Turn `trans_fulfill_obligation` into a query #44967

wesleywiser · 2017-10-02T00:53:30Z

rust-highfive · 2017-10-02T00:53:41Z

(rust_highfive has picked a reviewer for you, use r? to override)

wesleywiser · 2017-10-02T01:01:57Z

src/librustc/traits/trans/mod.rs

+/// Assumes that this is run after the entire crate has been successfully type-checked.
+pub fn trans_fulfill_obligation<'a, 'tcx>(ty: TyCtxt<'a, 'tcx, 'tcx>,
+                                          (def_id, param_env, trait_ref):
+                                          (DefId, ty::ParamEnv<'tcx>, ty::PolyTraitRef<'tcx>))


I changed this to take a DefId instead of a Span because it looked like all of the other queries used DefId and not Spans. However, there were two places where DUMMY_SP was passed in. I changed those to use DefIds that seemed appropriate but I'm not sure. I'll indicate those below.

If that wasn't the right thing to do, I can revert it.

Hm, @nikomatsakis would know more about how the span here is actually used.

wesleywiser · 2017-10-02T01:02:22Z

src/librustc_trans/monomorphize.rs

@@ -113,7 +111,7 @@ fn resolve_associated_item<'a, 'tcx>(

    let trait_ref = ty::TraitRef::from_method(tcx, trait_id, rcvr_substs);
    let vtbl = tcx.trans_fulfill_obligation(
-        DUMMY_SP, ty::ParamEnv::empty(traits::Reveal::All), ty::Binder(trait_ref));
+        (trait_id, ty::ParamEnv::empty(traits::Reveal::All), ty::Binder(trait_ref)));


This was the first place DUMMY_SP was passed in.

wesleywiser · 2017-10-02T01:02:36Z

src/librustc_trans/monomorphize.rs

        substs: tcx.mk_substs_trait(source_ty, &[target_ty])
    });

    match tcx.trans_fulfill_obligation(
-        DUMMY_SP, ty::ParamEnv::empty(traits::Reveal::All), trait_ref) {
+        (def_id, ty::ParamEnv::empty(traits::Reveal::All), trait_ref)) {


This was the other place.

michaelwoerister · 2017-10-02T13:52:44Z

src/librustc/ich/impls_ty.rs

+    fn hash_stable<W: StableHasherResult>(&self,
+                                          hcx: &mut StableHashingContext<'gcx>,
+                                          hasher: &mut StableHasher<W>) {
+        use traits::Vtable::*;


The enum discriminant also needs to be hashed:

mem::discriminant(self).hash_stable(hcx, hasher);

michaelwoerister · 2017-10-02T13:59:46Z

src/librustc/ich/impls_ty.rs

+    fn hash_stable<W: StableHasherResult>(&self,
+                                          hcx: &mut StableHashingContext<'gcx>,
+                                          hasher: &mut StableHasher<W>) {
+        self.impl_def_id.hash_stable(hcx, hasher);


So what we usually do here is use an exhaustive destructuring let statement, so in case a field gets added or removed, we don't miss updating the HashStable implementation:

let traits::VtableImplData { impl_def_id, substs, ref nested, } = *self; impl_def_id.hash_stable(hcx, hasher); substs.hash_stable(hcx, hasher); nested.hash_stable(hcx, hasher);

That means a bit more typing but has proven to be very useful in catching oversights.

michaelwoerister · 2017-10-02T14:02:35Z

Very nice, thank you, @wesleywiser!

Please update the HashStable implementation to use destructuring let statements.

I'll let @nikomatsakis do the rest of the review since trait selection is his field.

arielb1 · 2017-10-02T15:50:03Z

r? @arielb1

@wesleywiser

I think you should just not pass the span/defid. It's just used for error reporting, and passing the span around as the query key will prevent all caching.

nikomatsakis · 2017-10-03T11:51:46Z

I agree with @arielb1 -- the query key should just be ParamEnv and TraitRef, we don't need the def-id, we can instead use a dummy span.

nikomatsakis · 2017-10-03T11:52:29Z

src/librustc/traits/trans/mod.rs

+
+        let span = ty.def_span(def_id);
+
+        let obligation_cause = ObligationCause::misc(span,


here you can use ObligationCause::dummy() instead

nikomatsakis · 2017-10-03T11:53:31Z

src/librustc/traits/trans/mod.rs

+                debug!("Encountered ambiguity selecting `{:?}` during trans, \
+                        presuming due to overflow",
+                        trait_ref);
+                ty.sess.span_fatal(span,


I actually think this case .. hmm .. maybe cannot occur? That is, @arielb1, would that code you added that causes compilation to abort when types get too large perhaps kick in first?

bors · 2017-10-03T14:48:04Z

☔ The latest upstream changes (presumably #44896) made this pull request unmergeable. Please resolve the merge conflicts.

wesleywiser · 2017-10-04T19:38:54Z

src/librustc/traits/trans/mod.rs

+                                        (selection ambiguity)");
+            }
+            Err(e) => {
+                span_bug!(span, "Encountered error `{:?}` selecting `{:?}` during trans",


What should I do about this span here? Pass DUMMY_SP?

