Re: [OE-core] [PATCH] glibc: revert one upstream change to work around broken DEBUG_BUILD build

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Tue, 2022-07-26 at 14:18 +0200, Martin Jansa wrote:
> * work around for build failure with DEBUG_BUILD:
>   http://errors.yoctoproject.org/Errors/Details/661468/
> 
>   TOPDIR/tmp-glibc/work/core2-64-oe-linux/glibc/2.35-r0/recipe-sysroot-native/usr/bin/x86_64-oe-linux/../../libexec/x86_64-oe-linux/gcc/x86_64-oe-linux/11.3.0/ld.bfd:
>   TOPDIR/tmp-glibc/work/core2-64-oe-linux/glibc/2.35-r0/build-x86_64-oe-linux/libc.a(libc-tls.o):
>   in function `__libc_setup_tls':
>   /usr/src/debug/glibc/2.35-r0/git/csu/libc-tls.c:202: undefined reference to `_startup_fatal_not_constant'
> 
>   introduced in:
>   https://git.openembedded.org/openembedded-core/commit/?id=50ca0eea5424c89c9b39cf69d8bd73fd65f06e95
>   with:
>   https://sourceware.org/git/?p=glibc.git;a=commitdiff;h=a2a6bce7d7e52c1c34369a7da62c501cc350bc31
> 
>   Temporary reverting as discussed in:
>   https://lists.openembedded.org/g/openembedded-core/message/168463
>   (it's also reverted in buildroot now with: https://lore.kernel.org/buildroot/20220720213853.1A27184110@busybox.osuosl.org/t/)
> 
> Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
> ---
>  ...lement-a-useful-version-of-_startup_.patch | 128 ++++++++++++++++++
>  meta/recipes-core/glibc/glibc_2.35.bb         |   2 +
>  2 files changed, 130 insertions(+)
>  create mode 100644 meta/recipes-core/glibc/glibc/0001-Revert-Linux-Implement-a-useful-version-of-_startup_.patch
> 
> diff --git a/meta/recipes-core/glibc/glibc/0001-Revert-Linux-Implement-a-useful-version-of-_startup_.patch b/meta/recipes-core/glibc/glibc/0001-Revert-Linux-Implement-a-useful-version-of-_startup_.patch
> new file mode 100644
> index 0000000000..2421a63605
> --- /dev/null
> +++ b/meta/recipes-core/glibc/glibc/0001-Revert-Linux-Implement-a-useful-version-of-_startup_.patch
> @@ -0,0 +1,128 @@
> +From 6b8959add09e425df262bf9178b39ca35bc4003c Mon Sep 17 00:00:00 2001
> +From: Martin Jansa <Martin.Jansa@gmail.com>
> +Date: Sun, 24 Jul 2022 19:41:41 +0200
> +Subject: [PATCH] Revert "Linux: Implement a useful version of _startup_fatal"
> +
> +This reverts commit 2d05ba7f8ef979947e910a37ae8115a816eb4d08.
> +Upstream-Status: Inappropriate [temporary work around]
> +
> +Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>

Khem and I talked briefly about this. The upstream glibc community
doesn't support building glibc with -O1, they require it always be -O2
or higher. I don't really want to be stuck carrying a glibc patch for
something like this we can't resolve.

Is there an underlying issue that should be fixed here or is this
really the case that this is something upstream simply don't expect to
work?

Cheers,

Richard