changeset 9025 d09d4b578e77
parent 9023 276802355854
child 9754 fe6acfd4652c
     1.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     1.2 +++ b/docs/	Tue Jul 29 08:04:15 2014 -0700
     1.3 @@ -0,0 +1,129 @@
     1.4 +Dynamic API
     1.5 +================================================================================
     1.6 +Originally posted by Ryan at
     1.7 +
     1.8 +Background:
     1.9 +
    1.10 +- The Steam Runtime has (at least in theory) a really kick-ass build of SDL2, 
    1.11 +  but developers are shipping their own SDL2 with individual Steam games. 
    1.12 +  These games might stop getting updates, but a newer SDL2 might be needed later. 
    1.13 +  Certainly we'll always be fixing bugs in SDL, even if a new video target isn't 
    1.14 +  ever needed, and these fixes won't make it to a game shipping its own SDL.
    1.15 +- Even if we replace the SDL2 in those games with a compatible one, that is to 
    1.16 +  say, edit a developer's Steam depot (yuck!), there are developers that are 
    1.17 +  statically linking SDL2 that we can't do this for. We can't even force the 
    1.18 +  dynamic loader to ignore their SDL2 in this case, of course.
    1.19 +- If you don't ship an SDL2 with the game in some form, people that disabled the
    1.20 +  Steam Runtime, or just tried to run the game from the command line instead of 
    1.21 +  Steam might find themselves unable to run the game, due to a missing dependency.
    1.22 +- If you want to ship on non-Steam platforms like GOG or Humble Bundle, or target
    1.23 +  generic Linux boxes that may or may not have SDL2 installed, you have to ship 
    1.24 +  the library or risk a total failure to launch. So now, you might have to have 
    1.25 +  a non-Steam build plus a Steam build (that is, one with and one without SDL2 
    1.26 +  included), which is inconvenient if you could have had one universal build 
    1.27 +  that works everywhere.
    1.28 +- We like the zlib license, but the biggest complaint from the open source 
    1.29 +  community about the license change is the static linking. The LGPL forced this 
    1.30 +  as a legal, not technical issue, but zlib doesn't care. Even those that aren't
    1.31 +  concerned about the GNU freedoms found themselves solving the same problems: 
    1.32 +  swapping in a newer SDL to an older game often times can save the day. 
    1.33 +  Static linking stops this dead.
    1.34 +
    1.35 +So here's what we did:
    1.36 +
    1.37 +SDL now has, internally, a table of function pointers. So, this is what SDL_Init
    1.38 +now looks like:
    1.39 +
    1.40 +    UInt32 SDL_Init(Uint32 flags)
    1.41 +    {
    1.42 +        return jump_table.SDL_Init(flags);
    1.43 +    }
    1.44 +
    1.45 +Except that is all done with a bunch of macro magic so we don't have to maintain
    1.46 +every one of these.
    1.47 +
    1.48 +What is jump_table.SDL_init()? Eventually, that's a function pointer of the real
    1.49 +SDL_Init() that you've been calling all this time. But at startup, it looks more 
    1.50 +like this:
    1.51 +
    1.52 +    Uint32 SDL_Init_DEFAULT(Uint32 flags)
    1.53 +    {
    1.54 +        SDL_InitDynamicAPI();
    1.55 +        return jump_table.SDL_Init(flags);
    1.56 +    }
    1.57 +
    1.58 +SDL_InitDynamicAPI() fills in jump_table with all the actual SDL function 
    1.59 +pointers, which means that this _DEFAULT function never gets called again. 
    1.60 +First call to any SDL function sets the whole thing up.
    1.61 +
    1.62 +So you might be asking, what was the value in that? Isn't this what the operating
    1.63 +system's dynamic loader was supposed to do for us? Yes, but now we've got this 
    1.64 +level of indirection, we can do things like this:
    1.65 +
    1.66 +    export SDL_DYNAMIC_API=/my/actual/
    1.67 +    ./MyGameThatIsStaticallyLinkedToSDL2
    1.68 +
    1.69 +And now, this game that is staticallly linked to SDL, can still be overridden 
    1.70 +with a newer, or better, SDL. The statically linked one will only be used as 
    1.71 +far as calling into the jump table in this case. But in cases where no override
    1.72 +is desired, the statically linked version will provide its own jump table, 
    1.73 +and everyone is happy.
