summaryrefslogtreecommitdiff
path: root/ci/fuzzing/README.txt
blob: 5ee4b8d0d0ea0ebc5c720a3e0abd3a11df3b11ad (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
Fuzzing GStreamer
=================

  This directory contains the various fuzzing targets and helper
  scripts.

* Fuzzing targets

  Fuzzing targets as small applications where we can test a specific
  element or API. The goal is to have them be as small/targetted as
  possible.

    ex: appsrc ! <some_element> ! fakesink num-buffers=<small>
    
  Not all components can be tested directly and therefore will be
  indirectly tested via other targets (ex: libgstaudio will be tested
  by targets/elements requiring it)

  Anything that can process externally-provided data should be
  covered, but there are cases where it might not make sense to use a
  fuzzer (such as most elements processing raw audio/video).

* build-oss-fuzz.sh

  This is the script executed by the oss-fuzz project.

  It builds glib, GStreamer, plugins and the fuzzing targets.

* *.c

  The fuzzing targets where the data to test will be provided to a
  function whose signature follows the LibFuzzer signature:
  https://llvm.org/docs/LibFuzzer.html

* *.corpus

  A file matching a test name that contains a list of files to use when
  starting a fuzzing run.  Providing an initial set files can speed up
  the fuzzing process significantly.

* TODO

  * Add a standalone build script

    We need to be able to build and test the fuzzing targets outside
    of the oss-fuzz infrastructure, and do that in our continous
    integration system.

    We need:

    * A dummy fuzzing engine (given a directory, it opens all files and
      calls the fuzzing targets with the content of those files.
    * A script to be able to build those targets with that dummy engine
    * A corpus of files to test those targets with.

  * Build targets with dummy engine and run with existing tests.

  * Create pull-based variants

    Currently the existing targets are push-based only. Where
    applicable we should make pull-based variants to test the other
    code paths.

  * Add more targets

    core:
      gst_parse fuzzer ?
    base:
      ext/
        ogg
	opus
	pango
	theora
	vorbis
      gst/
        subparse
	typefind : already covered in typefind target
      gst-libs/gst/
        sdp
	other ones easily testable directly ?