MIR-borrowck: augmented assignment causes duplicate errors #47607

davidtwco · 2018-01-20T02:19:38Z

Fixes #45697. This PR resolves the error duplication. I attempted to replace the existing sets since there were quite a few but only managed to replace two of them.

r? @nikomatsakis

nikomatsakis · 2018-01-22T18:02:02Z

@bors r+

nice!

bors · 2018-01-22T18:02:03Z

📌 Commit 09d6682 has been approved by nikomatsakis

alexcrichton · 2018-01-25T21:50:08Z

@bors: r-

I think this may have caused this failure

davidtwco · 2018-01-25T22:18:17Z

It seems like the other of the two errors is being emitted now where it was being supressed before. I assume another PR in the rollup has changed something that causes the error order to change?

alexcrichton · 2018-01-25T22:34:08Z

Hm yeah that may also be true! I'm not sure as to which though :(

davidtwco · 2018-01-25T23:06:06Z

Just pushed rebased changes, locally I'm not able to reproduce the issue yet and looking through the files that the rollup PRs have touched, nothing is jumping out at me as being the likely culprit so I'll keep digging.

alexcrichton · 2018-01-25T23:09:40Z

@bors: r=nikomatsakis

Ok let's keep this in the queue and see what happens, I may also be diagnosing in error!

bors · 2018-01-25T23:09:41Z

📌 Commit fd67863 has been approved by nikomatsakis

alexcrichton · 2018-01-25T23:09:49Z

er sorry,

@bors: r=nikomatsakis

bors · 2018-01-25T23:09:50Z

💡 This pull request was already approved, no need to approve it again.

There's another pull request that is currently being tested, blocking this pull request: Libtest json output #46450

bors · 2018-01-25T23:09:50Z

📌 Commit fd67863 has been approved by nikomatsakis

bors · 2018-01-27T04:12:13Z

💔 Test failed - status-travis

kennytm · 2018-01-27T07:53:32Z

src/test/ui/issue-45697.stderr

+29 |         let z = copy_borrowed_ptr(&mut y);
+   |                                   ------ borrow of `*y.pointer` occurs here
+30 |         *y.pointer += 1;
+   |         ^^^^^^^^^^^^^^^ assignment to borrowed `*y.pointer` occurs here


The borrow checker says it's a "use of y", not "assignment of y.pointer".

[00:55:48] failures: [00:55:48] [00:55:48] ---- [ui] ui/issue-45697.rs stdout ---- [00:55:48] diff of stderr: [00:55:48] [00:55:48] 6 30 | *y.pointer += 1; [00:55:48] 7 | ^^^^^^^^^^^^^^^ assignment to borrowed `*y.pointer` occurs here [00:55:48] 8 [00:55:48] + error[E0503]: cannot use `*y.pointer` because it was mutably borrowed (Mir) [00:55:48] - error[E0506]: cannot assign to `*y.pointer` because it is borrowed (Mir) [00:55:48] 10 --> $DIR/issue-45697.rs:30:9 [00:55:48] 11 | [00:55:48] 12 29 | let z = copy_borrowed_ptr(&mut y); [00:55:48] [00:55:48] + | ------ borrow of `y` occurs here [00:55:48] - | ------ borrow of `*y.pointer` occurs here [00:55:48] 14 30 | *y.pointer += 1; [00:55:48] + | ^^^^^^^^^^^^^^^ use of borrowed `y` [00:55:48] - | ^^^^^^^^^^^^^^^ assignment to borrowed `*y.pointer` occurs here [00:55:48] 16 [00:55:48] 17 error: aborting due to 2 previous errors [00:55:48] 18

Hmm, seems like the issue we saw in the rollup again. I'll rebase it and take a look.

Alright, so after a healthy amount of confusion and recompiling, I've made a little progress - it seems my changes simultaneously work and don't work:

/mnt/c/Users/David/Projects/personal/rust issue-45697* ↓↑ ❯ ./build/x86_64-unknown-linux-gnu/stage1/bin/rustc src/test/ui/issue-45697.rs -Z borrowck=compare -Z emit-end-regions error[E0506]: cannot assign to `*y.pointer` because it is borrowed (Ast) --> src/test/ui/issue-45697.rs:30:9 | 29 | let z = copy_borrowed_ptr(&mut y); | - borrow of `*y.pointer` occurs here 30 | *y.pointer += 1; | ^^^^^^^^^^^^^^^ assignment to borrowed `*y.pointer` occurs here error[E0503]: cannot use `*y.pointer` because it was mutably borrowed (Mir) --> src/test/ui/issue-45697.rs:30:9 | 29 | let z = copy_borrowed_ptr(&mut y); | ------ borrow of `y` occurs here 30 | *y.pointer += 1; | ^^^^^^^^^^^^^^^ use of borrowed `y` error: aborting due to 2 previous errors /mnt/c/Users/David/Projects/personal/rust issue-45697* ↓↑ ❯ ./build/x86_64-unknown-linux-gnu/stage1/bin/rustc src/test/ui/issue-45697.rs -Z borrowck=compare -Z emit-end-regions -O error[E0506]: cannot assign to `*y.pointer` because it is borrowed (Ast) --> src/test/ui/issue-45697.rs:30:9 | 29 | let z = copy_borrowed_ptr(&mut y); | - borrow of `*y.pointer` occurs here 30 | *y.pointer += 1; | ^^^^^^^^^^^^^^^ assignment to borrowed `*y.pointer` occurs here error[E0506]: cannot assign to `*y.pointer` because it is borrowed (Mir) --> src/test/ui/issue-45697.rs:30:9 | 29 | let z = copy_borrowed_ptr(&mut y); | ------ borrow of `*y.pointer` occurs here 30 | *y.pointer += 1; | ^^^^^^^^^^^^^^^ assignment to borrowed `*y.pointer` occurs here error: aborting due to 2 previous errors

