Rustup #2673
Rustup #2673
Conversation
Implement condvars for Windows Adds 3 shims for Windows: `SleepConditionVariableSRW`, `WakeConditionVariable`, `WakeAllConditionVariable` to add support for condvars (which fixes rust-lang#2628). Salvaged from what was removed from rust-lang#2231
remove a stray stderr file
README: Add macos cache deletion command
bump rustc-build-sysroot
pthread: slight refactoring of how we access the sync object fields
Add miri_alloc, miri_dealloc `miri_alloc` and `miri_dealloc` are basically the same as `__rust_alloc` and `__rust_dealloc` respectively, but without the check for a global allocator. This should allow bootstrapping an allocator in environments, where no fundamental way of allocating memory is available (`no_std` + `alloc` in Miri).
update josh instructions josh-project/josh#965 and josh-project/josh#994 have been merged so we don't need a forked josh any more. :) However, this is blocked on josh-project/josh#1032 which currently prevents me from actually testing this...
adjust josh pushing, remove './miri toolchain HEAD/commit' Looks like pushing works much better if we use as base commit the last commit we pulled from rustc and merged into Miri. Which I guess is fair. Conveniently, the `rust-version` file should usually be pretty much exactly that commit. Also adjust `rustc-pull` to update that file. Pulling is now the way to update to a newer rustc, so I also removed `./miri toolchain HEAD`.
use josh feature to fetch a specific commit
clean up retagging recursive traversal a bit The last refactor left some strange things behind -- no need to have `qualify` be a separate function, really.
remove unused `shell_escape` dep
|
Of course pushing this back to rustc fails again due to josh-project/josh#998. I really don't have the time to debug something every single pull/push... so I am beginning to wonder if josh might just not be ready yet for our needs. :/ |
|
And even if I get it to push, re-pulling does not do a proper roundtrip... |
|
We might have to do a force-push to Miri again. :/ Basically, trying to get our current Miri master into rustc via josh, then re-fetching that and force-pushing Miri to that. |
…check in rustc-push
change handling of rust-version file in rustc-pull, and add a sanity check in rustc-push
|
I have identified the source of the non-roundtrip -- it's 218f60e. I don't think we can fix this without a force push, unfortunately. #2674 adjusts the script to
So at least next time this happens we should get a warning before things are in rustc history forever, which should let us avoid further force-pushes. |
|
This branch has identical content to #2675, so we use it as CI proxy. (The other PR has conflicts so CI is not even started.) |
This is unfortunately duplicating some Miri commits. When producing rust-lang/rust#104456, josh did something odd in rust-lang/rust@46d0df4 and added a rust-version file change. I have adjusted the way we update the rust-version file in
rustc-pulland hope this will not happen again...