Just do a bug! I think. I think for better error reporting we should just dump the query stack during each ICE.

wesleywiser · 2017-10-06T14:46:04Z

I believe I've resolved all of the review feedback.

theotherjimmy · 2017-10-08T15:21:39Z

src/librustc/ty/maps/plumbing.rs

@@ -765,6 +765,7 @@ pub fn force_from_dep_node<'a, 'gcx, 'lcx>(tcx: TyCtxt<'a, 'gcx, 'lcx>,
        DepKind::SpecializationGraph => { force!(specialization_graph_of, def_id!()); }
        DepKind::ObjectSafety => { force!(is_object_safe, def_id!()); }
        DepKind::TraitImpls => { force!(trait_impls_of, def_id!()); }
+        DepKind::FulfillObligation => { force!(item_body_nested_bodies, def_id!()); }


Should this be forcing fulfull_obligation instead?

Fixed by making it a bug. I don't know how we'd recreate the query key so that seems appropriate to me, but I'm not 100% sure.

Part of rust-lang#44891

nikomatsakis

So this looks right to me. The only question in my mind is if the one case -- which used to be a span_fatal and is now a bug! is truly a bug. If not, I suspect that we should convert the query to yield a Result or Option or something, and let the caller report the problem.

nikomatsakis · 2017-10-09T14:47:37Z

@bors r+

bors · 2017-10-09T14:47:38Z

📌 Commit 31f4b57 has been approved by nikomatsakis

bors · 2017-10-10T23:32:25Z

⌛ Testing commit 31f4b57 with merge bf27afeb1a9cf122bd5b2cdfbc765e7f02840119...

bors · 2017-10-10T23:40:12Z

💔 Test failed - status-travis

wesleywiser · 2017-10-11T01:45:22Z

Android builder failed. Seems related to some kind of network issue?

[00:03:56] + /android/sdk/tools/android update sdk -a --no-ui --filter platform-tools,android-18,sys-img-armeabi-v7a-android-18
[00:03:57] Refresh Sources:
[00:03:57] Fetching https://dl.google.com/android/repository/addons_list-2.xml
[00:04:02] Failed to fetch URL https://dl.google.com/android/repository/addons_list-2.xml, reason: peer not authenticated
[00:04:02] Fetched Add-ons List successfully
[00:04:02] Refresh Sources
[00:04:02] Fetching URL: https://dl.google.com/android/repository/repository-11.xml
[00:04:02] Failed to fetch URL https://dl.google.com/android/repository/repository-11.xml, reason: SSLPeerUnverified peer not authenticated
[00:04:02] Refresh Sources:
[00:04:02] Fetching URL: https://dl.google.com/android/repository/repository-11.xml
[00:04:02] Failed to fetch URL https://dl.google.com/android/repository/repository-11.xml, reason: SSLPeerUnverified peer not authenticated
[00:04:02] There is nothing to install or update.
[00:04:06] + true
[00:04:06] + echo yes
[00:04:06] + create_avd armeabi-v7a 18
[00:04:06] + abi=armeabi-v7a
[00:04:06] + api=18
[00:04:06] + echo no
[00:04:06] + /android/sdk/tools/android create avd --name armeabi-v7a-18 --target android-18 --abi armeabi-v7a
[00:04:06] Error: Target id is not valid. Use 'android list targets' to get the target ids.
[00:04:07] The command '/bin/sh -c . /scripts/android-sdk.sh && download_and_create_avd tools_r25.2.5-linux.zip armeabi-v7a 18' returned a non-zero code: 1
[00:04:07] The command has failed after 5 attempts.
The command "stamp sh -x -c "$RUN_SCRIPT"" exited with 1.

kennytm · 2017-10-11T06:32:28Z

@bors retry

Android SDK HTTPS issue (fixed in ci: Fix installing the Android SDK #45193)

bors · 2017-10-11T23:14:19Z

⌛ Testing commit 31f4b57 with merge 58bb6295629d03131e13923d939ef1401efb4451...