It seems that when the -O flag is passed (as it is with compiletest), the write pass happens first in the MIR borrowck, but when it isn't passed, the read pass happens first. It's the write pass error we want, so that's a problem. Now, I've got to assume that when the tests are run for a merge, -O isn't passed since that's the only thing I've found that really makes a difference here (edit: in fact, we can see that it isn't in the failed travis output).

When I was working on this PR initially, I ran all my tests through x.py, and therefore thought my tests passed (and they did, technically) and Travis on the PR confirms that. Not sure what the solution is here, but I'm going to keep digging.

I guess one solution would be to add -O to the compile-flags of the test? Doesn't seem ideal though.

Tried to dig around and find out what's causing this, but struggling to make any decent progress, would appreciate any pointers as to where I should be looking - @nikomatsakis.

In the paragraphs below, I'm defining passing run as when -O is passed as an argument and the intended error is output. Failing runs are everything else.

So, on the passing runs, the first error is from a Shallow/Write(Mutate) call to access_place from mutate_place which was called by visit_statement_entry. On the failing runs, we see a Deep/Read(Copy) call from the same places.

For the second error, on passing runs, we see a Deep/Read(Copy) run from access_place <- consume_operator <- visit_terminator_entry. On failing runs, we see a Shallow/Write(Mutate) call from access_place <- mutate_place <- visit_statement_entry.

Logs from the passing run and logs from the failing run. These logs include some logging statements I've added to help me trace the call stack back.

As best I can tell, the issue isn't in the borrow_check module but in the statements/terminator that is provided to it from the dataflow or transform module differing when -O is passed and otherwise.

…place/span.

shepmaster · 2018-02-03T01:47:08Z

Ping from triage, @davidtwco! Will you have time to address the strange errors?

davidtwco · 2018-02-03T09:11:19Z

@shepmaster Yes, just been looking at other in-progress PRs. Will turn focus back on this one.

nikomatsakis · 2018-02-05T21:34:23Z

OK, @davidtwco and I chatted a bit about this over gitter. First off, the problem arises because -- without -O -- we generate code like

tmp = CheckedAdd(*y.pointer, 1);
...
*y.pointer = tmp

but with -O we elide the bounds check and you get

*y.pointer = Add(*y.pointer, 1)

This, combined with the fact that we visit the LHS of an assignment first means that -- with -O only -- we see the write first:

rust/src/librustc_mir/borrow_check/mod.rs

Lines 351 to 357 in 6c04c41

    
           self.mutate_place( 
        
               ContextKind::AssignLhs.new(location), 
        
               (lhs, span), 
        
               Shallow(None), 
        
               JustWrite, 
        
               flow_state, 
        
           );

We could do a bunch of work to fix this, but I personally am unconvinced that showing an error about the write is better than showing an error about the read (after all, both are illegal). I'm wondering if we should just visit the RHS of an assignment so that the order is consistent.

…e pass ordering when -O is passed.

nikomatsakis · 2018-02-06T17:10:51Z

@davidtwco remind me, are we already testing (on travis) both with and without -O? I feel like it'd be nice to have a test to make sure we don't overlook the overflow checks in future.

davidtwco · 2018-02-06T17:13:25Z

@nikomatsakis when a PR is tested it uses -O, when bors runs tests before a merge, it does not.

nikomatsakis · 2018-02-06T17:24:28Z

@davidtwco that doesn't feel very intentional. It's not really about -O anyway, it's about overflow checks. How about we make two versions of the test (or two revisions), one using -C overflow-checks=on and one using -C overflow-checks=off? (You can pass flags with // compile-flags)

davidtwco · 2018-02-06T17:38:48Z

@nikomatsakis I've pushed a commit with a new ui test with overflow checks on, and modified the existing tests to have overflow checks on.

nikomatsakis · 2018-02-06T18:47:04Z

@bors r+

bors · 2018-02-06T18:47:05Z

📌 Commit bb6e54d has been approved by nikomatsakis

bors · 2018-02-07T02:20:30Z

⌛ Testing commit bb6e54d with merge 4f93357...

