Bug#503833: reassign, patch link

[dann frazier]

On Thu, Nov 06, 2008 at 08:33:08PM +0100, Andreas Metzler wrote:
> On 2008-11-06 dann frazier <dannf at debian.org> wrote:
> > reassign 503833 libtasn1-3
> > severity 503833 important
> > thanks
> 
> > Upstream has developed a patch:
> >   http://lists.gnu.org/archive/html/gnutls-devel/2008-11/msg00000.html
> 
> > If I apply this to libtasn1-3 it fixes the problem for me.
> > However, I had set the nocheck DEB_BUILD_OPTION because it causes
> > libtasn1-3 to fail at least one build-time test.
> 
> > It would be nice to get a fix for this into lenny as it will fix a
> > regression from etch's subversion.
> 
> Hello,
> 
> I have just uploaded a fixed version to experimental.

Thanks Andreas!

> I am reluctant
> to go immediately to unstable since it requires a shlibs bump due to
> an added function.

I definitely wouldn't call myself a shared lib packaging expert, but
does adding a function really break the ABI? It shouldn't break
backwards compatibility, and anything newly built that may use the new
function should automatically get a versioned runtime dependency.

> I am not also not sure about this change:
> 
> ----------------------
> --- a/lib/libtasn1.h
> +++ b/lib/libtasn1.h
> @@ -105,11 +105,14 @@ extern "C"
>  /* that represent an ASN.1 DEFINITION.                */
>  /******************************************************/
>  
> +#define SMALL_VALUE_SIZE 16
> +
>    struct node_asn_struct
>    {
>      char *name;                        /* Node name */
>      unsigned int type;         /* Node type */
>      unsigned char *value;      /* Node value */
> +    unsigned char small_value[SMALL_VALUE_SIZE]; /* if value is less than that store it here */
>      int value_len;
>      struct node_asn_struct *down;      /* Pointer to the son node */
>      struct node_asn_struct *right;     /* Pointer to the brother node */
>      struct node_asn_struct *left;       /* Pointer to the next list element */
>    };
> 
>    typedef struct node_asn_struct node_asn;
> 
>    typedef node_asn *ASN1_TYPE;
> ----------------------
> 
> Although the docs only talk about using the ASN1_TYPE pointer the
> struct definition still is in the public interface. Doesn't this break
> the ABI? - I am really not sure, to tired.

My guess would be yes[1].. which sucks because it would make it
significantly harder to get into lenny and fix the subversion
regression :( I wonder if there's an alternate way to handle it?

[1] but again, I'm still not an expert here :)
-- 
dann frazier