rustup to rustc 1.15.0-dev (ace092f56 2016-12-13) #93

oli-obk · 2016-12-14T16:07:34Z

Due to always_encode_mir we can now run tests that require auxiliary builds.

oli-obk · 2016-12-14T16:11:13Z

tests/compiletest.rs


+fn miri_pass(path: &str, target: &str) {
+    let mut config = compiletest::default_config();
+    config.mode = "mir-opt".parse().expect("Invalid mode");


mir-opt is not ideal

it runs the optimization passes (but I think we run miri before those are run)

it generates a final binary that is then executed on the host.

The clean solution would be to add a "compile-pass" test to compiletest

But it saves us this giant mess below, which would just get bigger if we started to play with auxiliary builds.

solson · 2016-12-15T05:51:26Z

src/memory.rs

+        let offset = ptr.offset as usize;
+        match alloc.bytes[offset..].iter().position(|&c| c == 0) {
+            Some(size) => {
+                if self.relocations(ptr, size as u64)?.count() != 0 {


I think this and the check_defined need to use size + 1 to ensure the found null byte also is defined and not part of a relocation.

solson · 2016-12-15T05:53:30Z

src/terminator/mod.rs

+
+            "getenv" => {
+                {
+                    let name = args[0].read_ptr(&self.memory)?;


s/name/name_ptr/ for clarity.

solson · 2016-12-15T05:57:42Z

src/terminator/mod.rs

+            }
+
+            // unix panic code inside libstd will read the return value of this function
+            "pthread_rwlock_rdlock" => {


This is fine for our current experimentation, but I guess we'll need a plan for stuff like this eventually. For the real CTFE that gets put into rustc, it would probably make more sense to have cfg(const) sections that avoid needing the libc calls.

@eddyb Have you thought about this? Panicking, printing when panicking, and even Vec allocation all currently require us to hook libc calls, which might be considered too dirty...

real ctfe can only evaluate const fn.

Also we can't do any cfg(const) if we don't want to end up compiling libstd twice (once for the target and once for const eval).

We want to make panic allowed in real CTFE in some form, though, don't we? Otherwise we can't call anything that has any code path leading to panic.

It doesn't need to be literally cfg(), but something to tell CTFE what to do instead of going into low-level code.

Allowing panicking in const fn implies that whatever functions panic!() calls (transitively) all need to be const fn unless we cut it off somewhere with a const fn-specific feature to tell CTFE to do something differently from what the native code does.

In fact, hooking libc calls is effectively just one version of "tell CTFE to do something differently", but we could perhaps do it in a more first-class, maintainable way if it was explicit in the code implementing panic.

But I'm still pretty unsure.

The easiest thing would be to create an unstable attribute that libstd can use to annotate functions that CTFE should not go into this function, but instead do some magic.

solson · 2016-12-15T09:12:26Z

Oh, it seems like these changes aren't in an actual nightly release yet, just in master?

solson · 2016-12-15T09:13:27Z

It looks like the latest nightly is 9 days old: https://static.rust-lang.org/dist/channel-rust-nightly-date.txt (2016-12-06 at time of writing)

solson · 2016-12-15T09:17:54Z

I will revert this to keep master building for anyone new who comes along, but I'm also keeping these changes alive in a new branch: https://github.com/solson/miri/tree/always_encode_mir

Introduce and use specialized `//@ ignore-auxiliary` for test support files instead of using `//@ ignore-test` ### Summary Add a semantically meaningful directive for ignoring test *auxiliary* files. This is for auxiliary files that *participate* in actual tests but should not be built by `compiletest` (i.e. these files are involved through `mod xxx;` or `include!()` or `#[path = "xxx"]`, etc.). ### Motivation A specialized directive like `//@ ignore-auxiliary` makes it way easier to audit disabled tests via `//@ ignore-test`. - These support files cannot use the canonical `auxiliary/` dir because they participate in module resolution or are included, or their relative paths can be important for test intention otherwise. Follow-up to: - #139705 - #139783 - #139740 See also discussions in: - [#t-compiler > Directive name for non-test aux files?](https://rust-lang.zulipchat.com/#narrow/channel/131828-t-compiler/topic/Directive.20name.20for.20non-test.20aux.20files.3F/with/512773817) - [#t-compiler > Handling disabled `//@ ignore-test` tests](https://rust-lang.zulipchat.com/#narrow/channel/131828-t-compiler/topic/Handling.20disabled.20.60.2F.2F.40.20ignore-test.60.20tests/with/512005974) - [#t-compiler/meetings > [steering] 2025-04-11 Dealing with disabled tests](https://rust-lang.zulipchat.com/#narrow/channel/238009-t-compiler.2Fmeetings/topic/.5Bsteering.5D.202025-04-11.20Dealing.20with.20disabled.20tests/with/511717981) ### Remarks on remaining unconditionally disabled tests under `tests/` After this PR, against commit 79a272c6402, only **14** remaining test files are disabled through `//@ ignore-test`: <details> <summary>Remaining `//@ ignore-test` files under `tests/`</summary> ``` tests/debuginfo/drop-locations.rs 4://@ ignore-test (broken, see #128971) tests/rustdoc/macro-document-private-duplicate.rs 1://@ ignore-test (fails spuriously, see issue #89228) tests/rustdoc/inline_cross/assoc-const-equality.rs 3://@ ignore-test (FIXME: #125092) tests/ui/match/issue-27021.rs 7://@ ignore-test (#54987) tests/ui/match/issue-26996.rs 7://@ ignore-test (#54987) tests/ui/issues/issue-49298.rs 9://@ ignore-test (#54987) tests/ui/issues/issue-59756.rs 2://@ ignore-test (rustfix needs multiple suggestions) tests/ui/precondition-checks/write.rs 5://@ ignore-test (unimplemented) tests/ui/precondition-checks/read.rs 5://@ ignore-test (unimplemented) tests/ui/precondition-checks/write_bytes.rs 5://@ ignore-test (unimplemented) tests/ui/explicit-tail-calls/drop-order.rs 2://@ ignore-test: tail calls are not implemented in rustc_codegen_ssa yet, so this causes 🧊 tests/ui/panics/panic-short-backtrace-windows-x86_64.rs 3://@ ignore-test (#92000) tests/ui/json/json-bom-plus-crlf-multifile-aux.rs 3://@ ignore-test Not a test. Used by other tests tests/ui/traits/next-solver/object-soundness-requires-generalization.rs 2://@ ignore-test (see #114196) ``` </details> Of these, most are either **unimplemented**, or **spurious**, or **known-broken**. The outstanding one is `tests/ui/json/json-bom-plus-crlf-multifile-aux.rs` which I did not want to touch in *this* PR -- that aux file has load-bearing BOM and carriage returns and byte offset matters. I think those test files that require special encoding / BOM probably are better off as `run-make` tests. See #139968 for that aux file. ### Review advice - Best reviewed commit-by-commit. - The directive name diverged from the most voted `//@ auxiliary` because I think that's easy to confuse with `//@ aux-{crate,dir}`. r? compiler

