Re: RFS: mbuffer (new package)
Hello again,
my clarification follows.
fredag den 11 december 2009 klockan 02:13 skrev Hector Oron detta:
> Hi,
>
> 2009/12/11 Mats Erik Andersson <mats.andersson@gisladisker.se>:
> > out of curiosity: Does mbuffer now prefer libmhash over libssl?
> > That has been troublesome in the past.
>
The documentation of the upstream author states that he prefers
libmhash over libssl. At about the same time as Peter Pentchev
made his first packaging of mbuffer this past summer, I found a
mistake in the upstreams configure.in template, which made libssl
the first choice, then libgnutls, and lastly libmhash, in case
more than one library would be present.
I did submit a patch to the upstream author Thomas Komor on this,
which he acknowledged. It seemed to me to be a better choice in
building a package for Debian by using the small library libmhash,
rather than to invoke libgnutls. The Debian Policy rules against
libssl in this situation because of mixed licences. But then again,
Peter Pentchev has invested time in his adaption for libgnutls,
which certainly is a factor here.
> $ cat 01-md5-gnutls.patch
> Description: Use the MD5 routines from GnuTLS instead of OpenSSL.
> Avoid linking to OpenSSL for license reasons.
> Forwarded: not-needed
> Author: Peter Pentchev <roam@ringlet.net>
> Last-Update: 2009-12-10
>
> --- a/configure
> +++ b/configure
> @@ -4912,7 +4912,7 @@
> return 0;
> }
> _ACEOF
> -for ac_lib in '' crypto; do
> +for ac_lib in '' gnutls-openssl; do
> if test -z "$ac_lib"; then
> ac_res="none required"
> else
> --- a/mbuffer.c
> +++ b/mbuffer.c
> @@ -81,7 +81,7 @@
> #define MD5_END(hash,ctxt) MD5Final(hash,&(ctxt))
> #define HAVE_MD5 1
> #elif defined HAVE_LIBCRYPTO
> -#include <openssl/md5.h>
> +#include <gnutls/openssl.h>
> static MD5_CTX MD5ctxt;
> #define MD5_INIT(ctxt) MD5_Init(&ctxt);
> #define MD5_UPDATE(ctxt,at,num) MD5_Update(&ctxt,at,num)
>
>
> --
> Héctor Orón
>
--
Mats Erik Andersson
Reply to: