Re: [Qemu-devel] [PATCH v7 11/35] util: introduce qemu_file_getlength()

[Xiao Guangrong <guangrong.xiao@linux.intel.com>]

On 11/04/2015 10:44 PM, Eduardo Habkost wrote:
> On Wed, Nov 04, 2015 at 11:17:09AM +0800, Xiao Guangrong wrote:
>>
>>
>> On 11/04/2015 07:21 AM, Eduardo Habkost wrote:
>>> On Mon, Nov 02, 2015 at 05:13:13PM +0800, Xiao Guangrong wrote:
>>> [...]
>>>> +size_t qemu_file_getlength(const char *file, Error **errp)
>>>> +{
>>>> +    int64_t size;
>>> [...]
>>>> +    return size;
>>>
>>> Can you guarantee that SIZE_MAX >= INT64_MAX on all platforms supported
>>> by QEMU?
>>>
>>
>> Actually, this function is abstracted from the common function, raw_getlength(),
>> in raw-posix.c whose return value is int64_t.
>>
>> And i think int64_t is large enough for block devices.
>
> int64_t should be enough, but I don't know if size_t is large enough on
> all platforms.
>
> I believe it's going to be either one of those cases:
>
> * If you are absolutely sure SIZE_MAX >= INT64_MAX on all platforms,
>    please explain why (and maybe add a QEMU_BUILD_BUG_ON?). (I don't
>    think this will be the case)
> * If SIZE_MAX < INT64_MAX is possible but you believe
>    size <= SIZE_MAX is always true here, please explain why (and maybe
>    add an assert()).
> * Otherwise, we need to set an appropriate error if size > SIZE_MAX
>    or change the type of qemu_file_getlength(). What about making it
>    uint64_t?
>

It sounds better, I will change the return value from size_t to uint64_t.

Thank you for pointing it out, Eduardo!