More refactoring to obey platform abstraction lint #36948
Conversation
| rtabort!(concat!("assertion failed: ", stringify!($e))) | ||
| } | ||
| }) | ||
| } |
There was a problem hiding this comment.
This macro is no longer used anywhere.
|
A lot of this is just general cleanup... |
How dare you! |
|
|
||
| // Turn warnings into errors, but only after stage0, where it can be useful for | ||
| // code to emit warnings during language transitions | ||
| #![cfg_attr(not(stage0), deny(warnings))] |
There was a problem hiding this comment.
If you're really ambitious, I've been meaning to remove all the cfg_attr here and just leave deny(warnings)
| @@ -425,46 +403,62 @@ pub use core::u16; | |||
| pub use core::u32; | |||
| #[stable(feature = "rust1", since = "1.0.0")] | |||
| pub use core::u64; | |||
| #[stable(feature = "rust1", since = "1.0.0")] | |||
| pub use alloc::boxed; | |||
There was a problem hiding this comment.
I think the ordering here affects the ordering in rustdoc, right? (just confirming you want to change that too)
There was a problem hiding this comment.
It does not appear so to me. rustdoc seems to be alphabetizing everything.
| @@ -184,6 +184,10 @@ macro_rules! assert_approx_eq { | |||
| }) | |||
| } | |||
|
|
|||
| macro_rules! rtabort { | |||
There was a problem hiding this comment.
aww I was hoping we could keep this contained within the sys module
There was a problem hiding this comment.
I'll see if I can put it there. I don't recall why I moved it, presumably just because it didn't work with my prefered module declaration order in lib.rs.
| #[doc(hidden)] | ||
| #[cfg(not(target_thread_local))] |
There was a problem hiding this comment.
I think we'll want to always compile in this module. The theory behind this is that we compile the standard library with OS-based TLS (e.g. this module) but then consumers of the standard library can use the "fast TLS". The primary motivation here is OSX 10.6 and 10.7+, where 10.6 has slow TLS only but 10.7 has fast. We could in theory compile a 10.6 standard library but everyone else would have fast TLS.
I kinda liked having everything contained in one nice "sys" directory, but I guess not enough to go against conventions of filesystem hierarchies and module structures. |
|
Updated to leave os-specific TLS key in with documentation for why, move rtabort back to sys_common. |
|
@bors: r+ |
|
📌 Commit ae10558 has been approved by |
|
⌛ Testing commit ae10558 with merge d3b75bb... |
|
💥 Test timed out |
|
@bors: retry
|
|
☔ The latest upstream changes (presumably #37002) made this pull request unmergeable. Please resolve the merge conflicts. |
|
@bors: retry |
|
🔒 Merge conflict |
|
ping @brson (for an update) |
Make the directory structure reflect the module structure. I've always found the existing structure confusing.
This makes it dissimilar to how core is structured on disk, but more predictable on its own.
|
@bors r+ |
|
📌 Commit 6135cbc has been approved by |
More refactoring to obey platform abstraction lint
The most interesting things here are moving `std/sys/common` to `std/sys_common`, and `std/num/{f32,f64}.rs` to `std/{f32,f64}.rs`, and adding more documentation to `std/lib.rs`.
r? @alexcrichton
|
💔 Test failed - auto-win-gnu-64-opt |
|
@bors: retry On Wednesday, November 2, 2016, bors notifications@github.com wrote:
|
More refactoring to obey platform abstraction lint
The most interesting things here are moving `std/sys/common` to `std/sys_common`, and `std/num/{f32,f64}.rs` to `std/{f32,f64}.rs`, and adding more documentation to `std/lib.rs`.
r? @alexcrichton
The most interesting things here are moving
std/sys/commontostd/sys_common, andstd/num/{f32,f64}.rstostd/{f32,f64}.rs, and adding more documentation tostd/lib.rs.r? @alexcrichton