@nikomatsakis

MIR-borrowck: augmented assignment causes duplicate errors Fixes #45697. This PR resolves the error duplication. I attempted to replace the existing sets since there were quite a few but only managed to replace two of them. r? @nikomatsakis

bors · 2018-02-07T05:08:25Z

☀️ Test successful - status-appveyor, status-travis
Approved by: nikomatsakis
Pushing 4f93357 to master...

rust-highfive assigned nikomatsakis Jan 20, 2018

shepmaster added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Jan 20, 2018

nikomatsakis added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jan 23, 2018

alexcrichton mentioned this pull request Jan 25, 2018

Rollup of 19 pull requests #47748

Merged

davidtwco force-pushed the issue-45697 branch from 09d6682 to fd67863 Compare January 25, 2018 23:05

alexcrichton closed this Jan 25, 2018

alexcrichton reopened this Jan 25, 2018

kennytm reviewed Jan 27, 2018

View reviewed changes

kennytm added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Jan 27, 2018

davidtwco added 5 commits January 27, 2018 13:33

Added test for rust-lang#45697

dd6b005

Updated other affected tests.

f1c1db6

Introduced a new set to stop duplicate errors from MIR passes on one …

3daa4d2

…place/span.

Encompassed error deduplication of some existing sets in the ctxt.

f92d679

Added logging for error suppression.

970fb1a

davidtwco force-pushed the issue-45697 branch from fd67863 to 970fb1a Compare February 5, 2018 20:47

Swapped order of left/right visits to ensure consistency in read/writ…

5cd4b4f

…e pass ordering when -O is passed.

Added and updated tests to enable/disable overflow checks.

bb6e54d

nikomatsakis approved these changes Feb 6, 2018

View reviewed changes

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Feb 6, 2018

bors merged commit bb6e54d into rust-lang:master Feb 7, 2018

davidtwco deleted the issue-45697 branch February 25, 2018 12:18

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

MIR-borrowck: augmented assignment causes duplicate errors #47607

MIR-borrowck: augmented assignment causes duplicate errors #47607

davidtwco commented Jan 20, 2018

nikomatsakis commented Jan 22, 2018

bors commented Jan 22, 2018

alexcrichton commented Jan 25, 2018

davidtwco commented Jan 25, 2018

alexcrichton commented Jan 25, 2018

davidtwco commented Jan 25, 2018

alexcrichton commented Jan 25, 2018

bors commented Jan 25, 2018

alexcrichton commented Jan 25, 2018

bors commented Jan 25, 2018

bors commented Jan 25, 2018

bors commented Jan 27, 2018

kennytm Jan 27, 2018 •

edited

Loading

davidtwco Jan 27, 2018

davidtwco Jan 27, 2018 •

edited

Loading

davidtwco Jan 27, 2018

shepmaster commented Feb 3, 2018

davidtwco commented Feb 3, 2018 •

edited

Loading

nikomatsakis commented Feb 5, 2018

nikomatsakis commented Feb 6, 2018

davidtwco commented Feb 6, 2018

nikomatsakis commented Feb 6, 2018

davidtwco commented Feb 6, 2018

nikomatsakis commented Feb 6, 2018

bors commented Feb 6, 2018

bors commented Feb 7, 2018

bors commented Feb 7, 2018

MIR-borrowck: augmented assignment causes duplicate errors #47607

MIR-borrowck: augmented assignment causes duplicate errors #47607

Conversation

davidtwco commented Jan 20, 2018

nikomatsakis commented Jan 22, 2018

bors commented Jan 22, 2018

alexcrichton commented Jan 25, 2018

davidtwco commented Jan 25, 2018

alexcrichton commented Jan 25, 2018

davidtwco commented Jan 25, 2018

alexcrichton commented Jan 25, 2018

bors commented Jan 25, 2018

alexcrichton commented Jan 25, 2018

bors commented Jan 25, 2018

bors commented Jan 25, 2018

bors commented Jan 27, 2018

kennytm Jan 27, 2018 • edited Loading

Choose a reason for hiding this comment

davidtwco Jan 27, 2018

Choose a reason for hiding this comment

davidtwco Jan 27, 2018 • edited Loading

Choose a reason for hiding this comment

davidtwco Jan 27, 2018

Choose a reason for hiding this comment

shepmaster commented Feb 3, 2018

davidtwco commented Feb 3, 2018 • edited Loading

nikomatsakis commented Feb 5, 2018

nikomatsakis commented Feb 6, 2018

davidtwco commented Feb 6, 2018

nikomatsakis commented Feb 6, 2018

davidtwco commented Feb 6, 2018

nikomatsakis commented Feb 6, 2018

bors commented Feb 6, 2018

bors commented Feb 7, 2018

bors commented Feb 7, 2018

kennytm Jan 27, 2018 •

edited

Loading

davidtwco Jan 27, 2018 •

edited

Loading

davidtwco commented Feb 3, 2018 •

edited

Loading