On Sat, Sep 06, 2008 at 12:22:11PM +0200, Philipp Kern wrote: > # oh noes, not a default build flavour, thus not RC > severity 497960 normal >
Yet another case of "policy > common sense". By looking at debian/rules you can see that USE_OPENSSL=1 is a debian package construction and breaking it is a sucky regression no matter what policy says about it. (I guess the reason why the users of ircd-hybrid needs to rebuild in the first place is because of the distributable problems of GPL vs OpenSSL.) Anyway, ... The problem comes from the apparently untested patch which was added in the latest version. Snipped from the package changelog: > * Add 19_sslonly.dpatch to allow setting a channel to be only joinable > by SSL-users or people on localhost, thanks to Joshua Kwan. > (Closes: #419934) Disabling this patch in debian/patches/00list would be a safe way of fixing the regression. (The feature in the patch might be a good idea when the patch is fixed, but all the patch does now is nothing in the non-ssl case and breaks build in the ssl case.) I've attached a build-fix which i have *no* *idea* if it's the correct fix, but it looks kind of like the localClient part was a cut'n'paste error from a couple of lines above where fd is used. -- Andreas Henriksson
--- ircd-hybrid-7.2.2.dfsg.2/debian/patches/19_sslonly.dpatch 2008-09-06 16:14:19.000000000 +0200 +++ ircd-hybrid-7.2.2.dfsg.2-fixed/debian/patches/19_sslonly.dpatch 2008-09-06 16:15:23.000000000 +0200 @@ -71,7 +71,7 @@ + int fd = source_p->localClient->fd.fd; + fde_t *F = lookup_fd(fd); + -+ if (F && !F->ssl && strcmp(source_p->localClient->sockhost, "127.0.0.1")) ++ if (F && !F->ssl && strcmp(source_p->sockhost, "127.0.0.1")) + return (ERR_SSLONLYCHAN); + } +#else
