Hi,
thanks for your reply. I found a very simple workaround (within the
DigitalNet library context), by decoding - removing attributes - then
encoding again, before entering Crypto++. This way it works fine.
Btw, I think that if we just fetch attributes in Crypto++ - without any
interpretation - it's all what is needed to solve this problem.
Regards,
Renzo
----- Original Message -----
From: "Wei Dai" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Thursday, May 06, 2004 11:37 AM
Subject: Re: optional PrivatekeyInfo attributes
> On Thu, May 06, 2004 at 10:31:45AM +0200, Renzo Tomaselli wrote:
> > RSASSA_PKCS1v15_MD5_Signer rsaPriv(privateKey), I discovered that
optional
> > PKCS#8 attributes are never decoded (rsa.h, line#98).
> > This in turn is called from asn.cpp, line#534. Nex line raises an
exception
> > since overal seq length does not match actual fetched length (in
> > MessageEnd).
> > I feel that the passed private key is ok even with such attributes, thus
it
> > looks like a Crypto++ bug.
>
> Yeah, there's a comment in asn.h explaining that "skip optional attributes
> if present" still needs to be implemented. One reason I haven't done it
> yet is because I haven't encountered a PKCS#8 private key with optional
> attributes until now. Unfortunately I'm leaving on a trip without my
> computer for at least 10 days. If you can wait until I come back to fix
> this bug, please go ahead and send me the key with optional attributes.
> Otherwise you might want to try to implement it yourself and send me a
> patch.
>
>
