LLVM Bugzilla is read-only and represents the historical archive of all LLVM issues filled before November 26, 2021. Use github to submit LLVM bugs

Bug 20183 - Libcxx test case failures on darwin (10.6): tor.pass, eval.pass, sleep_for.pass
Summary: Libcxx test case failures on darwin (10.6): tor.pass, eval.pass, sleep_for.pass
Status: RESOLVED FIXED
Alias: None
Product: libc++
Classification: Unclassified
Component: All Bugs (show other bugs)
Version: unspecified
Hardware: Macintosh MacOS X
: P normal
Assignee: Marshall Clow (home)
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-07-01 14:44 PDT by Chris Matthews
Modified: 2017-05-04 22:27 PDT (History)
4 users (show)

See Also:
Fixed By Commit(s):


Attachments
test log (476.80 KB, text/plain)
2014-07-01 15:16 PDT, Chris Matthews
Details
build log (12.18 KB, text/plain)
2014-07-01 15:17 PDT, Chris Matthews
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Chris Matthews 2014-07-01 14:44:30 PDT
The following tests have been failing on Darwin of a while now:

********************
Failing Tests (3):
    libc++ :: numerics/rand/rand.device/ctor.pass.cpp
    libc++ :: numerics/rand/rand.device/eval.pass.cpp
    libc++ :: thread/thread.threads/thread.thread.this/sleep_for.pass.cpp


The most recent results can be found here:

http://lab.llvm.org:8013/builders/libcxx_clang-x86_64-darwin11-RA?numbuilds=50
Comment 1 Marshall Clow (home) 2014-07-01 14:58:01 PDT
Interesting; I'm not seeing those failures in my nightly runs.
At least the first two appear to be a 10.8/10.9 thing; they pass on 10.8, fail on 10.9.
Comment 2 Marshall Clow (home) 2014-07-01 15:08:38 PDT
(In reply to comment #1)
> Interesting; I'm not seeing those failures in my nightly runs.
> At least the first two appear to be a 10.8/10.9 thing; they pass on 10.8,
> fail on 10.9.

This is just wrong; please disregard.
Comment 3 Marshall Clow (home) 2014-07-01 15:09:46 PDT
Are you running with a freshly built dylib?
These tests were added to test a change to src/random.cpp; code that gets included in the dylib.

If I run the tests with the libc++.dylib that Apple ships as part of 10.9, then (at least the first two tests) will fail.

P.S.  For some reason, I cannot connect to lab.llvm.org:8013
Comment 4 Chris Matthews 2014-07-01 15:16:53 PDT
Created attachment 12730 [details]
test log
Comment 5 Chris Matthews 2014-07-01 15:17:10 PDT
Created attachment 12731 [details]
build log
Comment 6 Chris Matthews 2014-07-01 15:17:50 PDT
I have added some logs that might be useful from a failing build.

These bots are running 10.6.
Comment 7 Marshall Clow (home) 2014-07-01 15:25:15 PDT
(In reply to comment #6)
> I have added some logs that might be useful from a failing build.
> 
> These bots are running 10.6.

Ah, that's the key.

Doug Gregor wrote (back in 2011, on the cfe-developers list):
> Apple only supports libc++ on Mac OS 10.7 ....

And Howard Hinnant wrote (in the same thread in 2011):
> Note that Apple never supported libc++ on 10.6, and I am not testing against it. 

[ The thread is was titled "libc++ on Mac OS X 10.6" and started on Oct 9, 2011 ]

Why are you running 10.6 on this bot?
Comment 8 Chris Matthews 2014-07-01 15:39:49 PDT
This bot is very old.  Even though it has not been officially supported, the bot has been green and finding occasional bugs.

If this bug is 10.6 specific, the bot could be taken off line.
Comment 9 Marshall Clow (home) 2014-07-01 15:41:59 PDT
I'm willing to try to figure out what's going on here; but if it's just a problem on 10.6, I'm going to give it a lower priority than other bugs.
Comment 10 Chris Matthews 2014-07-01 15:43:53 PDT
That sounds great. Thanks!
Comment 11 Eric Fiselier 2017-05-04 22:27:27 PDT
All of these bugs were fixed by David Majnemer!

`sleep_for.pass.cpp` test was fixed by commit r210210. Also I believe `ctor.pass.cpp` was fixed by commit r210060 and `eval.pass.cpp` was fixed by r210061.

All of these tests pass against the OS X 10.12 system library but not on 10.11.