Skip to contentSkip to conversation

qom: Rename Object::class into Object::klass

PATCH In review Felix Wu 4 participants 6 messages Jun 24, 2024 – Jun 25, 2024

lore
Share & tools
lore

Felix Wu

author
#0 lore raw

PATCH 1/2 qom: Rename Object::class into Object::klass

From: Roman Kiryanov <[email protected]>

'class' is a C++ keyword and it prevents from
using the QEMU headers with a C++ compiler.

Google-Bug-Id: 331190993

Change-Id I9ab7d2d77edef654a9c7b7cb9cd01795a6ed65a2
---
 hw/core/qdev-properties-system.c |  2 +-
 include/exec/memory.h            |  2 +-
 include/qom/object.h             |  2 +-
 qom/object.c                     | 90 ++++++++++++++++----------------
 4 files changed, 48 insertions(+), 48 deletions(-)
hw/core/qdev-properties-system.c +1 -1
index f13350b4fb..a6781841af 100644
@@ -431,7 +431,7 @@ static void set_netdev(Object *obj, Visitor *v, const char *name,
         }
 
         if (peers[i]->info->check_peer_type) {
-            if (!peers[i]->info->check_peer_type(peers[i], obj->class, errp)) {
+            if (!peers[i]->info->check_peer_type(peers[i], obj->klass, errp)) {
                 goto out;
             }
         }
include/exec/memory.h +1 -1
index 2d7c278b9f..e5bd75956e 100644
@@ -1808,7 +1808,7 @@ static inline IOMMUMemoryRegion *memory_region_get_iommu(MemoryRegion *mr)
 static inline IOMMUMemoryRegionClass *memory_region_get_iommu_class_nocheck(
         IOMMUMemoryRegion *iommu_mr)
 {
-    return (IOMMUMemoryRegionClass *) (((Object *)iommu_mr)->class);
+    return (IOMMUMemoryRegionClass *) (((Object *)iommu_mr)->klass);
 }
 
 #define memory_region_is_iommu(mr) (memory_region_get_iommu(mr) != NULL)
include/qom/object.h +1 -1
index 13d3a655dd..7afdb261a8 100644
@@ -153,7 +153,7 @@ struct ObjectClass
 struct Object
 {
     /* private: */
-    ObjectClass *class;
+    ObjectClass *klass;
     ObjectFree *free;
     GHashTable *properties;
     uint32_t ref;
qom/object.c +45 -45

254 lines not on this page · raw message

Signature · 1 line
2.45.2.741.gdbec12cfda-goog
author
#1 lore raw

PATCH 2/2 include/qom: Rename typename into type_name

From: Roman Kiryanov <[email protected]>

typename is a C++ keyword and it prevents from
using the QEMU headers with a C++ compiler.

Google-Bug-Id: 331190993

Change-Id Iff313ca5ec157a1a3826b4f5665073534d961a26
---
 hw/core/bus.c          |   8 +--
 include/hw/qdev-core.h |   4 +-
 include/qom/object.h   |  78 +++++++++++++--------------
 qom/object.c           | 120 ++++++++++++++++++++---------------------
 4 files changed, 105 insertions(+), 105 deletions(-)
hw/core/bus.c +4 -4
index b9d89495cd..07c5a83673 100644
@@ -152,18 +152,18 @@ static void bus_unparent(Object *obj)
     bus->parent = NULL;
 }
 
-void qbus_init(void *bus, size_t size, const char *typename,
+void qbus_init(void *bus, size_t size, const char *type_name,
                DeviceState *parent, const char *name)
 {
-    object_initialize(bus, size, typename);
+    object_initialize(bus, size, type_name);
     qbus_init_internal(bus, parent, name);
 }
 
-BusState *qbus_new(const char *typename, DeviceState *parent, const char *name)
+BusState *qbus_new(const char *type_name, DeviceState *parent, const char *name)
 {
     BusState *bus;
 
-    bus = BUS(object_new(typename));
+    bus = BUS(object_new(type_name));
     qbus_init_internal(bus, parent, name);
 
     return bus;
include/hw/qdev-core.h +2 -2
index 5336728a23..ede4b74bd8 100644
@@ -867,9 +867,9 @@ DeviceState *qdev_find_recursive(BusState *bus, const char *id);
 typedef int (qbus_walkerfn)(BusState *bus, void *opaque);
 typedef int (qdev_walkerfn)(DeviceState *dev, void *opaque);
 
-void qbus_init(void *bus, size_t size, const char *typename,
+void qbus_init(void *bus, size_t size, const char *type_name,
                DeviceState *parent, const char *name);
-BusState *qbus_new(const char *typename, DeviceState *parent, const char *name);
+BusState *qbus_new(const char *type_name, DeviceState *parent, const char *name);
 bool qbus_realize(BusState *bus, Error **errp);
 void qbus_unrealize(BusState *bus);
 
include/qom/object.h +39 -39

242 lines not on this page · raw message

qom/object.c +60 -60

389 lines not on this page · raw message

Signature · 1 line
2.45.2.741.gdbec12cfda-goog
Reviewed-by on v1 1/2
#2 lore raw
In reply to Felix Wu

Hi Felix,

Quoted 6 lines · Felix Wu

From: Roman Kiryanov <[email protected]>

'class' is a C++ keyword and it prevents from
using the QEMU headers with a C++ compiler.

