_string.h(226): STL error : Index
> out of bounds
> c:/mingw/include/stlport/stl/debug/_string.h(226): STL assertion
> failure:     __n < this->size()
>
> The source of this problem is a bug in the implementation of
> "PutDecodedDatumInto":
>
> void PutDecodedDatumInto(const TestData &data, const char *name,
> BufferedTransformation &target)
> {
> std::string s1 = GetRequiredDatum(data, name), s2;
>
> while (!s1.empty())
> {
> while (s1[0] == ' ') //<-- Bug is here
> s1 = s1.substr(1);
> ...
> }
> }
>
> It's obvious that if s1 is empty, "s1[0]" will be an invalid read
> because 0 == s1.size() (According to C++ standard, index is valid only
> if its value is less than the length of string).
>
> The workaround is to change:
> while (s1[0] == ' ')
> s1 = s1.substr(1);
>
> to:
> while (s1[0] == ' ') {
> s1 = s1.substr(1);
> if (s1.empty())
> return; //avoid invalid read if s1 is empty
> }
>
> Note:
> 1. This bug makes many tests failed because datum of some fields ends
> with a space character ("MAC" field in authenticated-symmetric cipher
> test for e.g).
>
> 2. To reproduce this bug with other compiler/STL platform (MSVC/
> Dinkumware for e.g), just replace "s1[0]" with "s1.at(0)" (because
> std::string::operator[ ]  doesnt check whether index is valid) and you
> will get the "std::out_of_range" exception (MSVC/Dinkumware will
> produce "invalid string position" error messages).
>
> Regards,
> An
>
>
> >
>

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the "Crypto++ Users" Google Group.
To unsubscribe, send an email to cryptopp-users-unsubscribe@....
More information about Crypto++ and this group is available at http://www.cryptopp.com.
-~----------~----~----~----~------~----~------~--~---