[PATCH] Do not split variables definitions into 2 DIEs.

[Frédéric Riss]

> On Oct 9, 2014, at 5:06 PM, Eric Christopher <echristo at gmail.com> wrote:
> 
> On Thu, Oct 9, 2014 at 4:19 PM, Frederic Riss <friss at apple.com> wrote:
>> I finally managed to get some test results for this patch. Indeed putting the static member variable definition broke some lldb tests, thus I left the behavior unchanged for static member variables. This is still a nice cleanup of the global variable DIE creation logic.
> 
> Curious what broke?

The resolution of static member variable (heh, who would have thought !). I haven’t looked into the lldb code yet, but I suppose it has something to do with the fact that the declarations are TAG_member and the definitions TAG_variable. With the old patch, the definitions became TAG_members. It might just be missing a case of handling TAG_members as potential variable definitions.

Fred

> -eric
> 
>> 
>> David, if you still like the direction I'd appreciate if you could give it a test run on the GDB testsuite to see if it doesn't break anything there. Note that I haven't included an additional test for type units as I removed the behavior change that might have caused issues with that usecase.
>> 
>> http://reviews.llvm.org/D5457
>> 
>> Files:
>>  lib/CodeGen/AsmPrinter/DwarfUnit.cpp
>>  test/DebugInfo/X86/gnu-public-names.ll