[PATCH v2 16/50] image: Add Kconfig options for FIT in the host build
sjg at chromium.org
Wed May 12 19:30:12 CEST 2021
On Wed, 12 May 2021 at 10:18, Alex G. <mr.nuke.me at gmail.com> wrote:
> On 5/12/21 10:54 AM, Simon Glass wrote:
> > Hi Alex,
> > On Wed, 12 May 2021 at 09:48, Alex G. <mr.nuke.me at gmail.com> wrote:
> >> On 5/12/21 9:51 AM, Simon Glass wrote:
> >>> Hi Alex,
> >>> On Tue, 11 May 2021 at 13:57, Alex G. <mr.nuke.me at gmail.com> wrote:
> >>>> On 5/6/21 9:24 AM, Simon Glass wrote:
> >> [snip]
> >>>>> +
> >>>>> +config HOST_FIT_PRINT
> >>>>> + def_bool y
> >>>>> + help
> >>>>> + Print the content of the FIT verbosely in the host build
> >>>> This option also doesn't make sense.This seems to do what 'mkimage -l'
> >>>> already supports.
> >>> Are you sure you have looked at the goal of the CONFIG_IS_ENABLED()
> >>> changes? This is here purely to avoid #ifdefs in the share code.
> >> On the one hand, we have the cosmetic inconvenience caused by #ifdefs.
> >> On the other hand we have the config system. To most users, the config
> >> system is likely more visible, while it's mostly developers who will
> >> ever see the ifdefs.
> >> Therefore, in order to get the developer convenience of less ifdefs, we
> >> have to sacrifice user convenience by cloberring the Kconfig options. I
> >> think this is back-to-front.
> > These Kconfig options are not visible to users. They cannot be updated
> > in defconfig, nor in 'make menuconfig', etc. They are purely there for
> > the build system.
> >> Can we reduce the host config count to just SLL/NOSSL?
> > The point here is that the code has a special case for host builds,
> > and this is a means to remove that special case and make the code
> > easier to maintain and follow.
> I understand where you're coming from. Without these changes, the code
> knows what it should and should not do, correct? My argument is that if
> the code has the logic to do the correct thing, that logic should not be
> split with the config system.
> I agree with the goal of reducing clutter in the source code. I disagree
> with this specific course of fixing it. Instead, I propose a single
> kconfig for host tools for SSL on/off.
> The disadvantage of my proposal is that we have to refactor the common
> code in a way consistent with the goals, instead of just changing some
> #ifdefs to if(CONFIG_IS_ENABLED()). I admit that, off the top of my
> head, I don't have a detailed plan on how to achieve this.
You are mostly describing the status quo, so far as I understand it.
The problem is with the code that is built for both boards and tools.
For boards, we want this code to be compiled conditionally, depending
on what options are enabled. For tools, we want the code to be
I can think of only three ways to do this:
- status quo (add #ifdefs USE_HOSTCC wherever we need to)
- my series (make use of hidden Kconfig options to avoid that)
- put every single feature and associated lines of code in separate
files and compile them conditionally for boards, but always for tools
I believe the last option is actually impossible, or at least
impractical. It would cause an explosion of source files to deal with
all the various combinations, and would be quite painful to maintain
More information about the U-Boot