bors · 2017-10-12T01:00:10Z

💔 Test failed - status-travis

wesleywiser · 2017-10-12T02:28:38Z

Not sure what this means:

[01:40:10] ------------------------------------------
[01:40:10] stderr:
[01:40:10] ------------------------------------------
[01:40:10]
[01:40:10] ------------------------------------------
[01:40:10]
[01:40:10] thread '[debuginfo-lldb] debuginfo/lexical-scopes-in-block-expression.rs' panicked at 'explicit panic', src/tools/compiletest/src/runtest.rs:2425:8
[01:40:10] note: Run with RUST_BACKTRACE=1 for a backtrace.
[01:40:10]
[01:40:10]
[01:40:10] failures:
[01:40:10] [debuginfo-lldb] debuginfo/lexical-scopes-in-block-expression.rs
[01:40:10]
[01:40:10] test result: �[31mFAILED�(B�[m. 97 passed; 1 failed; 10 ignored; 0 measured; 0 filtered out
[01:40:10]
[01:40:10] thread 'main' panicked at 'Some tests failed', src/tools/compiletest/src/main.rs:323:21
[01:40:10]
[01:40:10]
[01:40:10] command did not execute successfully: "/Users/travis/build/rust-lang/rust/build/x86_64-apple-darwin/stage0-tools/x86_64-apple-darwin/release/compiletest" "--compile-lib-path" "/Users/travis/build/rust-lang/rust/build/x86_64-apple-darwin/stage2/lib" "--run-lib-path" "/Users/travis/build/rust-lang/rust/build/x86_64-apple-darwin/stage2/lib/rustlib/x86_64-apple-darwin/lib" "--rustc-path" "/Users/travis/build/rust-lang/rust/build/x86_64-apple-darwin/stage2/bin/rustc" "--src-base" "/Users/travis/build/rust-lang/rust/src/test/debuginfo" "--build-base" "/Users/travis/build/rust-lang/rust/build/x86_64-apple-darwin/test/debuginfo" "--stage-id" "stage2-x86_64-apple-darwin" "--mode" "debuginfo-lldb" "--target" "x86_64-apple-darwin" "--host" "x86_64-apple-darwin" "--llvm-filecheck" "/Users/travis/build/rust-lang/rust/build/x86_64-apple-darwin/llvm/build/bin/FileCheck" "--nodejs" "/Users/travis/.nvm/versions/node/v6.9.1/bin/node" "--host-rustcflags" "-Crpath -O" "--target-rustcflags" "-Crpath -O -Lnative=/Users/travis/build/rust-lang/rust/build/x86_64-apple-darwin/native/rust-test-helpers" "--docck-python" "/usr/local/opt/python/bin/python2.7" "--lldb-python" "/usr/bin/python" "--lldb-version" "lldb-360.1.70" "--lldb-python-dir" "/Applications/Xcode.app/Contents/SharedFrameworks/LLDB.framework/Resources/Python" "--llvm-version" "4.0.1\n" "--cc" "" "--cxx" "" "--cflags" "" "--llvm-components" "" "--llvm-cxxflags" "" "--adb-path" "adb" "--adb-test-dir" "/data/tmp/work" "--android-cross-path" "" "--color" "always"
[01:40:10] expected success, got: exit code: 101
[01:40:10]
[01:40:10]
[01:40:10] failed to run: /Users/travis/build/rust-lang/rust/build/bootstrap/debug/bootstrap test
[01:40:10] Build completed unsuccessfully in 0:28:09
[01:40:10] make: *** [check] Error 1
[01:40:10] make: INTERNAL: Exiting with 1 jobserver tokens available; should be 2!

kennytm · 2017-10-12T06:08:55Z

@bors retry LLDB segfault.

bors · 2017-10-12T07:48:18Z

⌛ Testing commit 31f4b57 with merge 39fd958...

