LD_DEBUG=all to graphviz visualizer

parse.py

#!/usr/bin/env python3

'''
Usage:

> LD_DEBUG=all LD_DEBUG_OUTPUT=logname ${COMMAND} ${ARGS}...
> cat logname.${PID} | parse.py | dot -Tsvg > output.svg

Or copy output of script to online visualizer, like this https://dreampuf.github.io/GraphvizOnline
(dot is from graphviz package)

Legend:
Arrows goes from from executable/library to it dependency;
If we looked for some symbols from library it will have such fields in description:
 * symbols:    how much symbols was exported from this library
 * syms_miss:  how many attempts was done to find this symbol in other libraries (too high? reorder libraries)
 * file_miss:  hom many path was checked to find a library location (too high? clean LD_LIBRARY_PATH)
'''

import sys

def clearName(name):
    return name.split('/')[-1].split('.')[0]

def assignSplit(exp):
    return exp.split('=')[1]

def wrapName(name):
    return '"' + name + '"'

symbolMissCounter = {}
libraryHitCounter = {}

print("digraph Libraries {")

for line in sys.stdin:
    vals = line.split()
    length = len(vals)

    if length == 3:
        # 55046: trying file=/usr/lib/x86_64-linux-gnu/libtinyxml.so.2.6.2
        if vals[1] == "trying":
            lib = clearName(assignSplit(vals[2]))
            rec = libraryHitCounter.get(lib, {"syms_found":0, "syms_miss":0, "library_path_miss":0})
            rec["library_path_miss"] = rec["library_path_miss"] + 1
            libraryHitCounter[lib] = rec

    if length == 6:
        # 55046: symbol=stdup;  lookup in file=/usr/libcurl.so.4 [0]
        if vals[2] == "lookup":
            sym = assignSplit(vals[1])[0:-1]    # remove `;` symbol at the end
            symbolMissCounter[sym] = symbolMissCounter.get(sym, 0) + 1

    if length == 7 or length == 8:
        # 55046: file=libgz-sim7.so.7    [0]; needed by ./app_exe              [0]
        # 55046: file=/lib/libgdal.so.30 [0]; needed by /usr/libpoppler.so.118 [0] (relocation dependency)
        if vals[3] == "needed":
            want = clearName(assignSplit(vals[1]))
            requester = clearName(vals[5])
            print('{} -> {};'.format(wrapName(requester), wrapName(want)))

    if length == 11 or length == 12:
        # 55046: binding file /usr/libgeos.so.3.10.2   [0] to /usr/libgeos.so.3.10.2 [0]: normal symbol `strdup'
        # 55046: binding file /usr/libcurl-gnutls.so.4 [0] to /usr/libc.so.6         [0]: normal symbol `getenv' [GLIBC_2.2.5]
        if vals[1] == "binding":
            sym = vals[10][1:-1]    # remove quotes around symbol
            lib = clearName(vals[6])
            rec = libraryHitCounter.get(lib, {"syms_found":0, "syms_miss":0, "library_path_miss":0})
            rec["syms_found"] = rec["syms_found"] + 1
            rec["syms_miss"] = rec["syms_miss"] + symbolMissCounter.get(sym, 0)
            libraryHitCounter[lib] = rec

print()
for lib, stat in libraryHitCounter.items():
    print('{} [label="{}\\nsymbols={}\\nsyms_miss={}\\nfile_miss={}"];'.format(wrapName(lib), lib, stat["syms_found"], stat["syms_miss"], stat["library_path_miss"]))

print("}")