Hi Masahiro,
On Thu, 7 May 2020 at 20:31, Masahiro Yamada masahiroy@kernel.org wrote:
Hi Simon,
On Fri, May 8, 2020 at 10:37 AM Simon Glass sjg@chromium.org wrote:
Hi Masahiro,
On Thu, 7 May 2020 at 07:10, Masahiro Yamada yamada.masahiro@socionext.com wrote:
<common.h> pulls in a lot of bloat. <common.h> is unneeded in most of places.
Signed-off-by: Masahiro Yamada yamada.masahiro@socionext.com
I'm wary of this. I think that every file should include common.h
I disagree.
"Please include <common.h> at the beginning of every file" is a fragile rule. You have no way to check it.
We can add it to checkpatch.
Our goal is to get rid of the special treatment of <common.h>
As we get closer though I've been thinking about the goal. Do we want people to include config.h specifically if common.h has nothing in it? I feel it is safer to keep common.h, perhaps just with config.h included, until we fully understand what we need.
and the solution is to remove the bloat. I have been plugging away at that. There is a pending series that reduces it down further, to 14 includes. Please help review!
http://patchwork.ozlabs.org/project/uboot/list/?series=169491
I saw it.
Humans cannot check it. If buildman does not raise a flag, it is fine.
OK.
The problem is that when someone uses #ifdef CONFIG options the config.h has to be included. So your patch is a bit brittle. As soon as someone uses CONFIG it may break.
For the legacy CONFIG options, yes. The options in Kconfig are all safe.
How come? If config.h is included, the options are not defined.
Common options were mostly moved to Kconfig.
We still have lots in scripts/config_whitelist.txt but most of them are platform-specific craps.
Yes...perhaps we should try to have two whitelists, so we know which ones matter more.
Regards, Simon