On 2025-01-28 7:28 a.m., Zoltan Boszormenyi via lists.openembedded.org
wrote:
2025. 01. 28. 12:57 keltezéssel, Alexander Kanavin írta:
On Tue, 28 Jan 2025 at 12:53, Böszörményi Zoltán <[email protected]>
wrote:
Which is
- written in rust
rust is in core.
- needs libclang.
There was a motion to merge meta-clang into core,
what happened to it?
The effort to move clang to core is ongoing. Patches were posted and
reviewed, with a significant number of issues to be addressed. If you
want it to happen sooner, please do help out with them.
Even once clang lands, we likely won't re-enable signing by default,
as having both rust and clang on the way to do_package
That's bad. This would make some packages in my meta-clang-revival
layer non-buildable:
* older versions of intel-graphics-compiler/intel-compute-runtime
from meta-intel need llvm/clang 14
* python3-llvmlite (dependency of python3-numba) needs
llvm/clang 15
With Yocto 5.1 and older, all 3 libLLVM versions (14, 15 and the default
version from meta-clang) can be packaged / signed and the library
packages can be installed in parallel.
But clang and e.g. clang14 can't be both pulled in as build
dependencies, since they conflict.
Zoltan,
A patch to enable the signing using a PACKAGECONFIG for rpm would
likely be accepted. Off by default of course and a test would be ideal.
../Randy
(which uses
bits of rpm) will lengthen build times to unacceptable levels. I just
posted some unfortunate benchmarks for rust-native in particular. But
you might get the rpm signing oe-selftest working again.
Alex
--
# Randy MacLeod
# Wind River Linux
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#210337):
https://lists.openembedded.org/g/openembedded-core/message/210337
Mute This Topic: https://lists.openembedded.org/mt/110856095/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
