[llvm] r186634 - Debug Info: enable verifying by default and disable testing cases that fail.

[manman ren]

Hi Duncan,

That is exactly what happened, DIGlobalVariable::Verify enforces a display
name, which seems unnecessary.
Eric, any opinion on this?

Thanks,
Manman


On Fri, Jul 19, 2013 at 1:21 AM, Duncan Sands <baldrick at free.fr> wrote:

> Hi Manman,
>
> On 19/07/13 07:07, Manman Ren wrote:
>
>>
>> Hi Duncan,
>>
>> How can I debug the failures?
>>
>
> this is coming from Fortran that likes to create anonymous global constants
> like this:
>
>   @0 = internal constant i32 77
>
> I think it's so that it can pass a pointer to "77" around.  Anyway, the
> debug
> code uses the name of the global as the "display name" but here there is
> none.
> The verifier then barfs because it thinks everything should have a name.
>
> I'm not sure why LLVM has this obsession that everything should have a
> name.
> Is there a real reason for that?  If dwarf requires a name, then in this
> kind
> of case someone will have to invent one.  The question then is whether it
> is
> better to require front-ends to always supply a name (meaning the
> front-end has
> to invent one), or to accept empty names and have the backend make
> something up
> when outputting dwarf.  Any opinions?
>
> Ciao, Duncan.
>
>
>  Also getDICompositeType is not used on mainline but is used on dragonegg,
>> how
>> can I figure out the usage?
>> Do  I have access to the source?
>>
>> Thanks,
>> Manman
>>
>> On Jul 18, 2013, at 8:56 PM, Eric Christopher wrote:
>>
>>  Also this appears to have broken the dragonegg bots. :)
>>>
>>> -eric
>>>
>>> On Thu, Jul 18, 2013 at 6:08 PM, Eric Christopher <echristo at gmail.com
>>> <mailto:echristo at gmail.com>> wrote:
>>>
>>>> Debug Info: enable verifying by default and disable testing cases that
>>>>> fail.
>>>>>
>>>>> 1> Use DebugInfoFinder to find debug info MDNodes.
>>>>> 2> Add disable-debug-info-verifier to disable verifying debug info.
>>>>> 3> Disable verifying for testing cases that fail (will update the
>>>>> testing cases
>>>>>   later on).
>>>>> 4> MDNodes generated by clang can have empty filename for
>>>>> TAG_inheritance and
>>>>>   TAG_friend, so DIType::Verify is modified accordingly.
>>>>>
>>>>>
>>>> Cool, looks about like what I'd expected from when we discussed it.
>>>> I'd have preferred you split out the change to verify :)
>>>>
>>>>  Note that DebugInfoFinder does not list all debug info MDNode.
>>>>>
>>>>
>>>> Seems like we should fix this? :)
>>>>
>>>>  For example, clang can generate:
>>>>> metadata !{i32 786468}, which will fail to verify.
>>>>>
>>>>
>>>> This doesn't seem related to "not listing" all of the debug info, but
>>>> a problem because of:
>>>>
>>>>  This MDNode is used by debug info but not included in DebugInfoFinder.
>>>>> This MDNode is generated as a temporary node in
>>>>> DIBuilder::createFunction
>>>>>  Value *TElts[] = { GetTagConstant(VMContext, DW_TAG_base_type) };
>>>>>  MDNode::getTemporary(**VMContext, TElts)
>>>>>
>>>>>
>>>> This is the "we need a temporary placeholder" problem basically.
>>>> You'll see it in GVs, subprograms etc when we construct the CU as
>>>> well. Theoretically the verifier should catch these as "not
>>>> important". What's actually going on here that's causing it to fail?
>>>>
>>>> -eric
>>>> ______________________________**_________________
>>>> llvm-commits mailing list
>>>> llvm-commits at cs.uiuc.edu <mailto:llvm-commits at cs.uiuc.**edu<llvm-commits at cs.uiuc.edu>
>>>> >
>>>> http://lists.cs.uiuc.edu/**mailman/listinfo/llvm-commits<http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits>
>>>>
>>>
>>
> ______________________________**_________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/**mailman/listinfo/llvm-commits<http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130719/48aaf681/attachment.html>