| 1 |
|
| 2 |
|
| 3 |
Properties: |
| 4 |
|
| 5 |
"template_caps" |
| 6 |
The union of all possible caps that a pad might have in any |
| 7 |
condition. Is a class property, and doesn't change over the |
| 8 |
life of an object. Set by the element during class init. |
| 9 |
|
| 10 |
"internal_caps" |
| 11 |
The caps that the element will allow on a pad, given its current |
| 12 |
state and the external_caps of all the other pads of the element. |
| 13 |
Set by the element. |
| 14 |
|
| 15 |
"external_caps" |
| 16 |
The caps representing what the core will allow on a pad. This is |
| 17 |
a combination of the peer's allowed caps, and filtering caps, and |
| 18 |
any in-process negotiation restrictions. Set by the core. |
| 19 |
|
| 20 |
"negotiated_caps" |
| 21 |
The format that the element will use. Set by the core. |
| 22 |
|
| 23 |
"filter_caps" |
| 24 |
The caps intersected with everything else during negotiation. Set |
| 25 |
by the application. |
| 26 |
|
| 27 |
"negotiable" |
| 28 |
A boolean that indicates that the pad is ready to negotiate. Set |
| 29 |
by the element. |
| 30 |
|
| 31 |
|
| 32 |
|
| 33 |
Negotiation steps, from the element perspective: |
| 34 |
|
| 35 |
- During object initialization, an element must set template_caps |
| 36 |
and allowed_caps for each pad. |
| 37 |
|
| 38 |
- In response to changes in the internal state of the element or in |
| 39 |
the peer_allowed_caps of any other pad owned by the element, the |
| 40 |
element should update allowed_caps of a pad. |
| 41 |
|
| 42 |
- Pads are initially created with negotiable as FALSE. When an |
| 43 |
element wants to indicate that negotiation is allowable, it should |
| 44 |
set negotiable to TRUE. |
| 45 |
|
| 46 |
- In response to a negotiate signal, an element should copy the caps |
| 47 |
in "allowed_caps" to "negotiated_caps", and extract any information |
| 48 |
needed to handle the format. At this point, "allowed_caps" is |
| 49 |
guaranteed to be fixed. |
| 50 |
|
| 51 |
|
| 52 |
Negotiation steps, from the core perspective: |
| 53 |
|
| 54 |
|
| 55 |
|
| 56 |
|
| 57 |
|
| 58 |
|
| 59 |
|
| 60 |
<ds> I'd like to split it up into manageable "projects" |
| 61 |
<ds> hopefully some of which we can recruit others to do |
| 62 |
<Company> i'd like a decision about if we drop threads |
| 63 |
<ds> in what sense? |
| 64 |
<ds> threads being not required? |
| 65 |
<Company> threads being not part of the API |
| 66 |
<ds> or threads being not used |
| 67 |
<Company> so no thread safety considerations |
| 68 |
<ds> I would like threads to not be part of the API |
| 69 |
<ds> requiring apps/plugins to be written thread-safely is insane |
| 70 |
<Company> the question is this: If someone calls gst_element_set_state (element, state), does he have to marshal it into the correct thread himself or not? |
| 71 |
<ds> no |
| 72 |
<ds> reason being: we already have to deal with subthreads |
| 73 |
<ds> we == the core |
| 74 |
<Company> then you'll have a lot of problems |
| 75 |
<Company> i'd like to just drop subthreads and be done with it |
| 76 |
--> StyXman ([email protected]) has joined #gstreamer |
| 77 |
<ds> and run everything in one thread? |
| 78 |
<Company> yes |
| 79 |
<Company> schedulers are free to use threads though |
| 80 |
<ds> this is obviously a retreat |
| 81 |
<Company> right |
| 82 |
<ds> but perhaps a strategic one |
| 83 |
<Company> otherwise you have to guarantee that you get no races |
| 84 |
<ds> I've seen this as a viable option for some time |
| 85 |
