[U-Boot] [PATCH 0/6] stdio: Clean up

Marek Vasut marex at denx.de
Sat Sep 1 16:19:06 CEST 2012


Dear Wolfgang Denk,

> Dear Marek Vasut,
> 
> In message <1346453055-30888-1-git-send-email-marex at denx.de> you wrote:
> > Clean up a few things in STDIO. Mostly, don't export structures directly,
> > but introduce some kind of accessors if needed and remove dead code.
> > 
> > NOTE: I'm still working on the "compile tested on 2 different
> > architectures"
> > 
> >       part. I'll keep you posted about that. I'd be glad for a review
> >       though. I'd hate to pull in logic errors, especially into such
> >       critical code.
> > 
> > NOTE2: Cross-posting cover to DM list, so I get feedback from those guys.
> > 
> > Marek Vasut (6):
> >   stdio: dm: Murder dead code in console.c
> >   stdio: dm: Add accessors to stdio_devices[]
> >   stdio: dm: Make stdio_devices[] local
> >   stdio: dm: Add stdio_fd_to_name() call
> >   stdio: dm: Use stdio_fd_to_name() call to localize stdio_names
> >   stdio: dm: Optimize stdio_print_current_devices()
> 
> I can't make heads nor tails from this patch series.
> 
> 1) It was posted to the U-Boot list, but all patches carry a "dm:" in
>    the subject, which does not appear to make sense to me, as at
>    least some of the changes have no relation to DM work at all.

They very distantly are. I really needed to clean up the STDIO a bit to 
familiarize myself with the code I'm soon going to break.

But all in all, I think exporting structures for others to access them as they 
wish isn't the best of ideas. Therefore I encapsulated these into the file and 
added accessors. The direction these patches take with STDIO and console.c stuff 
in U-Boot is such that applying proper encapsulation will allow easier 
conversion to the driver model stuff later. Yet I'm getting there with really 
small steps as I need to be very careful here.

> 2) It appears this might be a RFC series, so why isn't it maked as
>    such in the Subject: ?

It's not RFC, why would it be RFC? I'm still working on the "NOTE" part though.

> 3) It appears that some code gets added - what is the impact of these
>    changes on the memory footprint?

So far I tested this on M28:

Before:
   text    data     bss     dec     hex filename
 415705    7688  288708  712101   adda5 ./u-boot
  11754     788      12   12554    310a ./spl/u-boot-spl

After:
   text    data     bss     dec     hex filename
 415590    7676  288700  711966   add1e ./u-boot
  11794     788      12   12594    3132 ./spl/u-boot-spl

As you can see, the SPL grows a bit, yet U-Boot shrunk. Tested with Debian GCC 
4.7.1 .

> 4) Besides the dead code removal - what exactly is the purpose of
>    these patches?

Mostly see 1).

> Best regards,
> 
> Wolfgang Denk

Best regards,
Marek Vasut


More information about the U-Boot mailing list