…, r=scottmcm Make explicit that `TypeId`'s layout and size are unstable Or worded differently, explicitly remark non-stable-guarantee of `TypeId` layout and size. This PR makes no *additional* guarantees or non-guarantees, it only emphasizes that `TypeId`'s size and layout are unstable like any other `#[repr(Rust)]` types. This was discussed during [#t-compiler/meetings > [weekly] 2025-10-30 @ 💬](https://rust-lang.zulipchat.com/#narrow/channel/238009-t-compiler.2Fmeetings/topic/.5Bweekly.5D.202025-10-30/near/547949347), where the compiler team discussed a request rust-lang/rust#148265 to have the standard library (and language) commit to `TypeId` guaranteeing a size upper bound of 16 bytes. In the meeting, the consensus was: - We were sympathetic to the use case discussed in the request PR, however we feel like this stability guarantee is premature, given that there are unresolved questions surrounding the intended purpose of `TypeId`, and concerns surrounding its collision-resistance properties rust-lang/rust#10389 and rust-lang/rust#129014. We would prefer not making any of such guarantee until the collision-resistance concerns are resolved. - Committing to a stability guarantee on the size upper bound now would close the door to making `TypeId` larger (even if unlikely for perf reasons). Given that we have previously broken people who asserted the size of `TypeId` is 8 bytes, it was also discussed in the meeting that we should *explicitly* note that the size and layout of `TypeId` is not a stable guarantee, and is subject to changes between Rust releases, and thus cannot be relied upon -- if breakage in people's code is due to that assumption, it will be considered a won't-fix. - So even if `#[repr(Rust)]` types have unstable size and layout, this PR makes it explicit for `TypeId` since this type can feel "special" and users can be lead into thinking its size and layout is something they can rely upon. r? `@scottmcm` (or libs/libs-api/lang)

oli-obk added 3 commits December 14, 2016 17:06

rustup to rustc 1.15.0-dev (ace092f56 2016-12-13) (always_encode_mir)

69fa3eb

re-use mir-opt compiletest instead of rolling our own

8b8c743

enable auxiliary builds

9ec97ba

oli-obk commented Dec 14, 2016

View reviewed changes

solson suggested changes Dec 15, 2016

View reviewed changes

oli-obk added 3 commits December 15, 2016 09:56

remove unused import

2420360

check that the null terminator is defined and not part of a pointer

fd0c21e

improve variable name

0a79304

solson approved these changes Dec 15, 2016

View reviewed changes

solson merged commit cbffb3b into rust-lang:master Dec 15, 2016

solson mentioned this pull request Dec 15, 2016

Revert "rustup to rustc 1.15.0-dev (ace092f56 2016-12-13)" #94

Merged

oli-obk deleted the always_encode_mir branch December 15, 2016 09:21

solson mentioned this pull request Dec 17, 2016

Revert "Revert "rustup to rustc 1.15.0-dev (ace092f56 2016-12-13)"" #97

Merged

rustup to rustc 1.15.0-dev (ace092f56 2016-12-13) #93

rustup to rustc 1.15.0-dev (ace092f56 2016-12-13) #93

Uh oh!

Conversation

oli-obk commented Dec 14, 2016

Uh oh!

oli-obk Dec 14, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

solson commented Dec 15, 2016

Uh oh!

solson commented Dec 15, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

solson commented Dec 15, 2016

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

oli-obk Dec 14, 2016 •

edited

Loading

solson commented Dec 15, 2016 •

edited

Loading