the object header in the dataset structure). Also deleted the support for a
"per file" error stack, since we've decided to go with a "per thread" model
(currently).
name, they operate to query or modify the corresponding symbol table entry.
They don't understand directory hierarchies (that's the next layer up).
Since the object header stuff isn't done yet, they call four stub functions
that all return failure:
not_implemented_yet__create_object_header()
not_implemented_yet__insert_symtab_message()
not_implemented_yet__get_symtab_message()
not_implemented_yet__update_symtab_message()
The interface is:
haddr_t H5G_new (file, initial_heap_size)
Creates a new symbol table.
haddr_t H5G_find (file, symtab_addr, symbol_name, *entry)
Returns a symbol table entry given the name.
herr_t H5G_modify (file, symtab_addr, symbol_name, *entry)
Modifies the symbol table entry for the specified name.
herr_t H5G_insert (file, symtab_addr, symbol_name, *entry)
Inserts a new name and symbol table entry pair.
intn H5G_list (file, symtab_addr, maxentries, names[], entries[])
Returns a list of all names and symbol table entries.
H5F_SIZEOF_SIZE. Modified files to use these instead of
accessing the hdf5_file_t struct directly.
Changed address return values from H5B to -1 for error.
Changed off_t to haddr_t for anything that's a file address.