9:22 a.m.
Detection based on gnutls_session doesn't work because GnuTLS 2.x.y
comes with a compat.h that defines gnutls_session to gnutls_session_t.
Instead detect this based on LIBGNUTLS_VERSION_MAJOR.
---
configure.ac | 22 +++++++++++++++-------
1 files changed, 15 insertions(+), 7 deletions(-)
diff --git a/configure.ac b/configure.ac
index c8d291b..6a20366 100644
--- a/configure.ac
+++ b/configure.ac
@@ -834,17 +834,25 @@ fi
AC_SUBST([GNUTLS_CFLAGS])
AC_SUBST([GNUTLS_LIBS])
-dnl Old versions of GnuTLS uses types like 'gnutls_session' instead
-dnl of 'gnutls_session_t'. Try to detect this type if defined so
-dnl that we can offer backwards compatibility.
+dnl Detect GnuTLS 1.x.y and enable backwards compatibility macros if found.
old_cflags="$CFLAGS"
old_libs="$LIBS"
CFLAGS="$CFLAGS $GNUTLS_CFLAGS"
LIBS="$LIBS $GNUTLS_LIBS"
-AC_CHECK_TYPE([gnutls_session],
- AC_DEFINE([GNUTLS_1_0_COMPAT],[],
- [enable GnuTLS 1.0 compatibility macros]),,
- [#include <gnutls/gnutls.h>])
+AC_MSG_CHECKING([whether GnuTLS 1.0 compatibility macros are required])
+AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <gnutls/gnutls.h>]],
+ [[
+ #if LIBGNUTLS_VERSION_MAJOR < 2
+ # error gnutls 1.x.y detected
+ #endif
+ ]])],
+ [have_gnutlsv1=no],
+ [have_gnutlsv1=yes])
+if test "x$have_gnutlsv1" = xyes; then
+ AC_DEFINE_UNQUOTED([GNUTLS_1_0_COMPAT], 1,
+ [enable GnuTLS 1.0 compatibility macros])
+fi
+AC_MSG_RESULT([$have_gnutlsv1])
CFLAGS="$old_cflags"
LIBS="$old_libs"
--
1.7.4.1
9:48 a.m.
On 08/03/2011 08:22 AM, Matthias Bolte wrote:
Detection based on gnutls_session doesn't work because GnuTLS
2.x.y
comes with a compat.h that defines gnutls_session to gnutls_session_t.
Instead detect this based on LIBGNUTLS_VERSION_MAJOR.
---
configure.ac | 22 +++++++++++++++-------
1 files changed, 15 insertions(+), 7 deletions(-)
ACK for correctness.
However, I wonder if it is the most efficient. We aren't using this as
a Makefile conditional, and LIBGNUTLS_VERSION_MAJOR is available at
compile time. That is, I think that we could skip the configure check
altogether, and just move the #if LIBGNUTLS_VERSION_MAJOR < 2 check into
a common header used by all files that want to use gnutls in the first
place, for a slightly smaller and faster configure.
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library http://libvirt.org
11:28 a.m.
2011/8/3 Eric Blake <eblake(a)redhat.com>:
On 08/03/2011 08:22 AM, Matthias Bolte wrote:
>
> Detection based on gnutls_session doesn't work because GnuTLS 2.x.y
> comes with a compat.h that defines gnutls_session to gnutls_session_t.
>
> Instead detect this based on LIBGNUTLS_VERSION_MAJOR.
> ---
> configure.ac | 22 +++++++++++++++-------
> 1 files changed, 15 insertions(+), 7 deletions(-)
ACK for correctness.
However, I wonder if it is the most efficient. We aren't using this as a
Makefile conditional, and LIBGNUTLS_VERSION_MAJOR is available at compile
time. That is, I think that we could skip the configure check altogether,
and just move the #if LIBGNUTLS_VERSION_MAJOR < 2 check into a common header
used by all files that want to use gnutls in the first place, for a slightly
smaller and faster configure.
Here's a v2 that does this.
--
Matthias Bolte
http://photron.blogspot.com
11:31 a.m.
On 08/03/2011 10:28 AM, Matthias Bolte wrote:
2011/8/3 Eric Blake<eblake(a)redhat.com>:
> On 08/03/2011 08:22 AM, Matthias Bolte wrote:
>>
>> Detection based on gnutls_session doesn't work because GnuTLS 2.x.y
>> comes with a compat.h that defines gnutls_session to gnutls_session_t.
>>
>> Instead detect this based on LIBGNUTLS_VERSION_MAJOR.
>> ---
>> configure.ac | 22 +++++++++++++++-------
>> 1 files changed, 15 insertions(+), 7 deletions(-)
>
> ACK for correctness.
>
> However, I wonder if it is the most efficient. We aren't using this as a
> Makefile conditional, and LIBGNUTLS_VERSION_MAJOR is available at compile
> time. That is, I think that we could skip the configure check altogether,
> and just move the #if LIBGNUTLS_VERSION_MAJOR< 2 check into a common header
> used by all files that want to use gnutls in the first place, for a slightly
> smaller and faster configure.
Here's a v2 that does this.
Nicer - thanks for that work.
Detection based on gnutls_session doesn't work because GnuTLS
2.x.y
comes with a compat.h that defines gnutls_session to gnutls_session_t.
Instead detect this based on LIBGNUTLS_VERSION_MAJOR. Move this from
configure/config.h to gnutls_1_0_compat.h and make sure that all users
include gnutls_1_0_compat.h properly.
Also fix header guard in gnutls_1_0_compat.h.
ACK.
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library http://libvirt.org
12:27 p.m.
2011/8/3 Eric Blake <eblake(a)redhat.com>:
On 08/03/2011 10:28 AM, Matthias Bolte wrote:
>
> 2011/8/3 Eric Blake<eblake(a)redhat.com>:
>>
>> On 08/03/2011 08:22 AM, Matthias Bolte wrote:
>>>
>>> Detection based on gnutls_session doesn't work because GnuTLS 2.x.y
>>> comes with a compat.h that defines gnutls_session to gnutls_session_t.
>>>
>>> Instead detect this based on LIBGNUTLS_VERSION_MAJOR.
>>> ---
>>> configure.ac | 22 +++++++++++++++-------
>>> 1 files changed, 15 insertions(+), 7 deletions(-)
>>
>> ACK for correctness.
>>
>> However, I wonder if it is the most efficient. We aren't using this as a
>> Makefile conditional, and LIBGNUTLS_VERSION_MAJOR is available at compile
>> time. That is, I think that we could skip the configure check
>> altogether,
>> and just move the #if LIBGNUTLS_VERSION_MAJOR< 2 check into a common
>> header
>> used by all files that want to use gnutls in the first place, for a
>> slightly
>> smaller and faster configure.
>
> Here's a v2 that does this.
Nicer - thanks for that work.
> Detection based on gnutls_session doesn't work because GnuTLS 2.x.y
> comes with a compat.h that defines gnutls_session to gnutls_session_t.
>
> Instead detect this based on LIBGNUTLS_VERSION_MAJOR. Move this from
> configure/config.h to gnutls_1_0_compat.h and make sure that all users
> include gnutls_1_0_compat.h properly.
>
> Also fix header guard in gnutls_1_0_compat.h.
ACK.
Thanks, pushed.
--
Matthias Bolte
http://photron.blogspot.com
4891
days inactive
4891
days old
4 comments
2 participants
participants (2)
-
Eric Blake -
Matthias Bolte