PATCH: Update libiberty/testsuite/demangle-expected

[H. J. Lu]

On Fri, Oct 17, 2003 at 12:17:23AM -0400, DJ Delorie wrote:
> 
> > I assumed the output of demanglers in libiberty and libstdc++ should
> > be the identical on the same C++ symbol. I copied that particular test
> > from libstdc++. Since the orginal test was changed, I thought I should
> > keep it in sync. I guess my assumption is wrong. What is the policy on
> > the outputs from demanglers between libiberty and libstdc++ on the same
> > C++ symbols?
> 
> My reply had nothing to do with the demangler policy, and everything
> to do with patch committal policy.  You requested approval for a
> patch, and committed it without waiting for approval.  Such sloppy
> adherence to procedures is especially bad in libiberty, which services
> many projects.
> 
> As for the patch itself, I answered that already - there are no
> obvious changes to testsuites.  Each and every change to existing test
> cases must be discussed, because they could potentially harm backwards
> compatibility.

I imported a bunch of demangler tests from libstdc++, which are failed
on the current demangler in libiberty, including the one I changed:

# c++filt _GLOBAL__I__Z2fnv
_GLOBAL__I__Z2fnv

I thought its purpose is to remind people of the problem with demangler
in libiberty and make it to generate the same output as the one in
libstdc++. If by backwards compatibility, you meant to keep the
current behavior of demangler in libiberty, should I back out those
new failed tests imported from libstdc++ so that you won't see any
failures? If it is not what you meant, should those imported new
tests be kept in sync with the original ones in libstdc++?


H.J.