[Lldb-commits] [lldb] r270745 - Mark some arm-linux specific xfails marking bug entries

[Pavel Labath via lldb-commits]

Omair,

please be careful about using arm xfails. You are using too wide
annotations and disabling tests even on configurations where they are
known to pass.


On 25 May 2016 at 19:48, Omair Javaid via lldb-commits
<lldb-commits at lists.llvm.org> wrote:
> Modified: lldb/trunk/packages/Python/lldbsuite/test/expression_command/anonymous-struct/TestCallUserAnonTypedef.py
> URL: http://llvm.org/viewvc/llvm-project/lldb/trunk/packages/Python/lldbsuite/test/expression_command/anonymous-struct/TestCallUserAnonTypedef.py?rev=270745&r1=270744&r2=270745&view=diff
> ==============================================================================
> --- lldb/trunk/packages/Python/lldbsuite/test/expression_command/anonymous-struct/TestCallUserAnonTypedef.py (original)
> +++ lldb/trunk/packages/Python/lldbsuite/test/expression_command/anonymous-struct/TestCallUserAnonTypedef.py Wed May 25 13:48:39 2016
> @@ -24,6 +24,7 @@ class TestExprLookupAnonStructTypedef(Te
>          self.line = line_number('main.cpp', '// lldb testsuite break')
>
>      @expectedFailureAll(oslist=["windows"])
> +    @expectedFailureAll(oslist=['linux'], archs=['arm'], bugnumber="llvm.org/pr27868")

These tests pass on soft float targets (e.g. android), so we should
not disable it there. I think we should figure out a way to make it
possible to disambiguate these. Right now it is possible to match
based on the triple of the target using a regular expression, but
maybe we could make that a bit easier. How about introducing a
"environment" variable, so that you could specify 'environment =
"eabihf"' or something like that?

> --- lldb/trunk/packages/Python/lldbsuite/test/functionalities/archives/TestBSDArchives.py (original)
> +++ lldb/trunk/packages/Python/lldbsuite/test/functionalities/archives/TestBSDArchives.py Wed May 25 13:48:39 2016
> @@ -21,6 +21,7 @@ class BSDArchivesTestCase(TestBase):
>          self.line = line_number('a.c', '// Set file and line breakpoint inside a().')
>
>      @expectedFailureAll(oslist=["windows"], bugnumber="llvm.org/pr24527.  Makefile.rules doesn't know how to build static libs on Windows")
> +    @expectedFailureAll(oslist=["linux"], archs=["arm"], bugnumber="llvm.org/pr27795")

This test was passing in all configurations we are testing. The fact
that you are getting a unicode error here tells me that the problem is
probably specific to your setup (different locale or something). I
don't think we should be disabling tests on all arm builds because of
that. The problem is probably not that hard to fix, and it would be
extremely valuable to weed out system dependencies like this in order
to get more reproducible test results. Can you look into this ASAP?

> Modified: lldb/trunk/packages/Python/lldbsuite/test/linux/builtin_trap/TestBuiltinTrap.py
> URL: http://llvm.org/viewvc/llvm-project/lldb/trunk/packages/Python/lldbsuite/test/linux/builtin_trap/TestBuiltinTrap.py?rev=270745&r1=270744&r2=270745&view=diff
> ==============================================================================
> --- lldb/trunk/packages/Python/lldbsuite/test/linux/builtin_trap/TestBuiltinTrap.py (original)
> +++ lldb/trunk/packages/Python/lldbsuite/test/linux/builtin_trap/TestBuiltinTrap.py Wed May 25 13:48:39 2016
> @@ -25,6 +25,7 @@ class BuiltinTrapTestCase(TestBase):
>
>      @expectedFailureAll("llvm.org/pr15936", compiler="gcc", compiler_version=["<=","4.6"])
>      @expectedFailureAll(archs="arm", compiler="gcc", triple=".*-android") # gcc generates incorrect linetable
> +    @expectedFailureAll(oslist=['linux'], archs=['arm'])
>      @skipIfWindows
>      def test_with_run_command(self):
>          """Test that LLDB handles a function with __builtin_trap correctly."""
>

You're saying that the problem is due to gcc linetables. Then please
specify compiler="gcc" here (if you know an approximate version range,
even better).

cheers,
pl