On 2010/03/19 11:30, Mark Bucciarelli wrote: > On Fri, Mar 19, 2010 at 11:05 AM, Stuart Henderson <[email protected]> > wrote: > > > > There are some problems with SSL checks which need configure > > > > Maybe an upstream fix didn't make it into the Feb2010 > tarballs. > > There was a patch to buildtools/conf/MonetDB.m4 applied > to HEAD to fix issues I was having with finding openssl. > > ref: > > http://sourceforge.net/tracker/index.php?func=detail&aid=2914578&group_id=56967&atid=482468 > > m >
Their fix leaves a seemingly pointless check for -lcrypto after the -lssl check, I just moved the ssl check up a bit and used the result in the -lcrypto check. The perl check isn't working either; -lperl needs -lm otherwise it can't locate symbols e.g. for sqrt(). I moved the MATH_LIBS check earlier and used the result when checking for -lperl. There might be a better way but it's enough to get going (and discover that there's some other problem with installing the perl bindings...) Another problem is that #include <malloc.h> triggers a #warning which with MonetDB's preferred -Werror causes the build to error out. I've just put in an os-dependent check as a workaround but that's not much good for upstream, it could do with being a bit more restrictive and only pulling in malloc.h when it's required. I ran into a problem with our autoconf ports; when regenerating the configure script using base m4, instead of SIZEOF_VOID_P, SIZEOF_VOID__ is used (and causes compilation to fail); I used a nasty hack to build with gm4 and work around it, but that needs tracking down in autoconf/m4 sometime.. Here's what I have up to now. Not tested yet but I thought I'd send something out rather than sit on it...
monetdb.tgz
Description: application/tar-gz
