Statically link libstdc++ on windows-gnu #65911
Conversation
|
r? @kennytm (rust_highfive has picked a reviewer for you, use r? to override) |
rust-highfive
added
the
S-waiting-on-review
|
Cc @Amanieu who found the cause and could be interested in this PR. |
| @@ -279,7 +279,11 @@ fn main() { | |||
| let path = PathBuf::from(s); | |||
| println!("cargo:rustc-link-search=native={}", | |||
| path.parent().unwrap().display()); | |||
| println!("cargo:rustc-link-lib=static={}", stdcppname); | |||
| if target.contains("windows") { | |||
| println!("cargo:rustc-link-lib=static-nobundle={}", stdcppname); | |||
There was a problem hiding this comment.
This may cause distribution issues since we ship rustc_llvm and users like clippy/plugins will link to rustc_llvm transitively through rustc_driver, was there a reason though static didn't work?
There was a problem hiding this comment.
Yes, I think quite nasty or I'm just misunderstanding.
First let's take a look at static libstdc++:
nm /mingw64/lib/gcc/x86_64-w64-mingw32/9.2.0/libstdc++.a | grep pthread_mutex_init
U __imp_pthread_mutex_init
U pthread_mutex_init
U pthread_mutex_init
[..]
There are references to both static and dynamic symbols!
Now using static failed undefined with 2 undefined symbols: __imp_pthread_mutex_{destroy,init}. I honestly have no idea why picked dynamic symbols, though I haven't chased that rabbit.
It could be mitigated by dynamically linking to winpthread library (by removing static in line 298), windows-gnu ships libwinpthread-1.dll anyway because. But I felt it's just replacing one problem with another.
With some guidance I could try to investigate why picks dynamic pthread symbols when using static but static-nobundle seems to fit here (I asked for try build in OP to be sure).
There was a problem hiding this comment.
Hm ok, I'm not really sure what's going on with static vs static-nobundle, but now that I think about this we're linking this into librustc_driver.so where the "nobundle" part won't keep propagating, so this doesn't actually affect consumers, and as a result should be fine to land.
|
Want to revert the CI configuration back to the default? Otherwise r=me? |
|
I thought you might want to give it |
mati865
force-pushed
the
static-libstdcxx-mingw
branch
from
c814d88
to
473a498
Compare
|
Removed CI changing commit. |
|
Oh sorry I misread the ci changes, can this run the mingw test on this pr to ensure they pass before we send to bors? |
mati865
force-pushed
the
static-libstdcxx-mingw
branch
5 times, most recently
from
c8b551a
to
b20d8d3
Compare
|
All mingw tests are green: https://dev.azure.com/rust-lang/rust/_build/results?buildId=12392&view=logs&jobId=5ac1211a-cbe7-52be-dcc2-57e3991fd20c |
|
👍 Sounds good to me! @bors: r+ |
|
📌 Commit b20d8d3 has been approved by |
bors
added
S-waiting-on-bors
…lexcrichton Statically link libstdc++ on windows-gnu Fixes rust-lang#61561 by not shipping `libstdc++-6.dll` which can conflict with the GCC.
…lexcrichton Statically link libstdc++ on windows-gnu Fixes rust-lang#61561 by not shipping `libstdc++-6.dll` which can conflict with the GCC.
Statically link libstdc++ on windows-gnu Fixes #61561 by not shipping `libstdc++-6.dll` which can conflict with the GCC.
|
☀️ Test successful - checks-azure |
Fixes #61561 by not shipping
libstdc++-6.dllwhich can conflict with the GCC.