[lldb-dev] tests xfailed globally with radar references
Vedant Kumar via lldb-dev
lldb-dev at lists.llvm.org
Tue Mar 13 14:39:33 PDT 2018
Thanks for bringing this up. I'll volunteer to take a look at these tests and file PRs where appropriate. I should get to this within a day or two.
> On Mar 13, 2018, at 10:31 AM, Ted Woodward via lldb-dev <lldb-dev at lists.llvm.org> wrote:
> I was investigating TestWatchedVarHitWhenInScope.py, which tests that a
> watchpoint is hit in scope and not hit when out of scope. It's xfailed due
> to a radar:
> This is an actual bug - watchpoints are hit when the address is modified but
> the variable is out of scope. Which can happen a lot if the variable being
> watched is a local on the stack. I manually ran the testcase on Linux, and
> it failed when it hit the watchpoint out of scope:
> Process 18652 stopped
> * thread #1, name = 'a.out', stop reason = watchpoint 1
> frame #0: 0x00007ffff722d0d2 libc.so.6`__run_exit_handlers(status=0,
> listp=0x00007ffff75b36c8, run_list_atexit=true) at exit.c:35
> A google search found this message, posted by Ed Maste to llvm-dev:
> http://lists.llvm.org/pipermail/llvm-dev/2017-August/116471.html .
> I echo Ed's request - could someone at Apple look at these radars, and open
> llvm bugs for the ones that aren't Darwin-only?
> Ed's message:
> I was recently looking at an unexpected pass test result on FreeBSD
> which is decorated with expectedFailureAll referencing
> rdar://problem/24599697. I'm looking for more details on this specific
> radar bug report in order to determine if the test is not exercising
> the original bug on FreeBSD, or if it's really an Apple-only issue.
> As it's rather unfortunate to have a bug report reference for which no
> further public details are available I looked at other radar
> references in the test suite, and found 118 of them. Many of these are
> in comments, in macosx-specific cases (e.g.
> @expectedFailureAll(oslist=["macosx"], bugnumber="rdar://28805064")),
> or in tests decorated with skipUnlessDarwin, and I've ignored those.
> There are 10 tests with expected failure decorators, and one skipped,
> that apply beyond macosx and reference a radar (and do not
> additionally reference an LLVM PR). Can someone at Apple check these
> radar references and submit LLVM PRs, reference existing PRs, add a
> reasonable description in the test source, etc. as appropriate?
> @expectedFailureAll(archs=['i386'], bugnumber="rdar://28658860")
> @unittest2.expectedFailure("rdar://problem/10808472 SBValue::Cast test
> case is failing (virtual inheritance)")
> self.skipTest("rdar://problem/9173060 lldb hangs while running
> unique-types for clang version < 3")
> Qualcomm Innovation Center, Inc.
> The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a
> Linux Foundation Collaborative Project
> lldb-dev mailing list
> lldb-dev at lists.llvm.org
More information about the lldb-dev