On 2012-05-26 04:41, Phil Pennock wrote:
> Wolfgang's patch fixing my Cyrus SASL mistake is in head. Anyone object
> to the next cut being 4.80 instead of RC6?
>
> If all goes well, I'll cut on Sunday, just to leave tomorrow for finding
> any remaining issues.
>
> -Phil
>
I don't fully understand the comment in tls_gnu.c at the call to
gnutls_dh_params_export_pkcs3(). Is there no possibility of
a bug resulting from unsigned int and size_t being different sizes?
What would be wrong with:
--- a/src/src/tls-gnu.c
+++ b/src/src/tls-gnu.c
@@ -517,8 +517,12 @@ if (rc < 0)
uses "unsigned int" for the size field, but this call takes separate data
and size fields, with the latter being a size_t*. For now, we live with
the error as being safer than throwing away type information. */
+ {
+ size_t size = m.size;
rc = gnutls_dh_params_export_pkcs3(dh_server_params, GNUTLS_X509_FMT_PEM,
- m.data, &m.size);
+ m.data, &size);
+ m.size= size;
+ }
if (rc != GNUTLS_E_SUCCESS)
{
free(m.data);
Mind, though it compiles without warning here, it make no difference to the handful
of testsuite errors (2014, 2025, 2026, 2027) I'm down to.
OpenSSL builds and tests clean. No-TLS builds clean.
--
Cheers,
Jeremy
