mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-12-24 17:20:36 +00:00
32 lines
1.3 KiB
Text
32 lines
1.3 KiB
Text
|
= Cleanups =
|
||
|
|
* often two g_object_set for same object one after the other
|
||
|
|
* use GST_*_OBJECT () more often
|
||
|
|
* there are two gst_element_set_state() one after each other
|
||
|
|
|
||
|
|
= Renaming =
|
||
|
|
* internal use of img->image, vid->video
|
||
|
|
|
||
|
|
= Refactorisation =
|
||
|
|
* gstcamerabin:gst_camerabin_rewrite_tags
|
||
|
|
- sounds fishy, should use normal tagsetter method
|
||
|
|
- gst_camerabin_rewrite_tags_to_bin(9 why don't we just send a tag-event?
|
||
|
|
|
||
|
|
* file-name property
|
||
|
|
- supplying an already opened filedeskriptor would be more safe
|
||
|
|
- need to check what filesink does if the file exists and cannot be overwritten
|
||
|
|
|
||
|
|
* imagbin
|
||
|
|
- we want async operation here (especialy for burst mode capture)
|
||
|
|
- right now, its a bit fragile as we muck with locked_state
|
||
|
|
- main problem is that the location for filesink can only be set in NULL/READY
|
||
|
|
and we need to do that sync'ed with the dataflow. we can't use multifilesink
|
||
|
|
as it does a file per pad_push
|
||
|
|
- one problem of the current approach is that we can't have an image in e.g,
|
||
|
|
postprocessing while anotherone is beeing saved
|
||
|
|
- we could use a pool of imagebins:
|
||
|
|
- configure one (set filename)
|
||
|
|
- push_buffer
|
||
|
|
- on eos, put it back to the pool
|
||
|
|
- for this we need to check that we can have multiple instances of e.g.
|
||
|
|
dsp jpeg encoders
|