From: Blue Swirl <blauwirbel@gmail.com>
To: Anthony Liguori <aliguori@us.ibm.com>
Cc: Gerd Hoffman <kraxel@redhat.com>,
Paul Brook <paul@codesourcery.com>,
qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH 2/2] Introduce macro for defining qdev properties
Date: Fri, 17 Jul 2009 21:32:02 +0300 [thread overview]
Message-ID: <f43fc5580907171132x71f65c80pae06a889b4b240ff@mail.gmail.com> (raw)
In-Reply-To: <4A60BEEF.2050601@us.ibm.com>
On Fri, Jul 17, 2009 at 9:11 PM, Anthony Liguori<aliguori@us.ibm.com> wrote:
> Paul Brook wrote:
>>
>> I think I agree. Using GCC extensions for error checking (e.g. DO_UPCAST)
>> or performance (__builtin_clz) is fine, but I'm a reluctant to rely on it
>> for correct operation.
>>
>
> Practically speaking, we've never supported anything but GCC and I doubt we
> ever will. In this case, it's an important part of something I'm trying to
> fix about the current property system.
>
> It seems very brittle to me. You have to specify a type in both the state
> structure and in the property definition. Those two things are very, very
> far apart in the code. Right now, the rules about type compatible are ill
> defined which makes it more likely to break beyond simple mistakes. For
> instance, uint32 is used for uint32_t, int32_t, and int. That seems odd.
>
> I also don't like the fact that we mix field type information with display
> information. I haven't thought about the best solution to this but I think
> it's either introducing new struct types or adding an optional decorator
> parameter.
>
> The system I'm aiming for looks like this:
>
> typedef struct {
> SysBusDevice parent;
>
> /* public */
> uint32_t queue_depth;
> uint32_t tx_mitigation_delay;
> CharDriverState *chr;
>
> /* private */
> ...
> } MyDeviceState;
>
> static Property my_device_properties[] = {
> QDEV_PROP(MyDeviceState, queue_depth),
> QDEV_PROP(MyDeviceState, tx_mitigation_delay),
> QDEV_PROP(MyDeviceState, chr),
> {}
> };
>
> Where there's a connection between properties and device state fields and
> there's no duplicate type information. That means that for the most part,
> the rules of type compatible can be ignored by most users.
>
> I'd like to see most uses of QDEV_PROP_NAME eliminated by renaming variables
> and accepting '-' in place of '_'. We'll always need a way to accept
> default values.
>
> I'm not sure how to do this without GCC extensions. We could potentially
> add macro decorators and use a sparse-like tool to extract property lists
> automatically from device state.
Then there is the template way:
typedef struct MyDeviceState {
#define PROP(type, name) type name;
#define C_TYPE
#include "mydevice_props.hx"
} MyDeviceState;
#undef PROP
#undef C_TYPE
static Property my_device_properties[] = {
#define PROP(type, name) glue(QDEV_PROP_, type)(MyDeviceState, name),
#include "mydevice_props.hx"
{}
};
Where mydevice_props.hx contains:
#ifdef C_TYPE
#define I32 uint32_t
#define I64 uint64_t
#define CHRDEV CharDriverState *
#endif
PROP(I32, queue_depth)
PROP(I32, tx_mitigation_delay)
PROP(CHRDEV, chr)
next prev parent reply other threads:[~2009-07-17 18:32 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-07-17 14:41 [Qemu-devel] [PATCH 0/2][RFC] Introduce macros for setting properties Anthony Liguori
2009-07-17 14:41 ` [Qemu-devel] [PATCH 1/2] Introduce CharDriverState qdev property type Anthony Liguori
2009-07-21 14:46 ` [Qemu-devel] " Gerd Hoffmann
2009-07-21 14:53 ` Anthony Liguori
2009-07-21 15:03 ` Gerd Hoffmann
2009-07-21 15:14 ` Gerd Hoffmann
2009-07-21 15:23 ` Anthony Liguori
2009-07-21 15:34 ` Gerd Hoffmann
2009-07-21 15:47 ` Anthony Liguori
2009-07-22 9:48 ` Gerd Hoffmann
2009-07-17 14:41 ` [Qemu-devel] [PATCH 2/2] Introduce macro for defining qdev properties Anthony Liguori
2009-07-17 17:23 ` Blue Swirl
2009-07-17 17:26 ` Anthony Liguori
2009-07-17 17:33 ` Paul Brook
2009-07-17 18:11 ` Anthony Liguori
2009-07-17 18:32 ` Blue Swirl [this message]
2009-07-17 20:05 ` Anthony Liguori
2009-07-17 22:58 ` Paul Brook
2009-07-18 12:43 ` Jamie Lokier
2009-07-18 16:13 ` Anthony Liguori
2009-07-20 2:29 ` Jamie Lokier
2009-07-21 8:30 ` [Qemu-devel] " Gerd Hoffmann
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=f43fc5580907171132x71f65c80pae06a889b4b240ff@mail.gmail.com \
--to=blauwirbel@gmail.com \
--cc=aliguori@us.ibm.com \
--cc=kraxel@redhat.com \
--cc=paul@codesourcery.com \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).