[llvm-commits] [llvm] r105528 - in /llvm/trunk: lib/Transforms/IPO/PartialSpecialization.cpp test/Transforms/PartialSpecialize/ test/Transforms/PartialSpecialize/dg.exp test/Transforms/PartialSpecialize/two-specializations.ll

[Daniel Dunbar]

On Fri, Jun 25, 2010 at 8:06 AM, Kenneth Uildriks <kennethuil at gmail.com> wrote:
> On Wed, Jun 23, 2010 at 1:03 PM, Daniel Dunbar <daniel at zuster.org> wrote:
>> Hi Kenneth,
>>
>> This test is failing on llvm-arm-linux. It looks to me like the order
>> of the partial specialization numbering isn't deterministic?
>>
>> Failure is here:
>>  http://google1.osuosl.org:8011/builders/llvm-arm-linux/builds/3516/steps/test-llvm/logs/two-specializations.ll
>>
>> Can you take a look?
>
> Is there a way to see what the opt output was on the buildbot?

No, but if you look & guess enough you can see that what is happening
is that the specialized functions are getting named in the opposite
order. If you swap UseCallback1 and UseCallback2 on the output on x86,
you will see the same failure:
--
ddunbar at lordcrumb:PartialSpecialize (master)$ opt -S
-partialspecialization two-specializations.ll | FileCheck
two-specializations.ll
ddunbar at lordcrumb:PartialSpecialize (master)$ opt -S
-partialspecialization two-specializations.ll | sed -e
"s#UseCallback1#UseCallbackTmp#" -e "s#UseCallback2#UseCallback1#" -e
"s#UseCallbackTmp#UseCallback2#" | FileCheck two-specializations.ll
two-specializations.ll:17:15: error: CHECK-NEXT: is not on the line
after the previous match
; CHECK-NEXT: call void @UseCallback1()
              ^
<stdin>:16:2: note: 'next' match was here
 call void @UseCallback1()
 ^
<stdin>:13:6: note: previous match was here
Entry:
     ^
ddunbar at lordcrumb:PartialSpecialize (master)$
--

 - Daniel

> Does llvm work on an ARM-QEMU?
>