On 11/11/2015 02:27 PM, Eduardo Habkost wrote:
> Instead of implementing separate check functions for each vga
> interface type, add a table enumerating the possible VGA
> interfaces.
>
> Signed-off-by: Eduardo Habkost <ehabk...@redhat.com>
> ---
> include/sysemu/sysemu.h | 1 +
> vl.c | 114
> ++++++++++++++++++++++++++----------------------
> 2 files changed, 63 insertions(+), 52 deletions(-)
>
> diff --git a/include/sysemu/sysemu.h b/include/sysemu/sysemu.h
> index f992494..6406906 100644
> --- a/include/sysemu/sysemu.h
> +++ b/include/sysemu/sysemu.h
> @@ -147,6 +147,7 @@ extern int autostart;
> typedef enum {
> VGA_NONE, VGA_STD, VGA_CIRRUS, VGA_VMWARE, VGA_XENFB, VGA_QXL,
> VGA_TCX, VGA_CG3, VGA_DEVICE, VGA_VIRTIO,
> + VGA_TYPE_MAX,
> } VGAInterfaceType;
Would it be worth exposing this enum in qapi someday? But doesn't affect
the correctness of this patch.
> static void select_vgahw (const char *p)
Worth dropping the space before '(' while in the neighborhood?
> {
> const char *opts;
> + int t;
>
> - assert(vga_interface_type == VGA_NONE);
Are you intentionally dropping the assert? It protects us from
select_vgahw() being called more than once.
> - } else if (strstart(p, "cg3", &opts)) {
> - if (cg3_vga_available()) {
> - vga_interface_type = VGA_CG3;
> - } else {
> - error_report("CG3 framebuffer not available");
> - exit(1);
> + for (t = 0; t < VGA_TYPE_MAX; t++) {
> + VGAInterfaceInfo *ti = &vga_interfaces[t];
> + if (ti->opt_name && strstart(p, ti->opt_name, &opts)) {
> + if (ti->available && !ti->available()) {
> + error_report("%s not available", ti->name);
> + exit(1);
> + }
> + vga_interface_type = t;
> + break;
Nice compression in code size.
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
