[lldb-dev] lldb fails to build under GNU/Linux mipsel

[Greg Clayton]

llvm::APFloat used to not support being able to give out a human readable string, it used to only emit the compiler hex float format. I believe this has changed and that llvm::APFloat now does support a "to string" format that is human readable. 

If this is the case, the right fix is to get rid of the code that is using native float/double/long double and use APFloat directly (always getting the current definition of APFloat from the appropriate AST.

Greg

On May 16, 2013, at 3:16 AM, Sylvestre Ledru <sylvestre at debian.org> wrote:

> Hello,
> 
> tools/lldb/source/Plugins/Process/Utility/RegisterContextMacOSXFrameBackchain.cpp:
> In member function 'virtual bool
> RegisterContextMacOSXFrameBackchain::ReadRegister(const
> lldb_private::RegisterInfo*, lldb_private::RegisterValue&)':
> tools/lldb/source/Plugins/Process/Utility/RegisterContextMacOSXFrameBackchain.cpp:153:9:
> error: duplicate case value
> tools/lldb/source/Plugins/Process/Utility/RegisterContextMacOSXFrameBackchain.cpp:137:9:
> error: previously used here
> 
> Line 137 :
>        case sizeof (double):
> Line 153:
> // TOOD: need a better way to detect when "long double" types are
> // the same bytes size as "double"
> #if !defined(__arm__)
>        case sizeof (long double):
> 
> I could commit the attached patch but I wonder if there is not a better
> solution.
> 
> Thanks
> Sylvestre
> 
> <disable_for_mipsel_too.diff>_______________________________________________
> lldb-dev mailing list
> lldb-dev at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev