Re: [PATCH 3/7] hw/ide: Move IDE device related definitions to ide-dev.h

[Thomas Huth <thuth@redhat.com>]

On 19/02/2024 12.32, Philippe Mathieu-Daudé wrote:
> On 19/2/24 11:49, Thomas Huth wrote:
>> Let's start to unentangle internal.h by moving public IDE device
>> related definitions to ide-dev.h.
>>
>> Signed-off-by: Thomas Huth <thuth@redhat.com>
>> ---
>>   include/hw/ide/ide-dev.h  | 145 +++++++++++++++++++++++++++++++++++++-
>>   include/hw/ide/internal.h | 145 +-------------------------------------
>>   hw/ide/ide-dev.c          |   1 +
>>   3 files changed, 146 insertions(+), 145 deletions(-)
>>
>> diff --git a/include/hw/ide/ide-dev.h b/include/hw/ide/ide-dev.h
>> index 7e9663cda9..de88784a25 100644
>> --- a/include/hw/ide/ide-dev.h
>> +++ b/include/hw/ide/ide-dev.h
>> @@ -20,9 +20,152 @@
>>   #ifndef IDE_DEV_H
>>   #define IDE_DEV_H
>> +#include "sysemu/dma.h"
> 
> Not required.

It's required for QEMUSGList that is used in struct IDEState.

>>   #include "hw/qdev-properties.h"
>>   #include "hw/block/block.h"
>> -#include "hw/ide/internal.h"
>> +
>> +typedef struct IDEDevice IDEDevice;
>> +typedef struct IDEState IDEState;
> 
>> +typedef struct IDEDMA IDEDMA;
>> +typedef struct IDEDMAOps IDEDMAOps;
>> +typedef struct IDEBus IDEBus;
> 
> Looking at next patches, better forward-declare IDEBus and
> IDEDMA in "qemu/typedefs.h".

I really dislike using qemu/typedefs.h for things that are not really part 
of the core framework, since it's a 
touch-it-once-and-everything-gets-recompiled header. So IMHO the typedefs 
here are the lesser evil.

> IDEDMAOps and "sysemu/dma.h" belong to "hw/ide/ide-dma.h.

Ok, I can move the typedef for IDEDMAOps to ide-dma.h instead.

  Thomas