Patch: Improve demangling w.r.t. possible underscore prefix

[Andrew Haley]

Ranjit Mathew writes:
 > >It seems like it will break if ever __USER_LABEL_PREFIX__is not '_'.
 > 
 > I'm guessing that you mean "...if ever __USER_LABEL_PREFIX__ is not '_'
 > [and is not empty either]". Well, I checked for all instances of
 > USER_LABEL_PREFIX in all files in "gcc/config" and its subfolders
 > and every target either defines it to an underscore or an empty
 > string. Besides, c++filt does not handle anything more complicated
 > than an underscore. That way the patch looks fine to me, unless I've
 > misunderstood your question completely.
 > 
 > 
 > >Why can't you just cat __USER_LABEL_PREFIX__ to the start of the name?
 > >Surround the whole thing by #ifdef __USER_LABEL_PREFIX__.
 > 
 > By "cat" if you mean the "##" preprocessor operator, it does not work.

No, I mean that you can stringify __USER_LABEL_PREFIX__ by using the #
token.

So, 

#define __USER_LABEL_PREFIX__ _
#define USER_LABEL_PREFIX_STRING0(s) #s
#define USER_LABEL_PREFIX_STRING(s) USER_LABEL_PREFIX_STRING0(s)
#define PREFIX USER_LABEL_PREFIX_STRING(__USER_LABEL_PREFIX__)

java::lang::String*
gnu::gcj::runtime::NameFinder::getExternalLabel (java::lang::String* name)
{
  jsize nameLen = JvGetStringUTFLength (name);
  char *newName = (char *) JvMalloc (strlen (PREFIX) + nameLen + 1);
  strcpy (newName, PREFIX);
  JvGetStringUTFRegion (name, 0, nameLen, newName + strlen (newName));
  *(newName + strlen (PREFIX) + nameLen) = '\0';
  return JvNewStringLatin1 (newName);
}

I haven't tested this, but you get the idea.  It doesn't use gotos or
onstruct special symbols and it's won't break if ever
__USER_LABEL_PREFIX__ is "$" or somesuch.

Andrew.