import tempfile
import os
from os import path
import sys
import re
import subprocess
from contextlib import contextmanager
from collections import defaultdict

import cxxfilt

from elftools.elf.elffile import ELFFile
from elftools.elf.descriptions import describe_symbol_type
import libarchive

@contextmanager
def chdir(newdir):
    old_cwd = os.getcwd()
    try:
        os.chdir(newdir)
        yield
    finally:
        os.chdir(old_cwd)


def trace_source_files(linker, cmdline, trace_sections=[]):
    with tempfile.TemporaryDirectory() as tempdir:
        out_path = path.join(tempdir, 'output.elf')
        output = subprocess.check_output([linker, '-o', out_path, f'-Wl,--cref', *cmdline])
        lines = [ line.strip() for line in output.decode().splitlines() ]
        # FIXME also find isr vector table references

        defs = {}
        for line in lines[lines.index('Cross Reference Table')+3:]:
            try:
                *left, right = line.split()
                if left:
                    defs[' '.join(left)] = right 
            except:
                pass

        refs = defaultdict(lambda: set())
        syms = {}
        for sym, obj in defs.items():
            fn, _, member = re.match('^([^()]+)(\((.+)\))?$', obj).groups()
            fn = path.abspath(fn)

            if member:
                subprocess.check_call(['ar', 'x', '--output', tempdir, fn, member])
                fn = path.join(tempdir, member)

            with open(fn, 'rb') as f:
                elf = ELFFile(f)

                symtab = elf.get_section_by_name('.symtab')
                
                symtab_demangled = { cxxfilt.demangle(nsym.name).replace(' ', ''): i
                        for i, nsym in enumerate(symtab.iter_symbols()) }

                def lookup_size(name):
                    name_normalized = name.replace(' ', '')
                    if name_normalized in symtab_demangled:
                        entry = symtab.get_symbol(symtab_demangled[name_normalized])
                        return entry['st_size']
                    else:
                        return None

                syms[sym] = fn, lookup_size(sym)

                s = set()
                sec_map = { sec.name: i for i, sec in enumerate(elf.iter_sections()) }
                sec_name = f'.rel.text.{sym}'
                matches = [ i for name, i in sec_map.items() if re.match(f'\.rel\..*\.{sym}', name) ]
                if matches:
                    sec = elf.get_section(matches[0])
                    for reloc in sec.iter_relocations():
                        refsym = symtab.get_symbol(reloc['r_info_sym'])
                        s.add(refsym.name)
                        
                        if refsym.name not in defs:
                            syms[refsym.name] = fn, lookup_size(refsym.name)
                refs[sym] = s

                for tsec in trace_sections:
                    matches = [ i for name, i in sec_map.items() if name == f'.rel{tsec}' ]
                    s = set()
                    if matches:
                        sec = elf.get_section(matches[0])
                        for reloc in sec.iter_relocations():
                            refsym = symtab.get_symbol(reloc['r_info_sym'])
                            s.add(refsym.name)
                    refs[tsec.replace('.', '_')] |= s

        syms_out = set()
        with open(out_path, 'rb') as f:
            elf = ELFFile(f)
            symtab = elf.get_section_by_name('.symtab')
            for sym in symtab.iter_symbols():
                if describe_symbol_type(sym['st_info']['type']) == 'FUNC':
                    syms_out.add(sym.name)
            #for sym in defs:
            #    entry = symtab.get_symbol_by_name(sym)
            #    if entry is None:
            #        syms[sym] = defs[sym], None
            #    else:
            #        syms[sym] = defs[sym], entry[0]['st_size']

        return syms, refs, syms_out

@contextmanager
def wrap(leader='', print=print, left='{', right='}'):
    print(leader, left)
    yield lambda *args, **kwargs: print('   ', *args, **kwargs)
    print(right)

def mangle(name):
    return re.sub('[^a-zA-Z0-9_]', '_', name)

if __name__ == '__main__':
    import argparse
    parser = argparse.ArgumentParser()
    parser.add_argument('--trace-sections', type=str, action='append', default=[])
    parser.add_argument('linker_binary')
    parser.add_argument('linker_args', nargs=argparse.REMAINDER)
    args = parser.parse_args()

    trace_sections = args.trace_sections
    trace_sections_mangled = { sec.replace('.', '_') for sec in trace_sections }
    syms, refs, syms_out = trace_source_files(args.linker_binary, args.linker_args, trace_sections)

    clusters = defaultdict(lambda: [])
    for sym, (obj, size) in syms.items():
        clusters[obj].append((sym, size))

    obj_size = defaultdict(lambda: 0)
    for name, (obj, size) in syms.items():
        if size is not None:
            obj_size[obj] += size

    with wrap('digraph G', print) as lvl1print:
        print('rankdir=LR;')
        print()

        for i, (obj, syms) in enumerate(clusters.items()):
            with wrap(f'subgraph cluster_{i}', lvl1print) as lvl2print:
                lvl2print(f'label = "{obj} <{obj_size[obj]}>";')
                lvl2print()
                for sym, size in syms:
                    if sym in syms_out:
                        lvl2print(f'{mangle(sym)}[label = "{sym} <{size}>"];')
            lvl1print()

        for start, ends in refs.items():
            for end in ends:
                if end and (start in syms_out or start in trace_sections_mangled) and end in syms_out:
                    lvl1print(f'{mangle(start)} -> {mangle(end)};')

        for sec in trace_sections:
            lvl1print(f'{sec.replace(".", "_")} [label = "section {sec}"];')