Google-Bug-Id: 331190993

I asked Roman twice about this tag meaning:
https://lore.kernel.org/qemu-devel/[email protected]/
https://lore.kernel.org/qemu-devel/[email protected]/
Since you are taking his work, do you mind clarifying?

Please include a cover letter for your series:
https://www.qemu.org/docs/master/devel/submitting-a-patch.html#include-a-meaningful-cover-letter

Also for headers refactors, enabling scripts/git.orderfile helps reviewers.

Since you are posting different C++ enablement cleanups, I suggest you add a section in our docs/devel/style.rst requesting to keep headers C++ compatible, by not using C++ reserved keywords, etc...
In particular because the mainstream project is not build-testing for C++, thus we will likely merge patches breaking C++ and make your life harder. That said, a C++ header smoke-build job in our CI could help.

Quoted 16 lines
Change-Id I9ab7d2d77edef654a9c7b7cb9cd01795a6ed65a2
---
  hw/core/qdev-properties-system.c |  2 +-
  include/exec/memory.h            |  2 +-
  include/qom/object.h             |  2 +-
  qom/object.c                     | 90 ++++++++++++++++----------------
  4 files changed, 48 insertions(+), 48 deletions(-)
include/qom/object.h +0 -0
index 13d3a655dd..7afdb261a8 100644

include/qom/object.h struct ObjectClass · line 159

Quoted 8 lines
include/qom/object.h +1 -1
@@ -153,7 +153,7 @@ struct ObjectClass
  struct Object
  {
      /* private: */
-    ObjectClass *class;
+    ObjectClass *klass;
      ObjectFree *free;
      GHashTable *properties;
      uint32_t ref;

I suppose the OBJECT_CLASS / OBJECT_CLASS_CHECK / OBJECT_GET_CLASS macros aren't compiled so "class" isn't a problem there.

Since it isn't worst than our INTERFACE_CLASS() use:

Regards,

Phil.

#3 lore raw
Quoted 9 lines · Philippe Mathieu-Daudé

Since you are posting different C++ enablement cleanups, I suggest you add a section in our docs/devel/style.rst requesting to keep headers C++ compatible, by not using C++ reserved keywords, etc...

In particular because the mainstream project is not build-testing for C++, thus we will likely merge patches breaking C++ and make your life harder. That said, a C++ header smoke-build job in our CI could help.

Unless there's some easy mechanism for contributors to check that they haven't broken whatever our C++ requirement is, I don't think we should define it in the style guide.

More generally, we specifically removed the handling we had for being able to include our headers from C++ source files. (cf the stuff we added in commit 875df03b221 for extern "C" blocks and then removed again later). If we're not bringing that back (and I don't think we should) then we're not actually trying to have our headers be C++ compatible, so what are we aiming for?

thanks
-- PMM

#4 lore raw
Quoted 15 lines · Peter Maydell
Quoted 9 lines · Philippe Mathieu-Daudé

Since you are posting different C++ enablement cleanups, I suggest you add a section in our docs/devel/style.rst requesting to keep headers C++ compatible, by not using C++ reserved keywords, etc...

In particular because the mainstream project is not build-testing for C++, thus we will likely merge patches breaking C++ and make your life harder. That said, a C++ header smoke-build job in our CI could help.

Unless there's some easy mechanism for contributors to check that they haven't broken whatever our C++ requirement is, I don't think we should define it in the style guide.

Quoted 7 lines

More generally, we specifically removed the handling we had for being able to include our headers from C++ source files. (cf the stuff we added in commit 875df03b221 for extern "C" blocks and then removed again later). If we're not bringing that back (and I don't think we should) then we're not actually trying to have our headers be C++ compatible, so what are we aiming for?

I really dislike the drip-feeding of patches fixing C++ related problems. As maintainers we've no idea what the end state is, is this the last patch, or are there another 100 of these patches to trickle out one at a time. Ultimately from the QEMU maintainer POV anything related to C++ compatibility is a distraction, given the general consensus has turned to Rust as the future for QEMU, not C++.

If we're going to take any C++ compat cleanups as a courtesy to ease burden of a downstream fork, then I'd like to see a complete series in one go, so we can sensibly evaluate whether the end state is something desirable from QEMU's POV.

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#5 lore raw
In reply to Felix Wu
Quoted 7 lines · Felix Wu

From: Roman Kiryanov <[email protected]>

'class' is a C++ keyword and it prevents from
using the QEMU headers with a C++ compiler.

Google-Bug-Id: 331190993
Change-Id: I9ab7d2d77edef654a9c7b7cb9cd01795a6ed65a2

Please remove both of these lines when posting patches. They are irrelevant to QEMU's git commit history.

Quoted 8 lines
---
 hw/core/qdev-properties-system.c |  2 +-
 include/exec/memory.h            |  2 +-
 include/qom/object.h             |  2 +-
 qom/object.c                     | 90 ++++++++++++++++----------------
 4 files changed, 48 insertions(+), 48 deletions(-)

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|