glibc: Fix conflict error when enbale multilib on aarch64.
v2 cover glibc: Fix conflict error when enbale multilib on aarch64.
From: Lei Maohui <[email protected]>
- Fix conflict of finclude/math-vector-fortran.h error when enable multilib on aarch64.
- Because multilib header doesn't work well for fortran compiler, a compile error will occur with above glibc patch. So, disable multilib_header when fortran is enabled.
Lei Maohui (2):
multilib_header.bbclass:Disable multilib_header when fortran is
enabled.
glibc: Fix conflict error when enbale multilib on aarch64.
meta/classes-recipe/multilib_header.bbclass | 6 ++++++ meta/recipes-core/glibc/glibc-package.inc | 1 + 2 files changed, 7 insertions(+)
Signature · 1 line
2.25.1
Quoted 11 lines · leimaohui via lists.openembedded.org
From: Lei Maohui <[email protected]>
- Fix conflict of
finclude/math-vector-fortran.herror when enable multilib on
aarch64.
- Because multilib header doesn't work well for fortran compiler, a compile
error will occur with above glibc patch. So, disablemultilib_headerwhen
fortran is enabled.
Quoted 4 lines
Lei Maohui (2):
multilib_header.bbclass:Disablemultilib_headerwhen fortran is enabled.
glibc: Fix conflict error when enbale multilib on aarch64.
I'm afraid this looks like a pretty horrible workaround where the header is broken in the non-fortran enabled case and mulitlib wouldn't work in the fortran case.
I don't really want to merge a patch like this.
Cheers,
Richard
Hi, Richard
Quoted 4 lines
I'm afraid this looks like a pretty horrible workaround where the header is broken in the non-fortran enabled case and mulitlib wouldn't work in the fortran case.
I'm sorry. It seems that I should submit a discussion instead of submitting these patches directly.
I found that the current oe_multilib_header() function doesn’t work for fortran at all.
I figured out two methods to fix this error:
1. Write a new example multilib herder example for fortran, such as multilib_header_fortran_wrapper.h or something else.
2. Disable finclude/math-vector-fortran.h for fortran.
And finally, I selected the second way for the following considerations:
- I searched the bug report about multilib and fortran, it seems that few user to enable the multilib for fortran compiler.
- I built the core-image-minimal with multilib for fortran compiler. There is only one build error(finclude/math-vector-fortran.h).
For the above-mentioned considerations, I selected the second way to fix the issue quickly. I wonder if you can give me some suggestions about how to fix this issue.
Best regards
Lei
Quoted 37 lines
-----Original Message-----
From: [email protected] <[email protected]> On Behalf Of Richard Purdie
Sent: Friday, March 15, 2024 3:36 PM
To: Lei, Maohui <[email protected]>; [email protected]
Subject: Re: [OE-core] [PATCH v2 0/2] glibc: Fix conflict error when enbale multilib on aarch64.Quoted 11 lines · leimaohui via lists.openembedded.org
From: Lei Maohui <[email protected]>
- Fix conflict of
finclude/math-vector-fortran.herror when enable multilib on
aarch64.
- Because multilib header doesn't work well for fortran compiler, a compile
error will occur with above glibc patch. So, disablemultilib_headerwhen
fortran is enabled.Quoted 4 lines
Lei Maohui (2):
multilib_header.bbclass:Disablemultilib_headerwhen fortran is enabled.
glibc: Fix conflict error when enbale multilib on aarch64.I'm afraid this looks like a pretty horrible workaround where the header is broken in the non-fortran enabled case and mulitlib wouldn't work in the fortran case.
I don't really want to merge a patch like this.
Cheers,
Richard