[lld] r194624 - Remove default label from fully covered switch.

[Rui Ueyama]

On Wed, Nov 13, 2013 at 10:31 PM, Duncan P. N. Exon Smith <
dexonsmith at apple.com> wrote:

> On 2013 Nov 13, at 21:41, Rui Ueyama <ruiu at google.com> wrote:
>
> > On Wed, Nov 13, 2013 at 7:21 PM, Duncan Exon Smith <dexonsmith at apple.com>
> wrote:
> > On Nov 13, 2013, at 3:33 PM, Rui Ueyama <ruiu at google.com> wrote:
> >
> > >   case NCR_Error:
> > >     llvm::errs() << "SymbolTable: error while merging " << name <<
> "\n";
> > > -    // FALLTHRU
> > > -  default:
> > > -    llvm::report_fatal_error("SymbolTable::addByName(): unhandled
> switch clause");
> > >   }
> >
> > Was the change in behavior intended, or should you add a new
> report_fatal_error() call?
> >
> > This change is kind of intended. If something wrong in the input, we
> sometimes terminate the linker immediately, and sometimes continue
> processing after printing an error with llvm::err(). Previously the linker
> is terminated with pretty vague error message ("unhandled switch clause").
> With this patch it proceed after printing an error. I was OK in either way,
> so I just removed default label. I should have made this point clear in the
> commit log. Sorry about that.
>
> NCR_Error specifically indicates that one of the symbols is absolute;
> interesting that this isn’t worth terminating over.  I see that some
> duplicate definitions result in termination.  However, as long as new
> behavior is intentional, you may want to change the llvm::errs printout to
> “lld warning: ...” and be more explicit than “error while merging”.
>

Yeah, on the second thought, terminating the program would make more sense
than continuing the process as you said. Thank you for pointing out. I will
address that in a followup patch.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20131113/caacdcfe/attachment.html>