[LLVMdev] DW_AT_[MIPS_]linkage_name inconsistency

Adrian Prantl aprantl at apple.com
Fri Mar 6 08:30:39 PST 2015


> On Mar 5, 2015, at 5:32 PM, Robinson, Paul <Paul_Robinson at playstation.sony.com> wrote:
> 
> Yeah. Other fun is various different vendor attributes that we probably don't want depending on debugger that's going to be used.
> 
>  
> 
> -fgdb-extensions, -flldb-extensions, -f<whatever your debugger is called>-extensions?
> 
>  
> 
> Yeah, we've been tolerating all those DW_AT_APPLE_* things for years.  Never quite annoying or space-consumingabout. J
> 
> Being able to specify debugger-specific behavior separately from target is an intriguing idea. Would sure make the testing easier even if end users don't care.
> 

Agreed! Many of the checks for the Darwin platform should really be a check for LLDB extensions, so platforms like FreeBSD that use LLDB as the system debugger can benefit from it, too.

-- adrian

> --paulr
> 
>  
> 
> From: Eric Christopher [mailto:echristo at gmail.com] 
> Sent: Thursday, March 05, 2015 4:57 PM
> To: Robinson, Paul; LLVM Developers Mailing List (llvmdev at cs.uiuc.edu)
> Subject: Re: [LLVMdev] DW_AT_[MIPS_]linkage_name inconsistency
> 
>  
> 
>  
> 
> On Thu, Mar 5, 2015 at 3:37 PM Robinson, Paul <Paul_Robinson at playstation.sony.com> wrote:
> 
> We're inconsistent about using DW_AT_MIPS_linkage_name versus
> DW_AT_linkage_name.
> 
> Variables (see DwarfCompileUnit::getOrCreateGlobalVariableDIE())
> get the standard attribute for DWARF >= 4, MIPS for older.
> 
> Subprograms (see DwarfUnit::applySubprogramDefinitionAttributes())
> always get DW_AT_MIPS_linkage_name (no version check).
> 
>  
> 
> Probably just a bug. It's definitely not intentional on my part.
> 
>  
> 
> Before I go drudging through however many tests will die if I add
> a version check... we DO want to fix this, right?
> 
>  
> 
> Yep.
> 
>  
> 
> Thanks,
> --paulr
> 
> P.S.  PS4 doesn't want linkage names at all, which will probably
> be even more painful test-wise, but it seemed like cleaning up
> this bit should come first regardless.
> 
>  
> 
> Yeah. Other fun is various different vendor attributes that we probably don't want depending on debugger that's going to be used.
> 
>  
> 
> -fgdb-extensions, -flldb-extensions, -f<whatever your debugger is called>-extensions?
> 
>  
> 
> -eric
> 
>  
> 
> 
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
> 
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev





More information about the llvm-dev mailing list