[llvm] r279966 - Mark test as XFAIL instead of disabling it everywhere.

Teresa Johnson via llvm-commits llvm-commits at lists.llvm.org
Mon Aug 29 10:14:30 PDT 2016


On Mon, Aug 29, 2016 at 10:11 AM, Mehdi Amini <mehdi.amini at apple.com> wrote:
>
> On Aug 29, 2016, at 10:07 AM, Teresa Johnson <tejohnson at google.com> wrote:
>
> On Mon, Aug 29, 2016 at 9:57 AM, Mehdi Amini <mehdi.amini at apple.com> wrote:
>
>
> On Aug 29, 2016, at 6:14 AM, Teresa Johnson <tejohnson at google.com> wrote:
>
> Hi Mehdi,
>
> Is it necessary for this test to only run on X86? If so, we should
> create an X86 subdirectory in test/tools/llvm-lto2 and test
> config.root.targets in
> the lit as done elsewhere for X86-specific tests.
>
> Looks like my recent common fix broke this test (but it was hidden
> since it wasn't actually running). Will take a look at fixing that
> part, but I'm not sure if it is safe to enable everywhere given the
> original attempt to limit it.
>
>
> The issue is that a triple/datalayout is needed for the mangling to be
> correct during LTO (symbols are prefixed with _ on Darwin).
> Not adding a triple/datalayout makes the tests behave differently depending
> on the default triple llvm is built with (usually the host).
>
> Adding only a data layout and no triple may be enough for the mangling, but
> the CodeGen will assert when the data layout does not match the
> TargetMachine, so I’m not sure it’ll work.
>
>
> Is it not enough to add both an x86 target triple and datalayout to the
> test?
>
>
> Which is what I did, but then you can only run when the X86 backend is
> configured, which I tried (and failed) to achieve with the REQUIRES line.

Oh ok (there isn't a target datalayout there, which is why I asked).

>
> You suggested the right thing earlier (adding a X86 subdirectory as usual) I
> think.

Ok, will do. I have the fixes ready to the non-xfailed version, so will do both.

Teresa

>
>> Mehdi
>
>
>
> Teresa
>
>
>> Mehdi
>
>
>
> Thanks,
> Teresa
>
> On Mon, Aug 29, 2016 at 5:41 AM, Benjamin Kramer via llvm-commits
> <llvm-commits at lists.llvm.org> wrote:
>
> Author: d0k
> Date: Mon Aug 29 07:41:32 2016
> New Revision: 279966
>
> URL: http://llvm.org/viewvc/llvm-project?rev=279966&view=rev
> Log:
> Mark test as XFAIL instead of disabling it everywhere.
>
> There is no lit feature 'X86' so this test is just disabled completely.
> Make it XFAIL until a solution is found.
>
> Modified:
>   llvm/trunk/test/tools/llvm-lto2/common.ll
>
> Modified: llvm/trunk/test/tools/llvm-lto2/common.ll
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/tools/llvm-lto2/common.ll?rev=279966&r1=279965&r2=279966&view=diff
> ==============================================================================
> --- llvm/trunk/test/tools/llvm-lto2/common.ll (original)
> +++ llvm/trunk/test/tools/llvm-lto2/common.ll Mon Aug 29 07:41:32 2016
> @@ -1,4 +1,4 @@
> -; REQUIRES: X86
> +; XFAIL: *
> ; RUN: llvm-as < %s > %t1.bc
> ; RUN: llvm-as < %p/Inputs/common.ll > %t2.bc
>
> @@ -63,4 +63,4 @@ target triple = "x86_64-apple-macosx10.1
>
> define i8 *@foo() {
> ret i8 *@v
> -}
> \ No newline at end of file
> +}
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>
>
>
>
> --
> Teresa Johnson | Software Engineer | tejohnson at google.com | 408-460-2413
>
>
>
>
>
> --
> Teresa Johnson | Software Engineer | tejohnson at google.com | 408-460-2413
>
>



-- 
Teresa Johnson | Software Engineer | tejohnson at google.com | 408-460-2413


More information about the llvm-commits mailing list