Lion ponderings

 Noted a lot of blank frames between the clips, after the align routine.  The problem is caused by the user concatenating a lot of clips by pasting.  The align function doesn't touch the video/master track.  Once blank frames are in there, they stay forever. The align function could test if the audio is from the same asset as the video, if the audio source is from the same time as the video & then if the audio end overruns the video.  It would be easiest as another pass.  It could be too much control taken away from the user & too much complexity.  An easier solution might be to select the end of the video track before pasting instead of the end of the EDL or just using concatenate.   ------------------------------------------------------------------------  After years of Ben Heck complaining about Star Trek Into Darkness, the lion kingdom finally watched it to find out why it was so bad.  It wasn't that bad.  Lions com...

 More cinelerra bugs piled up, as lions continued doing more keyframe editing than usual.  Lions need to produce more movies that use all its features, though it has many alternatives at this point.  Constant keyframes didn't work in reverse playback & FloatAutos::automation_is_constant didn't work with constant keyframes at all.  The problem is get_prev_auto & get_next_auto return different values based on the direction.  get_automation_constant was an optimization in the days of 166Mhz to look up only 1 fade value for an entire audio buffer.  It might be overrated but it's technically correct.  It's used in a curious way inside VAutomation::direct_copy_possible because it's always true for a single frame of video.  It might be an artifact from when video was going to be processed in chunks, with pointers to arrays of pointers to frames.  Then we have multiple ways of getting the constant value:  IntAutos::get_automation_constan...

Yet another bug got fixed, whereby it was clearing an edit handle before pasting between 2 edits with the same source, resulting in the after edit being lost & replaced by the rest of the movie in the before edit.  It was clearing in all cases where the selected area had zero length.  The EQUIV operator is more difficult to use than == & won't work if the frame rate is in the thousands.  Lions are now assuming a start & end coming from user input are always going to work with the == operator.     During this process, 1/16 sample zoom exposed a few anomalies.  Noted if the source samplerate doesn't match the project samplerate, it draws the samples offset from the sample ticks.   Also noted the 48khz samples line up with the 24 frame rate but the 44.1khz samples do not.  It's possible for the insertion point to line up on a frame but not a sample in certain cases: double click an edit, ctrl seek to label, alt seek to edit.  Al...

 The answer is yes.  Personal projects lacking enough money for an AI plan are still being developed the old fashioned way.  Lions aren't sure how much longer the manual tools are going to continue working.  GDB is manely an aberration in a world where static analysis has replaced runtime analysis.  Higher level languages are famously being phased out in favor of prose.  Like embedded systems 40 years ago, it's once again really expensive to access the tools that get jobs.  If you only have access to the tools at a day job, you can't take any breaks from working.  The mane barrier to having a free operating system was once developing a free compiler.  Now it's developing a free language model. So is software still free if it requires a paid language model to analyze it or work on it at all?   -----------------------------------------------------------------------------------   When it rains, it pours in debug land.  N...

  Particularly proud of having sub sample zoom again, after 30 years.  It was last seen in 1996, possibly on sound sculptor & definitely soundedit 16 for 68k mac.    https://www.macintoshrepository.org/2514-soundedit-16  Soundedit 16 jumped from 1x to a single sub sample zoom.  Cinelerra was based on that.  The early waveform editors didn't pile on tons of functionality through context menus & handles, which makes sub sample zoom quite difficult now.  There is still no manual drawing of the samples because it would be pointless.  Ctrl+arrow dragging of single sample edit works.  Single sample edit handles work.   Context menus & arrow dragging without CTRL don't work because the edit handle hit point overlaps all of the single sample edit.  Single sample zoom could be increased further or the hit point could be shrunk.  Bigger zoom would make it harder to see the context.  A smalle...

  Imagine not knowing if keyboards, joysticks, light pens, mice, touch pads, touch screens, or spoken language were going to be the chosen user interface.  In a time when lions could scarcely imagine reaching a double digit age, it really felt like we'd be drawing with joysticks forever.    That was us before 1984.  To lions old enough to recognize artifacts from that era, that's a blue/green screen.  They were never interacting with a user interface.  It was composited later.     Laid down a bruiser.  1 month after the last 10mph effort broke things down again, went back to alternating between walk & run days to try to increase the intensity.  A steady decline throughout August ended in barely being able to hit 6mph.  The walk/run pattern got back up to a 7m45 mile.  --------------------------------------------------------------------------  After fixing a regression in the keyframe paste function, embarked on fi...

