Should undefined symbols in sysv hash table be ignored in do_lookup_x?

[Roland McGrath]

Is there any reason not to move the change into check_match?  Its only
other caller is the DT_GNU_HASH case, which it doesn't affect since that
table never has such symbols.  But check_match already checks for SHN_UNDEF
contingent on type_class, so the existing check would be simplified by
always catching SHN_UNDEF.