Hack regression: Incorrect case throwing duplicate errors

[simonwelsh]

HHVM Version

3.14.1

Standalone code, or other way to reproduce the problem

<?hh // strict

namespace phpunit\hhi\test {

use phpunit\framework\TestCase as TestCase;

class Test extends TestCase {}

}

namespace PHPUnit\Framework {
class TestCase {}
}

Expected result

vagrant@homestead:~/test$ hh_client
test.php:7:20,27: Could not find phpunit\framework\TestCase (Naming[2006])
  test.php:12:7,14: Did you mean PHPUnit\Framework\TestCase?
vagrant@homestead:~/test$ 

Actual result

vagrant@homestead:~/test$ hh_client
test.php:7:20,27: Could not find phpunit\framework\TestCase (Naming[2006])
  test.php:12:7,14: Did you mean PHPUnit\Framework\TestCase?
test.php:7:20,27: Could not find phpunit\framework\TestCase (Naming[2006])
  test.php:12:7,14: Did you mean PHPUnit\Framework\TestCase?
vagrant@homestead:~/test$ 

This has only started happening after upgrading to 3.14, which is breaking my phpunit-hhi unit tests (3.14 build, 3.13 build, 3.9 build)

[jwatzman]

@simonwelsh any chance you could bisect?

cc @dlreeves -- duplicate errors like this usually indicate we're running the same checks twice, so there could be an easy perf win in fixing this

[jwatzman]

cc @andrewjkennedy too, who I just saw on an internal task about check_todo running more than once which is a potential cause of this :)

[simonwelsh]

c354e48 is the first bad commit

commit c354e482dc73ea4487ce22a75b11d0e554a938ac
Author: Jez Ng <jezng@fb.com>
Date:   Sat May 7 15:04:33 2016 -0700

    Kill the Nast heap

    Reviewed By: dabek

    Differential Revision: D3241214

    fb-gh-sync-id: fae8bc92260c4f7cc631ca371bfde71be254a20e
    fbshipit-source-id: fae8bc92260c4f7cc631ca371bfde71be254a20e

:040000 040000 6fd85470748b47713ad5380a9f6f5216fbf240db a215bc8248410cc1f6821a57bcf7dff3f4713770 M  php

[simonwelsh]

Back to only one error