Re: [PATCH] fstests: fix blksize_t printf format warnings across architectures

[Qu Wenruo <wqu@suse.com>]

在 2024/11/21 08:51, Darrick J. Wong 写道:
> On Thu, Nov 21, 2024 at 08:36:58AM +1030, Qu Wenruo wrote:
>>
>>
>> 在 2024/11/20 22:10, Anand Jain 写道:
>>> Fix format string warnings when printing blksize_t values that vary
>>> across architectures. The warning occurs because blksize_t is defined
>>> differently between architectures: aarch64 architectures blksize_t is
>>> int, on x86-64 it's long-int.  Cast the values to long. Fixes warnings
>>> as below.
>>>
>>>    seek_sanity_test.c:110:45: warning: format '%ld' expects argument of type
>>>    'long int', but argument 3 has type 'blksize_t' {aka 'int'}
>>>
>>>    attr_replace_test.c:70:22: warning: format '%ld' expects argument of type
>>>    'long int', but argument 3 has type '__blksize_t' {aka 'int'}
>>
>> Why not just use %zu instead?
> 
>  From printf(3):
> 
>         z      A  following  integer conversion corresponds to a size_t
>                or ssize_t argument, or a following n conversion  corre‐
>                sponds to a pointer to a size_t argument.
> 
> blksize_t is not a ssize_t.

You're right, it's indeed different and it's more complex than I thought.

blksize_t is __SYSCALL_SLONG_TYPE, which has extra handling on x86_64 
with its x32 mode handling.
Only for x86_64 x32 mode it is __SQUAD_TYPE (depending on wordsize 
again) otherwise it's __SLONGWORD_TYPE (fixed to long int).

Meanwhile ssize_t is __SWORD_TYPE, which is only dependent on word size.
For 32bit word size it's __int64_t, and for 64bit it's long int.

So blksize_t is more weird than ssize_t.

Then force converting to long (int) is fine.

> 
> --D
> 
>> Thanks,
>> Qu
>>
>>>
>>> Signed-off-by: Anand Jain <anand.jain@oracle.com>
>>> ---
>>>    src/attr_replace_test.c | 2 +-
>>>    src/seek_sanity_test.c  | 2 +-
>>>    2 files changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/src/attr_replace_test.c b/src/attr_replace_test.c
>>> index 1218e7264c8f..5d560a633361 100644
>>> --- a/src/attr_replace_test.c
>>> +++ b/src/attr_replace_test.c
>>> @@ -67,7 +67,7 @@ int main(int argc, char *argv[])
>>>    	if (ret < 0) die();
>>>    	size = sbuf.st_blksize * 3 / 4;
>>>    	if (!size)
>>> -		fail("Invalid st_blksize(%ld)\n", sbuf.st_blksize);
>>> +		fail("Invalid st_blksize(%ld)\n", (long)sbuf.st_blksize);

Although for this case, I'd prefer to use "long int" just for the sake 
of consistency.

Otherwise looks good to me.

Reviewed-by: Qu Wenruo <wqu@suse.com>

Thanks,
Qu

>>>    	size = MIN(size, maxsize);
>>>    	value = malloc(size);
>>>    	if (!value)
>>> diff --git a/src/seek_sanity_test.c b/src/seek_sanity_test.c
>>> index a61ed3da9a8f..c5930357911f 100644
>>> --- a/src/seek_sanity_test.c
>>> +++ b/src/seek_sanity_test.c
>>> @@ -107,7 +107,7 @@ static int get_io_sizes(int fd)
>>>    		offset += pos ? 0 : 1;
>>>    	alloc_size = offset;
>>>    done:
>>> -	fprintf(stdout, "Allocation size: %ld\n", alloc_size);
>>> +	fprintf(stdout, "Allocation size: %ld\n", (long)alloc_size);
>>>    	return 0;
>>>
>>>    fail:
>>
>>
>