…tsakis Turn `trans_fulfill_obligation` into a query Part of #44891

bors · 2017-10-12T10:09:43Z

☀️ Test successful - status-appveyor, status-travis
Approved by: nikomatsakis
Pushing 39fd958 to master...

rust-highfive assigned nikomatsakis Oct 2, 2017

wesleywiser commented Oct 2, 2017

View reviewed changes

michaelwoerister reviewed Oct 2, 2017

View reviewed changes

arielb1 assigned arielb1 and unassigned nikomatsakis Oct 2, 2017

carols10cents added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Oct 2, 2017

nikomatsakis reviewed Oct 3, 2017

View reviewed changes

wesleywiser commented Oct 4, 2017

View reviewed changes

wesleywiser force-pushed the trans_fulfill_obligation branch 2 times, most recently from dbddd4e to a9862b0 Compare October 6, 2017 03:52

theotherjimmy reviewed Oct 8, 2017

View reviewed changes

Turn trans_fulfill_obligation into a query

31f4b57

Part of rust-lang#44891

wesleywiser force-pushed the trans_fulfill_obligation branch from a9862b0 to 31f4b57 Compare October 8, 2017 15:59

nikomatsakis approved these changes Oct 9, 2017

View reviewed changes

nikomatsakis assigned nikomatsakis and unassigned arielb1 Oct 9, 2017

kennytm mentioned this pull request Oct 12, 2017

Spurious LLDB segfault (signal 11) #45230

Open

bors added a commit that referenced this pull request Oct 12, 2017

Auto merge of #44967 - wesleywiser:trans_fulfill_obligation, r=nikoma…

39fd958

…tsakis Turn `trans_fulfill_obligation` into a query Part of #44891

bors merged commit 31f4b57 into rust-lang:master Oct 12, 2017

This was referenced Oct 12, 2017

Queryify Vtable methods #45137

Merged

Create NormalizeTy query #44984

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Turn `trans_fulfill_obligation` into a query #44967

Turn `trans_fulfill_obligation` into a query #44967

wesleywiser commented Oct 2, 2017

rust-highfive commented Oct 2, 2017

wesleywiser Oct 2, 2017

michaelwoerister Oct 2, 2017

wesleywiser Oct 2, 2017

wesleywiser Oct 2, 2017

michaelwoerister Oct 2, 2017

michaelwoerister Oct 2, 2017 •

edited

Loading

michaelwoerister commented Oct 2, 2017

arielb1 commented Oct 2, 2017

nikomatsakis commented Oct 3, 2017

nikomatsakis Oct 3, 2017

nikomatsakis Oct 3, 2017 •

edited

Loading

bors commented Oct 3, 2017

wesleywiser Oct 4, 2017

arielb1 Oct 4, 2017 •

edited

Loading

wesleywiser commented Oct 6, 2017

theotherjimmy Oct 8, 2017

wesleywiser Oct 8, 2017

nikomatsakis left a comment

nikomatsakis commented Oct 9, 2017

bors commented Oct 9, 2017

bors commented Oct 10, 2017

bors commented Oct 10, 2017

wesleywiser commented Oct 11, 2017

kennytm commented Oct 11, 2017

bors commented Oct 11, 2017

bors commented Oct 12, 2017

wesleywiser commented Oct 12, 2017

kennytm commented Oct 12, 2017

bors commented Oct 12, 2017

bors commented Oct 12, 2017


		let span = ty.def_span(def_id);

		let obligation_cause = ObligationCause::misc(span,

Turn trans_fulfill_obligation into a query #44967

Turn trans_fulfill_obligation into a query #44967

Conversation

wesleywiser commented Oct 2, 2017

rust-highfive commented Oct 2, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

michaelwoerister Oct 2, 2017 • edited Loading

Choose a reason for hiding this comment

michaelwoerister commented Oct 2, 2017

arielb1 commented Oct 2, 2017

nikomatsakis commented Oct 3, 2017

Choose a reason for hiding this comment

nikomatsakis Oct 3, 2017 • edited Loading

Choose a reason for hiding this comment

bors commented Oct 3, 2017

Choose a reason for hiding this comment

arielb1 Oct 4, 2017 • edited Loading

Choose a reason for hiding this comment

wesleywiser commented Oct 6, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nikomatsakis left a comment

Choose a reason for hiding this comment

nikomatsakis commented Oct 9, 2017

bors commented Oct 9, 2017

bors commented Oct 10, 2017

bors commented Oct 10, 2017

wesleywiser commented Oct 11, 2017

kennytm commented Oct 11, 2017

bors commented Oct 11, 2017

bors commented Oct 12, 2017

wesleywiser commented Oct 12, 2017

kennytm commented Oct 12, 2017

bors commented Oct 12, 2017

bors commented Oct 12, 2017

Turn `trans_fulfill_obligation` into a query #44967

Turn `trans_fulfill_obligation` into a query #44967

michaelwoerister Oct 2, 2017 •

edited

Loading

nikomatsakis Oct 3, 2017 •

edited

Loading

arielb1 Oct 4, 2017 •

edited

Loading