    1.74 +
    1.75 +So now:
    1.76 +- Developers can statically link SDL, and users can still replace it. 
    1.77 +  (We'd still rather you ship a shared library, though!)
    1.78 +- Developers can ship an SDL with their game, Valve can override it for, say, 
    1.79 +  new features on SteamOS, or distros can override it for their own needs, 
    1.80 +  but it'll also just work in the default case.
    1.81 +- Developers can ship the same package to everyone (Humble Bundle, GOG, etc), 
    1.82 +  and it'll do the right thing.
    1.83 +- End users (and Valve) can update a game's SDL in almost any case, 
    1.84 +  to keep abandoned games running on newer platforms.
    1.85 +- Everyone develops with SDL exactly as they have been doing all along. 
    1.86 +  Same headers, same ABI. Just get the latest version to enable this magic.
    1.87 +
    1.88 +
    1.89 +A little more about SDL_InitDynamicAPI():
    1.90 +
    1.91 +Internally, InitAPI does some locking to make sure everything waits until a 
    1.92 +single thread initializes everything (although even SDL_CreateThread() goes 
    1.93 +through here before spinning a thread, too), and then decides if it should use
    1.94 +an external SDL library. If not, it sets up the jump table using the current 
    1.95 +SDL's function pointers (which might be statically linked into a program, or in
    1.96 +a shared library of its own). If so, it loads that library and looks for and 
    1.97 +calls a single function:
    1.98 +
    1.99 +    SInt32 SDL_DYNAPI_entry(Uint32 version, void *table, Uint32 tablesize);
   1.100 +
   1.101 +That function takes a version number (more on that in a moment), the address of
   1.102 +the jump table, and the size, in bytes, of the table. 
   1.103 +Now, we've got policy here: this table's layout never changes; new stuff gets 
   1.104 +added to the end. Therefore SDL_DYNAPI_entry() knows that it can provide all 
   1.105 +the needed functions if tablesize <= sizeof its own jump table. If tablesize is
   1.106 +bigger (say, SDL 2.0.4 is trying to load SDL 2.0.3), then we know to abort, but
   1.107 +if it's smaller, we know we can provide the entire API that the caller needs.
   1.108 +
   1.109 +The version variable is a failsafe switch. 
   1.110 +Right now it's always 1. This number changes when there are major API changes 
   1.111 +(so we know if the tablesize might be smaller, or entries in it have changed). 
   1.112 +Right now SDL_DYNAPI_entry gives up if the version doesn't match, but it's not 
   1.113 +inconceivable to have a small dispatch library that only supplies this one 
   1.114 +function and loads different, otherwise-incompatible SDL libraries and has the
   1.115 +right one initialize the jump table based on the version. For something that 
   1.116 +must generically catch lots of different versions of SDL over time, like the 
   1.117 +Steam Client, this isn't a bad option.
   1.118 +
   1.119 +Finally, I'm sure some people are reading this and thinking,
   1.120 +"I don't want that overhead in my project!"  
   1.121 +To which I would point out that the extra function call through the jump table 
   1.122 +probably wouldn't even show up in a profile, but lucky you: this can all be 
   1.123 +disabled. You can build SDL without this if you absolutely must, but we would 
   1.124 +encourage you not to do that. However, on heavily locked down platforms like 
   1.125 +iOS, or maybe when debugging,  it makes sense to disable it. The way this is 
   1.126 +designed in SDL, you just have to change one #define, and the entire system 
   1.127 +vaporizes out, and SDL functions exactly like it always did. Most of it is 
   1.128 +macro magic, so the system is contained to one C file and a few headers. 
   1.129 +However, this is on by default and you have to edit a header file to turn it 
   1.130 +off. Our hopes is that if we make it easy to disable, but not too easy, 
   1.131 +everyone will ultimately be able to get what they want, but we've gently 
   1.132 +nudged everyone towards what we think is the best solution.