[llvm-testresults] buildbot failure in llvm on llvm-gcc-x86_64-linux-selfhost

[Jim Grosbach]

Evan, this looks like it's from 118135. There's a missing vpush instruction in the prologue on Linux for the generated testcase.

Previously, we would generate:
        push    {r4, r5, r6, r7, r8, r9, r10, lr}
        vpush   {d8, d9, d10}
        sub     sp, sp, #64
Now we generate:
	push	{r4, r5, r6, r7, r8, r9, r10, r11, lr}
	sub	sp, sp, #64

That said, it looks like the register allocation for the floating point values changed, so the prologue change is likely just a side effect of that, but I don't know if your change was supposed to have any effect on allocation order(?) The Darwin generated code also stopped having the vpush for the same reason, but our revert of the prologue optimization masked the change there, as we still have two pushes without an intervening stack adjustment, which is what the test is looking for:
        push    {r4, r5, r6, r7, lr}
        add     r7, sp, #12
        push    {r8, r10, r11}


It may be the test just needs updated. Can you have a look?

-Jim


On Nov 2, 2010, at 8:39 PM, buildbot at google1.osuosl.org wrote:

> The Buildbot has detected a new failure of llvm-gcc-x86_64-linux-selfhost on llvm.
> Full details are available at:
> http://google1.osuosl.org:8011/builders/llvm-gcc-x86_64-linux-selfhost/builds/463
> 
> Buildbot URL: http://google1.osuosl.org:8011/
> 
> Buildslave for this Build: osu7
> 
> Build Reason: 
> Build Source Stamp: 118137
> Blamelist: dgregor,djg,echristo,evancheng,grosbach,jtsoftware,kremenek,lattner,resistor,stoklund,void
> 
> BUILD FAILED: failed test.llvm.stage1 test.llvm.stage2
> 
> sincerely,
> -The Buildbot
>