This repository has been archived by the owner on Jul 21, 2020. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 13
/
README
174 lines (133 loc) · 8.64 KB
/
README
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
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
What is OBS?
This project is a rewrite of what was formerly known as "Open Broadcaster
Software", software originally designed for recording and streaming live
video content, efficiently.
Bug Tracker: https://obsproject.com/mantis/
We are no longer using GitHub issues! Please use Mantis, and only report
bugs and major issues. Do NOT use mantis to ask questions or request
features, please keep that to the forums.
Forum accounts are now linked to Mantis Bug Tracker. To use the bug
tracker, simply log in to the forums and then go to the bug tracker link
above.
What's the goal of rewriting OBS?
- Make it multiplatform. Use multiplatform libraries/functions/classes where
possible to allow this. Multi-platform support was one of the primary
reasons for the rewrite. This also means using a UI toolkit will be
necessary for user interface. It also means allowing the use of OpenGL as
well as Direct3D.
- Separate the application from the core, allowing custom application of
the core if desired, and easier extending of the user interface.
- Simplify complex systems to not only make it easier to use, but easier to
maintain.
- Write a better core API, and design the entire system to be modular.
- Now that we have much more experience, improve the overall design of all
the subsystems/API, and minimize/eliminate design flaws. Make it so we can
do all the things we've had trouble with before, such as custom outputs,
multiple outputs at once, better handling of inputs, custom services.
- Make a better/cleaner code base, use better coding standards, use standard
libraries where possible (not just STL and C standard library, but also
things like ffmpeg as well), and improve maintainability of the project as a
whole.
- Implement a new API-independent shader/effect system allowing better and
easier shaders usage and customization without having to duplicate shader
code.
- Better device support. Again, I didn't know what I was getting into when
I originally started writing code for devices. It evolved into a totally
convoluted mess. I would have improved the existing device plugin code, but
it was just all so fundamentally bad and flawed that it would have been
detrimental to progression to continue working on it rather than rewrite it.
What was wrong with the original OBS?
The original OBS was rewritten not because it was bad, at least in terms of
optimization. Optimization and graphics are things I love. However, there
were some serious problems with the code and design that were deep and
fundamental, which prevented myself and other developers from being able to
improve/extend the application or add new features very easily.
First, the design flaws:
- The original OBS was completely and hopelessly hard-coded for windows,
and only windows. It was just totally impossible to use it on other
systems.
- All the sub-systems were written before I really knew what I was getting
into. When I started the project, I didn't really fully comprehend the
scope of what I would need or how to properly design the project. My
design and plans for the application were just to write something that
would "stream games and a webcam, with things like overlays and such."
This turned out fine for most casual gamers and streamers (and very
successful), but left anyone wanting to do anything more advanced left
massively wanting.
- Subsystems and core functionalities intermingled in such a way that it
was a nightmare to get proper custom functionality out of it. Things
like QSV had to be meshed in with the main encoding loop, and it just
made things a nightmare to deal with. Custom outputs were nigh
impossible.
- The API was poorly designed because most of it came after I originally
wrote the application, it was more of an afterthought, and plugin API
would routinely break for plugin developers due to changing C++
interfaces (one of the reasons the core is now C).
- API was intermeshed with the main executable. The OBSApi DLL was
nothing more than basically this mutant growth upon OBS.exe that allowed
plugin developers to barely write plugins, but all the important API
code was actually stored in the executable. Navigation was a total mess.
- The graphics subsystem, while not bad, was incomplete, and though far
easier to use than bare D3D, wasn't ideal, and was hard-coded for D3D
specifically.
- The devices and audio code was poor, I had no idea what I was getting into
when I started writing them in. I did not realize beforehand all the
device-specific quirks that each device/system could have. Some devices
had bad timing and quirks that I never aniticipated while writing them.
I struggled with devices, and my original design for the audio subsystem
for example morphed over and over into an abomination that, though works,
is basically this giant duct-taped zombie monster.
- Shaders were difficult to customize because they had to be duplicated if
you wanted slightly different functionality that required more than just
changing shader constants.
- Orientation of sources was fixed, and required special code for each
source to do any custom modification of rotation/position/scale/etc.
This is one of those fundamental flaws that I look back on and regret, as
it was a stupid idea from the beginning. I originally thought I could
get more accurate source position/sizes, but it just turned out to be
totally bad. Should have been matrices from the beginning just like with
a regular 3D engine.
Second, the coding flaws:
- The coding style was inconsistent.
- C++98, C-Style C++, there was no exception usage, no STL. C++ used
poorly.
- Not Invented Here Syndrome everywhere. Custom string functions/classes,
custom templates, custom everything everywhere. To be fair, it was all
hand-me-down code from the early 2000s that I had become used to, but
that was no excuse -- C-standard libraries and the STL should have been
used from the beginning over anything else. That doesn't mean to say
that using custom stuff is always bad, but doing it to the extent I did
definitely was. Made it horrible to maintain as well, required extra
knowledge for plugin developers and anyone messing with the code.
- Giant monolithic classes everywhere, the main OBS class was paricularly
bad in this regard. This meant navigation was a nightmare, and no one
really knew where to go or where to add/change things.
- Giant monolithic functions everywhere. This was particularly bad
because it meant that functions became harder to debug and harder to
keep track of what was going on in any particular function at any given
time. These large functions, though not inefficient, were delicate and
easily breakable. (See OBS::MainCaptureLoop for a nightmarish example,
or the listbox subclass window procedure in WindowStuff.cpp)
- Very large file sizes with everything clumped up into single files (for
another particularly nightmarish example, see WindowStuff.cpp)
- Bad formatting. Code could go beyond 200 columns in some cases, making
it very unpleasant to read with many editors. Spaces instead of tabs,
K&R mixed with allman (which was admittedly my fault).
New (actual) coding guidelines
- For the C code (especially in the core), guidelines are pretty strict K&R,
kernel style. See the linux kernel "CodingStyle" document for more
information. That particular coding style guideline is for more than just
style, it actually helps produce a better overall code base.
- For C++ code, I still use CamelCase instead of all_lowercase just because
I prefer it that way, it feels right with C++ for some reason. It also
helps make it distinguishable from C code.
- I've started using 8-column tabs for almost everything -- I really
personally like it over 4-column tabs. I feel that 8-column tabs are very
helpful in preventing large amounts of indentation. A self-imposed
limitation, if you will. I also use actual tabs now, instead of spaces.
Also, I feel that the K&R style looks much better/cleaner when viewed with
8-column tabs.
- Preferred maximum columns: 80. I've also been doing this because in
combination with 8-column tabs, it further prevents large/bad functions
with high indentation. Another self-imposed limitation. Also, it makes
for much cleaner viewing in certain editors that wrap (like vim).