Fix shards regression #487
Conversation
Octogonapus
force-pushed
the
fix_shards_regression
branch
from
fe3d85a to
46f9c46
Compare
| "x86_64-w64-mingw32"=>"x86_64-w64-windows-gnu", | ||
| ) | ||
|
|
||
| triple2target(triple::String) = get(JLL_ENV_CLANG_TARGETS_MAPPING, triple, "unknown") | ||
|
|
||
| function __triplet(p::Platform) | ||
| for k in keys(p.tags) | ||
| if k != "arch" && k != "os" && k != "libc" | ||
| if k != "arch" && k != "os" && k != "libc" && k != "call_abi" && k != "os_version" |
There was a problem hiding this comment.
call_abi is needed for armv7l, os_version is needed for freebsd and sometimes darwin
| t *= "14" | ||
| elseif os(p) == "macos" && arch(p) == "aarch64" | ||
| t *= "20" | ||
| if os_version(p) === nothing |
There was a problem hiding this comment.
I don't know why the user wouldn't specify the version, but here I am just maintaining backward compatibility if they didn't.
|
looks like there are download issues on windows platforms. |
| @testset "get_default_args for all triples" begin | ||
| # just testing this does not crash on any default triple | ||
| Clang.get_default_args.(Clang.JLLEnvs.JLL_ENV_TRIPLES) | ||
| end |
There was a problem hiding this comment.
this test downloads artifacts for all platforms. running it every time is a bit overkill.
is there a way to somehow cache it?
There was a problem hiding this comment.
or maybe we can control it by env variables.
There was a problem hiding this comment.
sorry i didn't realize it did that, i've simplified the test
I'm not sure what that error means, do you know? |
no. probably just a mysterious connection issue. |
|
The new test looks good. Thanks! |
|
just noticed we just hit 1000 commits ;) |
This PR is my attempt at fixing the shards regression in #486. I added a test that checks all the default triples.
Fixes #486.