 Fixed a bug where the integer arguments to Resample were overflowing.  It kept the resamplert plugin from resampling long segments, while seeking still worked.  It might explain why using resamplert to fix synchronization didn't work.  It might have seeked to the right starting sample but drifted during playback.  The whole matter made lions ponder how a modern language model could seek out every case of an overflowing integer argument in the program while it would take forever do to it manually.  A lot of stuff isn't worth doing anymore, in lieu of waiting for a more effective language model to come down in price.  Part of the game is knowing what the language models are good at, expressing the tasks in spoken language & minimizing the prompt lottery.  That's constantly changing & not widely known.  Directed, static analysis was 1 area no-one knew about until July.  Until then, we thought it needed a working executable...

 Pondering the idea of a cruise again, it wouldn't be that unbearable, as long as it didn't spend too many days at sea.      https://www.youtube.com/watch?v=0UiHaCX2i8E     An inside facing cabin would work, since lions were more interested in touring land than the ocean.  It would still be quite expensive to travel to the starting point.   It's slim pickings from San Francisco.  For $3k, they go to a remote port in Canadia by legal mandate & then the very southern panhandle of Alaska instead of Alaska proper.  Too far south to see a real glacier, though Greenspun said they go ashore.  Viking has the ultimate cruise.     $7k & up, not even countable in fiat money.  Cheaper in the dead of winter. This travel thing might be getting out of paw.  Lions have escalated spending over the decades, starting with a dumpy motel in Burbank, 2012, slowly escalating to a super expensive hotel in 2025....

https://code.videolan.org/videolan/dav1d  There is finally a software decoder for AV1, after many years of only hardware decoding.  It's very difficult to compile, requiring yet another large build system, specific versions of nasm, yasm.  It got stuck on some kind of checkasm.asm file.  Upgrading nasm & yasm to the latest versions, deleting & recreating the build directory got it to compile.   Av1d seems to be funded by alphabet corporation which requires them to use no less than a certain number of academic dependencies.   ffmpeg has simply https://www.ffmpeg.org/general.html#dav1d  --enable-libdav1d     It requires package_cfg so you have to run 'ninja install;ldconfig' in dav1d-master/build.  There's no local library option.  The latest dav1d actually compiled with ffmpeg 5, but couldn't output any frames to Cinelerra.  The standalone ffmpeg transcoded properly.  It's proven easier to use ffmpe...

  Took a few weeks after Warsh came on board, but now we know the balance sheet charges upward despite the talk.     Contrarian opinion: interest rates are only going down until the national debt is back below 100% of GDP.  The game continues to be jawboning the markets with inactivity.   It...can't...be   -------------------------------------------------------------------------------------------   https://www.instagram.com/reels/DbEJdStifPB/  Should pay more attention to the city center concert series.  They had some big names this year instead of the usual cover bands.  David Benoit was there.  Lions played the living daylights out of his xmas album & wrote a program to try to transcribe his recording of xmas time is here.  After the last performance, 74 year old Lee Ritenour, saw a really serious looking, black guitar player coming off stage.  They used 1 of the many empty stores as a dressing room.  He had 2 r...

   Didn't know the master of looking perplexed, Anderson Cooper was an heir to the Vanderbilt fortune, 5 generations down.  His mother Gloria Vanderbilt was briefly married to Leopold Stokowski.  The enemy of the rich is the rich.   There was a photo of her in an opulent NY apartment.  He wasn't afraid to say she never went near the subway except once.  Oh the number of millionaires who ride it today, even to JFK. The Vanderbilt fortune only grew for 2 generations.  It was gradually whittled down ever since.  Lions only pondered this after Wolf contained footage of a Vanderbilt mansion. --------------------------------------------------------------------------------------------  Speaking of body counts,   https://www.tiktok.com/@yourowkelsey     World class trophy wife.  Helas,    the tomboy outfit came off & the magic was gone.  31 year olds definitely pop after 51.  Lions were decidedly ...