Re: [PATCH 02/15] mdadm/Grow: fix coverity issue CHECKED_RETURN

[Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>]

On Thu, 18 Jul 2024 10:29:50 +0800
Xiao Ni <xni@redhat.com> wrote:

> > lseek64 errors are unusual, they are exceptional, and I'm fine with logging
> > same error message but I would prefer to avoid repeating same message in
> > code. In case of debug, developer can do some backtracking, starting from
> > this function rather than hunt for the particular error message you used
> > multiple times.  
> 
> Hi Mariusz
> 
> If we use the above way, pr_err only prints the line of code in the
> function lseek64_log_err. We can't know where the error is. pr_err
> prints the function name and code line number. We put pr_err in the
> place where lseek fails, we can know which function and which lseek
> fails. It should be easy for debug. Is it right?
> >  

Oh right, I forgot about this from simple reason - I'm never using this.
Line is printed with make CXFLAGS=DDEBUG. If you will enable DEBUG flag then
mdadm will also spam with debug messages (prr_dbg messages). Anyway, you are
right it is an option.

Enabling DDEBUG in the past many times broke my time sensitive reproductions. I
don't like it.

From the top of my head, there is a rule "do not repeat same code". If we can
avoid repeating same error message by wrapping that to function then you will
satisfy my perfectionist sense. That's all.

I'm not going to force you to follow it - you proved that there is a way to
hunt the line by DDEBUG so I'm satisfied.

Thanks,
Mariusz