Re: [Buildroot] [PATCH] arch: Set max/common-page-size for libgcc & libstdc++

[Thomas Petazzoni via buildroot <buildroot@buildroot.org>]

Hello Alexey,

Thanks for the patch. See below some comments.

On Wed,  3 Jan 2024 02:23:11 -0800
Alexey Brodkin via buildroot <buildroot@buildroot.org> wrote:

> I would agree that it gets more and more ugly, but so far I cannot see
> any more elegant solution. So if there're suggestions how to improve,
> let's discuss.

Indeed, it's not super great as you realized yourself. Also, I believe
it's not the first time we identify a discrepancy in CFLAGS that are in
the wrapper, but that we forget to pass when building gcc/libc. See
your own commit d2ae7eb2a24c9033f8d8a9f81c5fe2ff2febdb5f from 2020,
which also added -matomic in gcc.mk.

Things are also a bit messy in the different C libraries:

- uClibc is built with $(TARGET_ABI) and $(TARGET_DEBUGGING)

- glibc is built with $(TARGET_OPTIMIZATION) and a bunch of additional
  flags

- musl is built with the full $(TARGET_CFLAGS)

Seems a bit meh to me.

So I think ideally we would to distinguish in a clear manner:

- CFLAGS that are passed through the wrapper, and which therefore need
  to be manually passed when building gcc and C libraries

- CFLAGS that are passed through the environment

I think at least for now what could be done is to use
$(ARCH_TOOLCHAIN_WRAPPER_OPTS) in gcc.mk.

But I'm surprised they are not needed also in the C library build. Why
are you not seeing the same problem with the C library build, it should
also need the same page size flags, doesn't it?

Thomas
-- 
Thomas Petazzoni, co-owner and CEO, Bootlin
Embedded Linux and Kernel engineering and training
https://bootlin.com
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot