bootstrap: Disable initial-exec TLS model on powerpc #85807
Conversation
|
(rust-highfive has picked a reviewer for you, use r? to override) |
rust-highfive
added
the
S-waiting-on-review
|
Please note: I have not fully tested this change yet. So please don't merge it yet although I'm sure it's correct. |
src/bootstrap/builder.rs
Outdated
| @@ -1237,6 +1237,7 @@ impl<'a> Builder<'a> { | |||
| // efficient initial-exec TLS model. This doesn't work with `dlopen`, | |||
| // so we can't use it by default in general, but we can use it for tools | |||
| // and our own internal libraries. | |||
| #[cfg(not(target_arch = "powerpc"))] | |||
There was a problem hiding this comment.
I think cfg! macro would be nicer here.
There was a problem hiding this comment.
Hm, I'm surprised to see cfg here at all; I'd expect this to check target.contains("powerpc") or so...
There was a problem hiding this comment.
I believe cfg would be wrong when cross-compiling - since "target" here refers to the target of bootstrap which would be the host (native) platform?
There was a problem hiding this comment.
@glaubitz try && !target.triple.startswith("powerpc-") instead
There was a problem hiding this comment.
Done. I'm testing the change now on a real PowerPC machine to see if it actually works.
There was a problem hiding this comment.
The change as suggested by @infinity0 works fine and fixes the problem on 32-bit PowerPC.
But please feel free to use a different approach if anyone has a better idea. I don't insist on my particular approach.
|
If this TLS mode is completely unsupported on powerpc, then I think we should emit an error whenver someone tries to specify it on the command line. Otherwise, anyone else using |
Mark-Simulacrum
added
S-waiting-on-author
glaubitz
force-pushed
the
powerpc-disable-initial-exec-tls
branch
from
c4b54f3
to
674e27e
Compare
This comment has been minimized.
This comment has been minimized.
glaubitz
force-pushed
the
powerpc-disable-initial-exec-tls
branch
from
674e27e
to
283619c
Compare
Is there some sort of feature matrix that Rust is carrying internally for every architecture where this information could be stored as a flag? I would guess this bug could affect the upcoming m68k port as well. Plus, such a feature matrix could be used to determine whether a certain test is run or not. Currently, all tests seem to be checking per architecture rather than per feature. |
This mode is an optimization hint, if an "initial-exec" hint is useless for a given target (like e.g. TLS models not generally mapping on Windows-based target), then it should be demoted to a weaker hint. |
joelpalmer
added
S-waiting-on-author
|
@glaubitz Ping from triage, what's next steps here? |
|
@glaubitz Are you intending to add anything more to this PR, or shall I tag this for review? |
|
@bkaestner I don't have any strong preference, the fix works for me. Feel free to merge it as-is or replace it with a better solution. |
@bstrie I believe this message was meant for you. |
|
Oops, sorry. Two people starting with "@b" in the thread and I just had only one coffee this morning :). |
I don't want to add anything else. Could you tag this for review? Thanks! |
bstrie
added
S-waiting-on-review
|
📌 Commit 283619c has been approved by |
bors
added
S-waiting-on-bors
Rollup of 7 pull requests Successful merges: - rust-lang#85807 (bootstrap: Disable initial-exec TLS model on powerpc) - rust-lang#87761 (Fix overflow in rustc happening if the `err_count()` is reduced in a stage.) - rust-lang#87775 (Add hint for unresolved associated trait items if the trait has a single item) - rust-lang#87779 (Remove special case for statement `NodeId` assignment) - rust-lang#87787 (Use `C-unwind` ABI for `__rust_start_panic` in `panic_abort`) - rust-lang#87809 (Fix typo in the ptr documentation) - rust-lang#87816 (Sync rustc_codegen_cranelift) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup
Fixes #81334.