Start patching for support linux arm64 as a host OS #5377
Conversation
kohnish
commented
Nov 13, 2024
kohnish
commented
- Fix isl pointed to deprecated url
- Add precompiled dependencies for linux aarch64 (still incomplete probably)
- Add Linux ARM 64 enum (still incomplete)
- Add aarch64 linux clang in konan.property (untested)
4u7
left a comment
4u7
left a comment
There was a problem hiding this comment.
It seems verification-metadata.xml is updated manually. We have a section in the readme about dependency-verification.
https://github.com/JetBrains/kotlin?tab=readme-ov-file#dependency-verification
Thanks. Somehow nodejs is gone if I do it automatically, I'll investigate. In any case, I'm ok if this PR gets merged. I still haven't managed to compile on aarch64 linux. |
|
Hope we poor linux/Aarch64 owners will get some working prebuilt KN compiler binary. I want to use this on my android device so bad, please!
|
SvyatoslavScherbina
left a comment
SvyatoslavScherbina
left a comment
There was a problem hiding this comment.
What is the status of this PR? Is it ready for review?
What is the scope of this PR?
Changes in toolchain_builder seem unrelated to supporting Linux/arm64 as a host OS.
| #!/bin/bash | ||
|
|
||
| set -eou pipefail | ||
| set -eo pipefail |
There was a problem hiding this comment.
What is the reason for this change?
There was a problem hiding this comment.
TOOLCHAIN_VERSION_SUFFIX is optional as you see and the script was failing
There was a problem hiding this comment.
This can and should be handled the same way as in run_container.sh:
TOOLCHAIN_VERSION_SUFFIX="${3:-""}"So, removing -u shouldn't even be required there.
it's ready to be merged as I mentioned. It's meant to "start" fixing something that is broken. Any change here is reaction to the build failure by following the documentation. |
There was a problem hiding this comment.
Changes in toolchain_builder are not necessary to support linux arm64 as a host OS: the built toolchain already exists, it is used for linux arm64 target support.
Change in HostManager doesn't seem to improve anything without actually supporting linux arm64 as a host. It would make sense as a part of a larger PR that actually does unblock anything.
An example of a meaningful contribution to linux arm64 host support: build the actual missing dependencies of the compiler (like LLVM and libffi) and check whether it works with them.
| #!/bin/bash | ||
|
|
||
| set -eou pipefail | ||
| set -eo pipefail |
There was a problem hiding this comment.
This can and should be handled the same way as in run_container.sh:
TOOLCHAIN_VERSION_SUFFIX="${3:-""}"So, removing -u shouldn't even be required there.
kohnish
force-pushed
the
linux-arm64
branch
2 times, most recently
from
8852060 to
e327baa
Compare
* Fix isl pointed to deprecated url * Add precompiled dependencies for linux aarch64 * Add Linux ARM 64 enum (still incomplete)
4 participants
