On 09/12/2016 04:18 AM, Marc-André Lureau wrote: > There are better chances to find what went wrong at build time than a > later assert in qmp_query_version > > Signed-off-by: Marc-André Lureau <marcandre.lur...@redhat.com> > --- > qmp.c | 16 +++------------- > scripts/create_config | 6 ++++++ > 2 files changed, 9 insertions(+), 13 deletions(-) >
> -
> - err = qemu_strtoll(tmp, &tmp, 10, &info->qemu->micro);
> - assert(err == 0);
> + info->qemu->major = QEMU_VERSION_MAJOR;
> + info->qemu->minor = QEMU_VERSION_MINOR;
> + info->qemu->micro = QEMU_VERSION_MICRO;
> info->package = g_strdup(QEMU_PKGVERSION);
>
> return info;
The old code silently ignores any garbage after the third integer (it
populates &tmp, but never checks the value of tmp).
> diff --git a/scripts/create_config b/scripts/create_config
> index 1dd6a35..e6929dd 100755
> --- a/scripts/create_config
> +++ b/scripts/create_config
> @@ -7,7 +7,13 @@ while read line; do
> case $line in
> VERSION=*) # configuration
> version=${line#*=}
> + major=$(echo "$version" | cut -d. -f1)
> + minor=$(echo "$version" | cut -d. -f2)
> + micro=$(echo "$version" | cut -d. -f3)
> echo "#define QEMU_VERSION \"$version\""
> + echo "#define QEMU_VERSION_MAJOR $major"
> + echo "#define QEMU_VERSION_MINOR $minor"
> + echo "#define QEMU_VERSION_MICRO $micro"
The new code likewise ignores any fourth field.
Do we care either way? Unless someone else has a reason for why we
should care, I'm fine with:
Reviewed-by: Eric Blake <ebl...@redhat.com>
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
