[PATCH] Fix error in tablegen when second operand of an !if is an empty list.

[Matt Arsenault]

On Feb 16, 2014, at 3:37 PM, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:

> Testcase?

This other patch doesn’t work without this:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20140203/204278.html


> 
> On 2014 Feb 7, at 15:00, Matt Arsenault <Matthew.Arsenault at amd.com> wrote:
> 
>> !if([Something], []) would error with "No type for list".
>> 
>> 
>> http://llvm-reviews.chandlerc.com/D2724
>> 
>> Files:
>> lib/TableGen/TGParser.cpp
>> 
>> Index: lib/TableGen/TGParser.cpp
>> ===================================================================
>> --- lib/TableGen/TGParser.cpp
>> +++ lib/TableGen/TGParser.cpp
>> @@ -1028,7 +1028,13 @@
>>    }
>>    Lex.Lex();  // eat the ','
>> 
>> -    Init *RHS = ParseValue(CurRec);
>> +    // Get the type in case the second operand might be an empty list. We can
>> +    // infer the next item's type is the same.
>> +    RecTy *MHSTy = 0;
>> +    if (TypedInit *MHSt = dyn_cast<TypedInit>(MHS))
>> +      MHSTy = MHSt->getType();
>> +
>> +    Init *RHS = ParseValue(CurRec, MHSTy);
>>    if (RHS == 0) return 0;
>> 
>>    if (Lex.getCode() != tgtok::r_paren) {
>> <D2724.1.patch>_______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> 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