Re: Possible bug

[Aris Adamantiadis <aris@xxxxxxxxxxxx>]

Hi Andreas,

Good fix, you didn't even forget the size_t cast.
It would be interesting to understand why we didn't catch this with an
automated test. Is it because the OpenSSH we use for nightly testing are
too old for Curve25519 kex?
Sometimes being ahead of OpenSSH isn't always an advantage :)

Aris

Le 25/08/14 12:24, Andreas Schneider a écrit :
> On Monday 25 August 2014 11:48:09 Giovanni Panozzo wrote:
>> Hi, I'm new on this list. I'm using libssh and I think I have found a
>> bug which crashes my program.
>>
>> I'm posting here because the site red.libssh.org is currently broken.
>>
>>
>> In src/dh.c the call:
>>
>> rc = ssh_buffer_pack(buf,
>> 	"dPdP",
>> 	CURVE25519_PUBKEY_SIZE,
>> 	session->next_crypto->curve25519_client_pubkey,
>> 	CURVE25519_PUBKEY_SIZE,
>> 	session->next_crypto->curve25519_server_pubkey);
>>
>> Is missing the data length parameter for "P"
>>
>>
>> Should it be ?
>>
>> rc = ssh_buffer_pack(buf,
>> 	"dPdP",
>> 	CURVE25519_PUBKEY_SIZE,
>> 	CURVE25519_PUBKEY_SIZE,
>> 	session->next_crypto->curve25519_client_pubkey,
>> 	CURVE25519_PUBKEY_SIZE,
>> 	CURVE25519_PUBKEY_SIZE,
>> 	session->next_crypto->curve25519_server_pubkey);
>>
> Yes, you're correct. Good catch! I've pushed a fix upstream.
>
>
>
> 	-- andreas
>