Timofey Peshin
Timofey Peshin
Are there any workarounds for the DEVELOPER_DIR issue on OS-X? I tried setting `BAZEL_USER_CPP_ONLY_TOOLCHAIN=1`, made sure `DEVELOPER_DIR` wasn't empty (but it was empty for me before... what is it even?),...
> Looks great! One comment is that we should then mark the run as failed somehow (assign a special exit code?), otherwise the FE will show a successfully finished run....
PR for printing to the console: https://github.com/wandb/wandb/pull/7517
> should we leave a TODO in conftest to remove it once we stop testing non-core? maybe link to a ticket? not sure what's the best way... alternatively leave it...
Note, using `install_name_tool -change` to manually rewrite the `@rpath/` entries to `/usr/lib/swift/` makes `delocate` work, which makes me think that `delocate` ignores certain paths, since `/usr/lib/swift/` does not contain any...
Thank you! Any clue as to why these libraries might be output with "@rpath/" rather than "/usr/lib" by a compiler?
I see, thanks for looking into it, I really appreciate it! It's clear to me now this isn't a bug or unexpected behavior with Delocate, so I'll close the issue.
I know a little more now. On my arm64 (aka aarch64) macOS device: * when I invoke `swiftc` with `--arch arm64`, the output binary uses `/usr/lib/swift/...` paths for Swift standard...
[otool_l_aarch64.txt](https://github.com/matthew-brett/delocate/files/14795676/otool_l_aarch64.txt) [otool_l_x86_64.txt](https://github.com/matthew-brett/delocate/files/14795677/otool_l_x86_64.txt)
A workaround for now is to use the `--ignore-missing-dependencies` option.