Re: [OE-core][PATCH] debugedit: fix build failure when enabling DEBUG_BUILD

[Randy MacLeod <randy.macleod@windriver.com>]

[-- Attachment #1: Type: text/plain, Size: 5779 bytes --]

On 2025-01-20 2:33 a.m., Chen Qi via lists.openembedded.org wrote:
> From: Chen Qi<Qi.Chen@windriver.com>
>
> When DEBUG_BUILD is enabled, we use "-Og -g" gcc options. In such
> case, the xxhash functions are considered not inline, yet debugedit.c
> defined XXH_INLINE_ALL to force inline, thus causing build failure.

I asked people about this PR on IRC yesterday and they said they wanted 
to hear what
upstream had to say. So far there is no reply to your email. Maybe 
create a defect in:

https://sourceware.org/bugzilla/buglist.cgi?quicksearch=debugedit
to get the upstream devs attention?

This inlining gcc-14 compile error has been discussed in the context of 
xxhash.

In the Binary Size control section of the README.md for:
https://github.com/Cyan4973/xxHash

we have:

|XXH_NO_INLINE_HINTS|:
By default, xxHash uses |__attribute__((always_inline))| and 
|__forceinline| to improve performance at the cost of code size.
Defining this macro to 1 will mark all internal functions as |static|, 
allowing the compiler to decide whether to
inline a function or not. This is very useful when optimizing for 
smallest binary size, and is automatically defined
when compiling with |-O0|, |-Os|, |-Oz|, or |-fno-inline| on GCC and Clang.
*It may also be required to successfully compile  using |-Og|, depending 
on compiler version.*


This was updated in:
https://github.com/Cyan4973/xxHash/commit/574aabad87b2ab9440403e92e1075ef48554eb87

and mentioned in:
https://github.com/Cyan4973/xxHash/issues/943#issuecomment-2563205130


Hope that helps push things along when you're back from holiday and
have time to work on this Qi.


../Randy


>
> This patch fixes this issue by not assuming those are always inline and
> instead link the xxhash library.
>
> The 0003-Makefile.am-do-not-update-manual.patch is moved from musl
> specific patch to SRC_URI, because we now have a patch to modify debugedit.c
> and this will cause manual to be generated again. This is unnecessary and
> will report help2man missing.
>
> Signed-off-by: Chen Qi<Qi.Chen@windriver.com>
> ---
>   .../debugedit/debugedit_5.1.bb                |  4 +-
>   ...-not-assume-xxhash-can-always-inline.patch | 48 +++++++++++++++++++
>   2 files changed, 49 insertions(+), 3 deletions(-)
>   create mode 100644 meta/recipes-devtools/debugedit/files/0001-Do-not-assume-xxhash-can-always-inline.patch
>
> diff --git a/meta/recipes-devtools/debugedit/debugedit_5.1.bb b/meta/recipes-devtools/debugedit/debugedit_5.1.bb
> index 00c40473c4..02429afcda 100644
> --- a/meta/recipes-devtools/debugedit/debugedit_5.1.bb
> +++ b/meta/recipes-devtools/debugedit/debugedit_5.1.bb
> @@ -10,9 +10,7 @@ LIC_FILES_CHKSUM ="file://COPYING;md5=59530bdf33659b29e73d4adb9f9f6552 \ 
> file://COPYING3;md5=d32239bcb673463ab874e80d47fae504"
>   
>   SRC_URI ="https://sourceware.org/ftp/debugedit/${PV}/debugedit-${PV}.tar.xz \ - "
> -
> -SRC_URI:append:libc-musl = "\
> +file://0001-Do-not-assume-xxhash-can-always-inline.patch \
>              file://0003-Makefile.am-do-not-update-manual.patch \
>              "
>   
> diff --git a/meta/recipes-devtools/debugedit/files/0001-Do-not-assume-xxhash-can-always-inline.patch b/meta/recipes-devtools/debugedit/files/0001-Do-not-assume-xxhash-can-always-inline.patch
> new file mode 100644
> index 0000000000..8e9256ec01
> --- /dev/null
> +++ b/meta/recipes-devtools/debugedit/files/0001-Do-not-assume-xxhash-can-always-inline.patch
> @@ -0,0 +1,48 @@
> +From 76eb6ce69349a05d0ff069062182550ebd88f146 Mon Sep 17 00:00:00 2001
> +From: Chen Qi<Qi.Chen@windriver.com>
> +Date: Sun, 19 Jan 2025 22:56:36 -0800
> +Subject: [PATCH] Do not assume xxhash can always inline
> +
> +When building with "-Og -g", we'll get errors like below:
> +
> +  include/xxhash.h:5210:1: error: inlining failed in call to
> +  'always_inline' 'XXH3_scrambleAcc_sse2': function not considered for inlining
> +
> +Upstream-Status: Submitted [https://sourceware.org/pipermail/debugedit/2025-January/000359.html]
> +
> +Signed-off-by: Chen Qi<Qi.Chen@windriver.com>
> +---
> + Makefile.am       | 4 ++--
> + tools/debugedit.c | 1 -
> + 2 files changed, 2 insertions(+), 3 deletions(-)
> +
> +diff --git a/Makefile.am b/Makefile.am
> +index c590edf..35fd947 100644
> +--- a/Makefile.am
> ++++ b/Makefile.am
> +@@ -42,8 +42,8 @@ find-debuginfo: $(top_srcdir)/scripts/find-debuginfo.in Makefile
> +
> + debugedit_SOURCES = tools/debugedit.c \
> + 		    tools/hashtab.c
> +-debugedit_CFLAGS = @LIBELF_CFLAGS@ @LIBDW_CFLAGS@ $(AM_CFLAGS)
> +-debugedit_LDADD = @LIBELF_LIBS@ @LIBDW_LIBS@
> ++debugedit_CFLAGS = @LIBELF_CFLAGS@ @LIBDW_CFLAGS@ @XXHASH_CFLAGS@ $(AM_CFLAGS)
> ++debugedit_LDADD = @LIBELF_LIBS@ @LIBDW_LIBS@ @XXHASH_LIBS@
> +
> + sepdebugcrcfix_SOURCES = tools/sepdebugcrcfix.c
> + sepdebugcrcfix_CFLAGS = @LIBELF_CFLAGS@ $(AM_CFLAGS)
> +diff --git a/tools/debugedit.c b/tools/debugedit.c
> +index beefd65..43f9cee 100644
> +--- a/tools/debugedit.c
> ++++ b/tools/debugedit.c
> +@@ -82,7 +82,6 @@ typedef struct Ebl_Strtab	Strtab;
> +
> + #include "tools/hashtab.h"
> +
> +-#define XXH_INLINE_ALL
> + #include "xxhash.h"
> +
> + #define DW_TAG_partial_unit 0x3c
> +--
> +2.25.1
> +
>
> -=-=-=-=-=-=-=-=-=-=-=-
> Links: You receive all messages sent to this group.
> View/Reply Online (#210023):https://lists.openembedded.org/g/openembedded-core/message/210023
> Mute This Topic:https://lists.openembedded.org/mt/110711482/3616765
> Group Owner:openembedded-core+owner@lists.openembedded.org
> Unsubscribe:https://lists.openembedded.org/g/openembedded-core/unsub [randy.macleod@windriver.com]
> -=-=-=-=-=-=-=-=-=-=-=-
>

-- 
# Randy MacLeod
# Wind River Linux

[-- Attachment #2: Type: text/html, Size: 8644 bytes --]