<Company> there's stuff that just does not work as is currently |
| 86 |
<ds> it always seemed difficult to convince other people that it's the right thing to do |
| 87 |
<ds> for example, wtay |
| 88 |
<Company> like setting xwindow ids, changing states or app-events |
| 89 |
<Company> like seeks |
| 90 |
<-- iain has quit ("leaving") |
| 91 |
<Company> all of this has the typical thread issues currently: it sometimes works and sometimes crashes the app |
| 92 |
<Company> and that's not acceptable to me |
| 93 |
<ds> let's do it |
| 94 |
<Company> if you look at my 0.10 code, I've tried to have GstObjects belong to a "context" |
| 95 |
<ds> this all fits into the idea that I think we should be a lot more like gtk architecturally |
| 96 |
<ds> for example, gtk doesn't have anything like caps negotiation, where elements call each other recursively |
| 97 |
<Company> and have functions ensure they're in the right context or otherwise marshal |
| 98 |
<Company> yes it does |
| 99 |
<ds> ? |
| 100 |
<Company> size requests and allocations |
| 101 |
<ds> but that's more like how I want it -- setting a property and firing a signal telling others that it's updated |
| 102 |
<ds> as i understand it |
| 103 |
<Company> though gtk has the bonus of being a tree which we don't have, we have a directed, maybe cyclic graph |
| 104 |
<Company> and for caps, the graph isn't even directed |
| 105 |
<Company> gtk is far too simple for gst |
| 106 |
<ds> true |
| 107 |
<Company> i can't even come up with an analogy to a video transcoder |
| 108 |
<ds> but we do some relatively simple stuff in a complicated way |
| 109 |
<-- StyXman ([email protected]) has left #gstreamer |
| 110 |
<Company> like? |
| 111 |
<ds> caps negotiation |
| 112 |
<Company> that's not simple |
| 113 |
<ds> passing around the data should be |
| 114 |
<ds> was it you or thomasvs that wanted to split core headers into app headers and plugin headers? |
| 115 |
<Company> I definitely wanted that |
| 116 |
<Company> app/plugin/registry to be exact |
| 117 |
<ds> registry? |
| 118 |
<ds> for functions that access "internal" stuff? |
| 119 |
<Company> stuff an app normally doesn't touch |
| 120 |
<Company> so we can do something like declare the "app" stuff supported for a longer time |
| 121 |
<ds> oooh, good call |
| 122 |
<ds> I'm sold |
| 123 |
<Company> so, now that threads are gone, what are the biggies I want in 0.10? |
| 124 |
<Company> 1) non-blocking main-loop integrated processing |
| 125 |
<ds> that's rather important if you don't have threads :) |
| 126 |
<Company> right |
| 127 |
<Company> but it works in my gst-01.0 stuff already |
| 128 |
<Company> minus the elements that need to be ported of cause |
| 129 |
<Company> which aren't that many fortunately |
| 130 |
<Company> though some are hard (like gnomevfs), because they have no non-blocking API |
| 131 |
<ds> a subclass of GstElement that fires up a helper thread? |
| 132 |
<ds> (dirty, but it works) |
| 133 |
<Company> probably |
| 134 |
<ds> - autopluggers would be easier to write if there was a core method |
| 135 |
<ds> to do what plugidentities do. |
| 136 |
<Company> no clue how to implement that |
| 137 |
<ds> there's also a note about bins being more involved with negotiation |
| 138 |
<Company> that's the next biggie: 2) fix the negotiation process to allow good autoplugging |
| 139 |
<ds> that was something I thought about last time around |
| 140 |
<Company> bins? |
| 141 |
<ds> that sounds like something I'd work on |
| 142 |
<ds> GstBin |
| 143 |
