public inbox for gcc@gcc.gnu.org
 help / color / mirror / Atom feed
From: Richard Sandiford <richard.sandiford@arm.com>
To: Keith Packard via Gcc-patches <gcc-patches@gcc.gnu.org>
Cc: gcc@gcc.gnu.org,  Keith Packard <keithp@keithp.com>
Subject: Re: [PATCH 0/3] picolibc: Add picolibc linking help
Date: Fri, 02 Sep 2022 13:36:23 +0100	[thread overview]
Message-ID: <mptwnamar7c.fsf@arm.com> (raw)
In-Reply-To: <20220824180426.820576-1-keithp@keithp.com> (Keith Packard via Gcc-patches's message of "Wed, 24 Aug 2022 11:04:23 -0700")

Keith Packard via Gcc-patches <gcc-patches@gcc.gnu.org> writes:
> Picolibc is a C library for embedded systems based on code from newlib
> and avr libc. To connect some system-dependent picolibc functions
> (like stdio) to an underlying platform, the platform may provide an OS
> library.
>
> This OS library must follow the C library in the link command line. In
> current picolibc, that is done by providing an alternate .specs file
> which can rewrite the *lib spec to insert the OS library in the right
> spot.
>
> This patch series adds the ability to specify the OS library on the
> gcc command line when GCC is configured to us picolibc as the default
> C library, and then hooks that up for arm, nds32, riscv and sh targets.

Not really my area, but the approach LGTM FWIW.  Main question/points:

- In:

  +case "${with_default_libc}" in
  +glibc)
  +    default_libc=LIBC_GLIBC
  +    ;;

  should there be a default case that raises an error for unrecognised
  libcs?  Command-line checking for configure isn't very tight, but we
  do raise similar errors for things like invalid --enable-threads values.

- I'm not sure either way about adding LIBC_NEWLIB.  On the one hand
  it makes sense for completeness, but on the other it's write-only.
  Adding it means that --with-default-libc=newlib toolchains have a
  different macro configuration from a default toolchain even in cases
  where newlib is the default.

  On balance I think it would be better to accept
  --with-default-libc=newlib but set default_libc to the empty string.

- Should we raise an error for toolchains that don't support the given
  C library?  It feels like we should, but I realise that could be
  difficult to do.

- Very minor, but in lines like:

  +#if defined(DEFAULT_LIBC) && defined(LIBC_PICOLIBC) && DEFAULT_LIBC == LIBC_PICOLIBC

  is LIBC_PICOLIB ever undefined?  It looks like config.gcc provides
  an unconditional definition.  If it is always defined then:

  #if DEFAULT_LIBC == LIBC_PICOLIB

  would be clearer.

Thanks,
Richard

>
> Keith Packard (3):
>   Allow default libc to be specified to configure
>   Add newlib and picolibc as default C library choices
>   Add '--oslib=' option when default C library is picolibc
>
>  gcc/config.gcc            | 56 ++++++++++++++++++++++++++++++++-------
>  gcc/config/arm/elf.h      |  5 ++++
>  gcc/config/nds32/elf.h    |  4 +++
>  gcc/config/picolibc.opt   | 26 ++++++++++++++++++
>  gcc/config/riscv/elf.h    |  4 +++
>  gcc/config/sh/embed-elf.h |  5 ++++
>  gcc/configure.ac          |  4 +++
>  7 files changed, 95 insertions(+), 9 deletions(-)
>  create mode 100644 gcc/config/picolibc.opt

  parent reply	other threads:[~2022-09-02 12:36 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-08-24 18:04 Keith Packard
2022-08-24 18:04 ` [PATCH 1/3] picolibc: Allow default libc to be specified to configure Keith Packard
2022-08-24 18:04   ` [PATCH 2/3] picolibc: Add newlib and picolibc as default C library choices Keith Packard
2022-08-24 18:04     ` [PATCH 3/3] picolibc: Add '--oslib=' option when default C library is picolibc Keith Packard
2022-08-24 18:24 ` [PATCH 0/3] picolibc: Add picolibc linking help Andrew Pinski
2022-09-02 12:36 ` Richard Sandiford [this message]
2022-09-03  6:02   ` Keith Packard

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=mptwnamar7c.fsf@arm.com \
    --to=richard.sandiford@arm.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=gcc@gcc.gnu.org \
    --cc=keithp@keithp.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).