Re: toolchain/49187: ATF detect_unused_tests tests failing

[Julio Merino <jmmv%netbsd.org@localhost>]

The following reply was made to PR toolchain/49187; it has been noted by GNATS.

From: Julio Merino <jmmv%netbsd.org@localhost>
To: Andreas Gustafsson <gson%gson.org@localhost>
Cc: gnats-bugs%netbsd.org@localhost
Subject: Re: toolchain/49187: ATF detect_unused_tests tests failing
Date: Thu, 11 Sep 2014 08:10:44 -0400

 On Thu, Sep 11, 2014 at 7:41 AM, Andreas Gustafsson <gson%gson.org@localhost> 
wrote:
 > Hi Julio,
 >
 > I would like to mark the failing detect_unused_tests test cases as
 > "expected failures"
 
 How is it failing?  Is the failure consistent?  In all platforms?
 Does it happen with gcc and clang?
 
 I remember fixing some failures in this specific test months ago, and
 now see that the fixes went into atf 0.20 (which is in the tree).  In
 fact, the test tries first to see if the compiler "works" by running a
 simple "pre-check" and then executes the real test.  I recall some
 compiler bug (can't remember if gcc or clang) that prevented this from
 working properly though so this test is actually reporting a real
 issue somewhere.
 
 > but it's not clear to me how to to do that since
 > the source for the tests in case is in src/external/bsd/atf rather
 > than src/tests.
 
 Ideally, just like with any other software in src/external/: you can
 patch it, but you should feed back any fixes upstream. And if the
 fixes/changes are not clear, try to discuss them with upstream first.
 Which is what you are doing here, so cool!
 
 > Does this need to go in the ATF distribution itself, or is there some
 > other way to do it?
 
 Well, if it really turns out to be a compiler problem... patching the
 test in NetBSD directly is probably reasonable.