On Wed, 13 Sep 2006 00:48:07 JST [EMAIL PROTECTED] (Andrew Church) wrote: > >Since we're talking about src/* content, feel free to comment/blame me > >about new framebuffer code recently committed by myself. > > Okay, if you insist ;) I was working on getting the tests in > newtest.pl to pass again, and noticed that the "free" index and video_buf > pointer in vframe_list_t are getting out of sync (the current code will > output a "BUG" warning every time you start transcoding and fix the > problem on its own). Do you have any idea where that might be happening?
Yes I have a clue (see my very last commit). :) I think that pops out from a misinterpretation of mine of vframe_list_t.free field meaning. I thought the meaning was exactly the opposite of truth :) After a bit of thinking, the correct interpretation should be: 'free' field is the index of buffer that DOES NOT have the 'real' video frame data in the pair of buffers that vframe_list_t carries (so it has 0/1 possible values). example: it's always safe to do (having something like video_operation(source_buffer, dest_buffer)) as follows vframe_list_t vf; /* ... */ video_operation(vf.video_buf_Y[!vf.free], vf.video_buf_Y[vf.free]); /* ... */ vf.free = !vf.free; /* that efficiently 'swaps' buffer [indexes]. I hope my english is good enough to express my thought here :) If I'm right, my very last commit will fix this issue. > > * --keep_asr is now disabled, as the code implementing it is broken (it > >> doesn't interact properly with other options). If you need to keep > >> the same aspect ratio, calculate the appropriate values manually. > > > >Maybe that should be reintroduced for 1.1.0? I'll take a look later to > >figure out what's wrong (I've a very few clues right now :) ). > I don't know; I'd tend more toward dropping it completely, both > because there's really no way to get it working (that I can think of) > without making the interactions with -j -Z -Y etc. very confusing-- > remember all the trouble we had with export profiles doing the same > thing?--and also because you can already use e.g. -Z 352x or -Z x288 to > maintain aspect ratio when zooming, so there doesn't seem to be too much > need for --keep_asr in the first place. Mostly agree; I don't agree completely (yet) because I've not yet take a look at the code :) > I'm pretty sure you're right about what it was supposed to do, but > since libtc didn't pay it any attention it wasn't really meaningful. ;) > It might still be a good idea to implement a switch like that, but that > can be re-added after the code to handle it is in place, rather than > trying to drag around old code that wasn't designed for it. Fixing this should'nt be too much work, I'll append this task at the (very) end of my TODO list :) Best regards, -- Francesco Romani - Ikitt ['people always complain, no matther what you do'] IM contact : (email first, Antispam default deny!) icq://27-83-87-867 known bugs : http://www.transcoding.org/cgi-bin/transcode?Bug_Showcase tiny homepage : http://fromani.exit1.org (see IDEAS if you want send code!)
