[Olivier Brunet]

I have an old app using RSA signature through crypto 4.2. I migrated to
5.2, and now the RSA signature code doesn't work anymore (ie it compiles,
but valid signatures aren't verified anymore). I tried some change on the
code to use new 5.2 interface but no luck. Here is the actual code, if you
see obvious errors:
static RandomPool & GlobalRNG()
{
         static AutoSeededRandomPool randomPool;
         return randomPool;
}


SignCode:
         RSASSA_PKCS1v15_SHA_Signer
privsign(StringSource(secKey,secKeySize,true));
         SecByteBlock signature(privsign.MaxSignatureLength());
         signlen = privsign.SignMessage(GlobalRNG(),(const byte *)buf, len,
signature);
         StringSink sink(sign);
         sink.Put(signature,signlen);
         sink.MessageEnd();
Where secKey/secKeySize is the private key (generated by crypto++) and
buf/len is the buffer to sign. The final sink stores data in 'sign' which
is a std::string.
Now for verify:
         RSASSA_PKCS1v15_SHA_Verifier
pub(StringSource(pubKey,pubKeySize,true));
         return pub.VerifyMessage((const byte *)buf, len, (const byte
*)sign, signlen);
where pubKey/pubKeySize is the public key corresponding to secKey, and
buf/len the same buffer than before. It always return false.
The secret key is generated elsewhere with:
         AutoSeededRandomPool randPool;
         RSAES_OAEP_SHA_Decryptor priv(randPool, keylength);
         TransparentFilter privFile(new FileSink(holder));
         priv.DEREncode(privFile);
         privFile.MessageEnd();
and pubkey is derived from it.

I traced down both sign and verify opertions in crypto++, and found that
during verify in
PK_DeterministicSignatureMessageEncodingMethod::VerifyMessageRepresentative()
the ComputeMessageRepresentative() seems to return some kind of null value.
Is there a visible mistake in my code ????

Thanks for your help,
Regards.