Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 30 Apr 2020 09:56:00 -0500
From:      Kyle Evans <kevans@freebsd.org>
To:        Dewayne Geraghty <dewayne@heuristicsystems.com.au>
Cc:        "freebsd-ports@freebsd.org" <freebsd-ports@freebsd.org>
Subject:   Re: Ports failing with -fno-common with clang 9/gcc 9
Message-ID:  <CACNAnaHAo6jFpRu3j0rB19oK7WrSvafRqiMOvy7JdWa7uvB2wg@mail.gmail.com>
In-Reply-To: <fd56ab9b-82f2-3022-2773-47767afa39c3@heuristicsystems.com.au>
References:  <fd56ab9b-82f2-3022-2773-47767afa39c3@heuristicsystems.com.au>

next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Apr 30, 2020 at 2:20 AM Dewayne Geraghty
<dewayne@heuristicsystems.com.au> wrote:
>
> As -fno-common will become the default in gcc10/llvm11 per
> https://lists.freebsd.org/pipermail/svn-src-stable-12/2020-April/004761.html
>

First- thanks! It's nice to see confirmation that the messages being
written here aren't just being broadcasted to the void, doubly so when
said confirmation is in the form of e-mails like this. :-)

> I thought I might share the list of ports that failed to build for
> maintainers to be aware of using -fno-common:
>
> I'm sorry this is a small list but I only use 1488 ports.

Indeed, it looks like a small list, but you've just identified 42
problems to work on while only using ~5% (? based on 30,000 ports,
which I know is probably low) of the ports tree.

That is to say, I think this speaks to the magnitude of the issue
we're dealing with here. Many will escape unscathed, but it's a good
idea for everyone to start taking action *now* to identify and fix or
report these so that we're not overloading our relatively small group
of toolchain folks in the next ~6 months. My experience in base
revealed that most of the third party software we had there was
already aware of the coming change and had patches to accommodate, so:

- For projects with a faster release cadence, this may be easily
overcome by events
- For projects with a slower release cadence, you may find patches
that can be or have been backported

In any event, I would urge folks to be proactive and identify this
stuff, reporting issues upstream and spreading awareness of the
impending default change for those projects that may not already be
actively aware.

On a closing note, I'm just going to kinda drop these patches here for
anyone that's willing/able to help make this a collective effort to
identify/fix/report problems here; they backport the default
-fno-common patch from future-LLVM11 to the base system compiler on a
system near you:

HEAD: https://people.freebsd.org/~kevans/llvm-fnocommon-head.diff
stable/12: https://people.freebsd.org/~kevans/llvm-fnocommon-12.diff

**Note that FreeBSD branches earlier than these (in the past couple of
weeks!) will not fully build with this patch applied:**

- r360072 (head)
- r360496 (stable/12)
- r360494 (stable/11)

Userland builds (all that's needed for most poudriere jails) should be
fine in the past couple of weeks earlier than those commits, but no
active releases will build with these patches applied.

Folks can feel free to tag me in -fno-common related PRs if assistance
is needed for any reason.

Thanks,

Kyle Evans



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CACNAnaHAo6jFpRu3j0rB19oK7WrSvafRqiMOvy7JdWa7uvB2wg>