<Company> the only reason bhins still exist for me is to group elements |
| 144 |
<ds> so that a bin subclass could override how its children are negotiated |
| 145 |
<ds> much like GtkBin, actually |
| 146 |
<ds> and allocation |
| 147 |
<Company> hum |
| 148 |
<Company> how do you solve that issue with links that are only partly inside the bin and partly inside another one? |
| 149 |
<ds> that would be up to the parent (grandparent?) |
| 150 |
<Company> so all links must be in the same toplevel bin? |
| 151 |
<ds> yes |
| 152 |
<ds> isn't that the case currently? |
| 153 |
<ds> probably not checked |
| 154 |
<Company> pfff |
| 155 |
<Company> currently you can do anything |
| 156 |
<ds> heh |
| 157 |
<ds> what about state changes? |
| 158 |
<Company> dunno |
| 159 |
<Company> it's not really important for me |
| 160 |
<Company> apart from the EOS stuff |
| 161 |
<ds> I'd like states to be split into "what state does the app want me to be in" and "what state am I in" |
| 162 |
<ds> with the latter being under control of the element |
| 163 |
<Company> yeah, that'd be nice |
| 164 |
<ds> that way, you don't have multiple people trying to control the state |
| 165 |
<Company> we definitely need to fix EOS |
| 166 |
<Company> i'm not sure how we should handle errors |
| 167 |
<ds> element errors? |
| 168 |
<Company> yeah, if we want to explicitly unset errors or not |
| 169 |
<Company> currently on error the element goes to paused |
| 170 |
<ds> it's easier with actual_state |
| 171 |
<Company> in theory i just set it to PLAYING again and continue? |
| 172 |
<ds> the actual_state goes to PAUSED and fires a signal |
| 173 |
<Company> dunno, maybe actual_state should go to ERROR |
| 174 |
<ds> yeah |
| 175 |
<Company> that'd require gst_element_unset_error though |
| 176 |
<ds> with actual_state, we can add states without thinking too hard |
| 177 |
<Company> but if you can just unset it, why not just do it and continue |
| 178 |
<ds> the app may want to be involved |
| 179 |
<ds> especially to get the error message |
| 180 |
<ds> do we want a separate api section for autopluggers? |
| 181 |
<Company> why would we want that? |
| 182 |
<ds> or make it part of registry |
| 183 |
<Company> maybe s/registry/internal/ or some other name |
| 184 |
<ds> because autopluggers may want to access internal information that we wouldn't expose to apps or plugins |
| 185 |
<ds> could you add gst/gstcompat.c to your arch repo? |
| 186 |
<Company> hm |
| 187 |
<Company> what internal info would an autoplugger need |
| 188 |
<ds> depends |
| 189 |
<ds> whatever replaces plugidentities |
| 190 |
<Company> a simple one (decoder) just needs caps and ranks |
| 191 |
<ds> I meant for negotiation information |
| 192 |
<Company> hm |
| 193 |
<Company> no, i don't want a seperate section for it |
| 194 |
<Company> just put it in "internal", "extended" or whatever |
| 195 |
<ds> ok |
| 196 |
<ds> do we want to plan for the app api to be ABI compatible past 0.10? |
| 197 |
<ds> er, that's silly |
| 198 |
<ds> er, kinda silly |
| 199 |
<Company> API compat |
| 200 |
<Company> you could achieve ABI compat if you expose all objects just as typedefs |
| 201 |
<Company> lemme reboot into linux so i can add gstcompat.c |
| 202 |
<Company> brb |
| 203 |
<-- Company has quit (Read error: 104 (Connection reset by peer)) |
| 204 |
--> Company ([email protected]) has joined #gstreamer |
| 205 |
--> ploum ([email protected]) has joined #gstreamer |
| 206 |
<-- marv has quit (Read error: 104 (Connection reset by peer)) |
| 207 |
