On 11/17/2010 07:32 AM, Daniel P. Berrange wrote:
On Tue, Nov 16, 2010 at 10:53:03AM -0700, Eric Blake wrote:
> In some cases, we were just plain confusing (BINDIR expanded to
> $(libexecdir), GETTEXT_PACKAGE redefined PACKAGE for no apparant
> reason); in other cases, we might as well benefit from gnulib doing
> the work for us (PKGDATADIR), sometimes with different spellings
> (LOCAL_STATE_DIR vs. LOCALSTATEDIR).
>
> * bootstrap.conf (gnulib_modules): Add configmake.
> * daemon/Makefile.am (libvirtd_CFLAGS): Drop defines provided by
> gnulib.
> 26 files changed, 79 insertions(+), 72 deletions(-)
ACK
Pushing as is. If we decide to do more cleanups, they can come as later
patches.
> +++ b/src/secret/secret_driver.c
> @@ -42,6 +42,7 @@
> #include "uuid.h"
> #include "virterror_internal.h"
> #include "files.h"
> +#include "configmake.h"
Should we just put that in 'internal.h' so all source files
(in src/ at least) get it ? We'd still need it separately
in some places
The documentation for "configmake.h" recommends that it be included
after all system headers. In particular, "internal.h" is included prior
to <windows.h> in util/threads-win32.h, and I'm afraid that moving
"configmake.h" into "internal.h" would break a mingw build, since
<windows.h> is notorious for littering the namespace with conflicting
defines. However, I have not tested that theory; would you like me to
try that as a possible subsequent cleanup?
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library
http://libvirt.org