rand_seed declarations not consistent...is that ok?

rand_seed declarations not consistent...is that ok?

[Grant Grundler]

cscope tells me:
C symbol: rand_seed

  File             Function                  Line
0 ioengine.h       <global>                    59 unsigned long rand_seed;
1 options.c        <global>                  1660 .off1 =
                                                  td_var_offset(rand_seed),
2 thread_options.h <global>                   103 unsigned int rand_seed;
3 thread_options.h <global>                   327 uint32_t rand_seed;
4 verify.h         <global>                    34 uint64_t rand_seed;

Any reason we have 4 different declarations of the same name?
Thoughts on if/how this should be fixed?

Even if it's not wrong, it's definitely confusing.

thanks,
grant

Re: rand_seed declarations not consistent...is that ok?

[Jens Axboe]

On Tue, Jan 28 2014, Grant Grundler wrote:
> cscope tells me:
> C symbol: rand_seed
> 
>   File             Function                  Line
> 0 ioengine.h       <global>                    59 unsigned long rand_seed;
> 1 options.c        <global>                  1660 .off1 =
>                                                   td_var_offset(rand_seed),
> 2 thread_options.h <global>                   103 unsigned int rand_seed;
> 3 thread_options.h <global>                   327 uint32_t rand_seed;
> 4 verify.h         <global>                    34 uint64_t rand_seed;
> 
> Any reason we have 4 different declarations of the same name?
> Thoughts on if/how this should be fixed?
> 
> Even if it's not wrong, it's definitely confusing.

Heh no. But they are not all the same. Could be cleaned up, but the
parser has some expectations and the thread_options.h one has the local
and over-the-net client version (which is why they are different, one
matches the parser and the other is independent). I would suggest we
make 0/3/4 a uint64_t to unify that part. The option could be made
unsigned long long and the type changed.

-- 
Jens Axboe

Re: rand_seed declarations not consistent...is that ok?

[Grant Grundler]

On Tue, Jan 28, 2014 at 11:24 AM, Jens Axboe <axboe@kernel.dk> wrote:
> On Tue, Jan 28 2014, Grant Grundler wrote:
>> cscope tells me:
>> C symbol: rand_seed
>>
>>   File             Function                  Line
>> 0 ioengine.h       <global>                    59 unsigned long rand_seed;
>> 1 options.c        <global>                  1660 .off1 =
>>                                                   td_var_offset(rand_seed),
>> 2 thread_options.h <global>                   103 unsigned int rand_seed;
>> 3 thread_options.h <global>                   327 uint32_t rand_seed;
>> 4 verify.h         <global>                    34 uint64_t rand_seed;
>>
>> Any reason we have 4 different declarations of the same name?
>> Thoughts on if/how this should be fixed?
>>
>> Even if it's not wrong, it's definitely confusing.
>
> Heh no. But they are not all the same. Could be cleaned up, but the
> parser has some expectations and the thread_options.h one has the local
> and over-the-net client version (which is why they are different, one
> matches the parser and the other is independent).

I gathered there were some differences. :)

> I would suggest we make 0/3/4 a uint64_t to unify that part.

That's easy. Patch sent: [PATCH] fio: consolidate rand_seed to uint64_t
Did I interpret that correctly?

BTW, I've not tested that patch...will do that and reply to the patch
if there are any new issues.


> The option could be made
> unsigned long long and the type changed.

I think this needs a bit more research (but only very little). Anyone
else want a relatively quick/easy project?
I don't need all the glory. :)

cheers,
grant