<ds> so do you still support caps registration? |
| 208 |
<-- ploum has quit (Client Quit) |
| 209 |
<ds> registering media types, fields, descriptions, and default fixate functions? |
| 210 |
<ds> and validation |
| 211 |
<Company> yes |
| 212 |
<Company> that's an important thing we need |
| 213 |
<Company> otherwise people mess up by either carelessly doing stuff wrong or not understanding what they're supposed to do |
| 214 |
<ds> whose responsibility is it to register these? |
| 215 |
<ds> libgstplugins? |
| 216 |
<ds> a plugin? |
| 217 |
<ds> do we want this info in the registry so that a plugin can automatically be loaded to provide registration for video/x-raw-rgb? |
| 218 |
<Company> haven't really thought about that part yet |
| 219 |
<ds> it's somewhat similar to typefind functions |
| 220 |
<Company> yeah |
| 221 |
--> ultrafunk ([email protected]) has joined #gstreamer |
| 222 |
<Company> i'd probably have a seperate Factory for caps info |
| 223 |
<Company> and have the system load them when you use the caps info stuff |
| 224 |
<ds> should caps info be manditory? |
| 225 |
<ds> have you synced your repo? |
| 226 |
<Company> i'm trying to, but some commit was interrupted last time... |
| 227 |
<ds> how does one sync a local repository to fd.o? |
| 228 |
<ds> rsync? |
| 229 |
<Company> tla archive-mirror |
| 230 |
<ds> ah |
| 231 |
<Company> see rhythmbox.org => developers for how it works |
| 232 |
<Company> and _never_ press ctrl-c in arch |
| 233 |
<-- Misirlou has quit (Remote closed the connection) |
| 234 |
--> Misirlou ([email protected]) has joined #gstreamer |
| 235 |
<ds> is there a reason why a pad might know (and set) allowed_caps, but not be ready to negotiate? |
| 236 |
<Company> i don't know a good one |
| 237 |
<ds> know a bad one? |
| 238 |
<Company> "arch-mirror: unable to acquire revision lock (internal error in archive-pfs.c(pfs_lock_revision))" |
| 239 |
<Company> stupidly ordered code |
| 240 |
<Company> no, don't know one |
| 241 |
<ds> here's a good question: |
| 242 |
<ds> should the core negotiate links in a pipeline that are negotiatable, even though some other links are not negotiatable? |
| 243 |
<ds> this ends up making us negotiate some things multiple times |
| 244 |
<ds> filesrc ! mad ! audioconvert ! audioscale ! osssink |
| 245 |
<Company> the core doesn't know if links are negotiable |
| 246 |
<ds> it will in the future :) |
| 247 |
<Company> sinesrc ! identity ! osssink |
| 248 |
<Company> how does it know if sinesrc ! identity is negotiable? |
| 249 |
<ds> because the allowed_caps on the two pads intersect |
| 250 |
--> marv ([email protected]) has joined #gstreamer |
| 251 |
<ds> I'm pretty sure that I don't want gst_pad_link() to return negotiation errors |
| 252 |
<ds> ever |
| 253 |
<Company> you can't avoid that |
| 254 |
<ds> I want to separate "linkable" from "negotiable" |
| 255 |
<Company> ah, gst_pad_link |
| 256 |
<ds> so you can link videotestsrc ! identity ! osssink |
| 257 |
<Company> i confused it with GstPadLink |
| 258 |
<Company> yeah, i was thinking about that, too |
| 259 |
<ds> vts ! osssink won't link, since the pad templates don't intersect |
| 260 |
<Company> http://web.ics.purdue.edu/~kuliniew/wp/archives/2004/08/07/error-handling/ |
| 261 |
<Company> :) |
| 262 |
<ds> right now the vts ! id ! osssink pipeline fails to link |
| 263 |
<Company> yeah |
| 264 |
<Company> and goom ! xvimagesink only fails when xvimagesink is at least READY |
| 265 |
<Company> (assuming your xv can't do RGB) |
| 266 |
|
| 267 |
|
| 268 |
|