| Age | Commit message (Collapse) | Author |
|---|
|
Now modules allocate and return an opaque setup pointer in
res_setup when they implement a setup method.
Defaults are utilized when ${module}_create_context() receives a
NULL setup. The default setup used in this case should match the
defaults/preferred values emitted by the module's setup method.
But performing setup should always be optional, so a NULL setup
provided to create_context() is to be expected.
No cleanup of these setup instances is currently performed, so
it's a small memory leak for now. Since these are opaque and may
contain nested references to other allocations, simply using
free() somewhere in the frontend is insufficient. There will
probably need to be something like a til_module_t.setup_free()
method added in the future which modules may assign libc's free()
to when appropriate, or their own more elaborate version.
Lifecycle for the settings is very simple; the setup method
returns an instance, the caller is expected to free it when no
longer needed (once free is implemented). The create_context
consumer of a given setup must make its own copy of the settings
if necessary, and may not keep a reference - it must assume the
setup will be freed immediately after create_context() returns.
This enables the ability to reuse a setup instance across
multiple create_context() calls if desired, one can imagine
something like running the same module with the same settings
multiple times across multiple displays for instance. If the
module has significant entropy the output will differ despite
being configured identically...
With this commit one may change settings for any of the modules
*while* the modules are actively rendering a given context, and
the settings should *not* be visible. They should only affect
the context they're supplied to.
|
|
This is a preparatory commit for cleaning up the existing sloppy
global-ish application of settings during the iterative _setup()
call sequences.
Due to how this has evolved from a very rudimentary thing
enjoying many assumptions about there ever only being a single
module instance being configured by the settings, there's a lot
of weirdness and inconsistency surrounding module setup WRT
changes being applied instantaneously to /all/ existing and
future context's renderings of a given module vs. requiring a new
context be created to realize changes.
This commit doesn't actually change any of that, but puts the
plumbing in place for the setup methods to allocate and
initialize a private struct encapsulating the parsed and
validated setup once the settings are complete. This opaque
setup pointer will then be provided to the associated
create_context() method as the setup pointer. Then the created
context can configure itself using the provided setup when
non-NULL, or simply use defaults when NULL.
A future commit will update the setup methods to allocate and
populate their respective setup structs, adding the structs as
needed, as well as updating their create_context() methods to
utilize those setups.
One consequence of these changes when fully realized will be that
every setting change will require a new context be created from
the changed settings for the change to be realized.
For settings appropriately manipulated at runtime the concept of
knobs was introduced but never finished. That will have to be
finished in the future to enable more immediate/interactive
changing of settings-like values appropriate for interactive
manipulation
|
|
Just mechanical replacement of some remaining ad-hoc
til_module_t.create_context() calls.
The montage module continues using an ad-hoc call because it
forces num_cpus=1 since it's already a threaded using a fragment
per module's tile. This suggests the til_module_create_context()
call should probably accept a num_cpus parameter, perhaps
treating a 0 value as the "automagic" discover value so callers
can explicitly set it when necessary.
|
|
Previously if you supplied an empty setting value like so:
"--module=compose,layers="
The interactive setup would get itself into an infinite loop
because the layers setting is already present, but has a NULL
value. This wasn't a NULL value, it was a "" value.
The parser should just fallthrough to the value state from the
equal state after recording the value's start pointer. This will
result in a "" value getting allocated and assigned to the value
before the loop breaks out on the '\0' immediately following the
'='.
There are probably other edge cases which need better handling
here.
|
|
Commit 7ff8ef included some fast and dirty fixups to
til_settings_apply_desc_generators(), but left an inappropriate
return path out of the iterator resulting in the caller accessing
a NULL res_desc.
The practical result of this was segfaulting in rototiller when
configuring anything utilizing desc generators, like drm_fb.
|
|
Now that there's a decoupled libtil usable by alternative
frontends by vendoring rototiller, the build should support
fb-less rototiller-less configurations.
In lieu of this change glimmer's build requires sdl2 despite not
actually utilizing sdl_fb. Now that shouldn't be necessary,
should there be neither libdrm or sdl2 present we'll only produce
libtil and no rototiller binary at all.
|
|
Introduce drawing styles, adding a line style in addition to the
existing points.
Settings are style={points,lines}, default is now lines.
|
|
No idea why this was in there. Though it makes the leader jump
by huge values, since it still winds up as samples on the same
path, just non-continuous, the swarm still appears correct,
hiding this braino.
If you reduce the SWARM_SIZE to a tiny number like 2 though, it
becomes very obvious that the leader is teleporting all over the
place.
|
|
Particularly in implementing a stateful/"retained" GUI it can be
desirable to embed something like a widget pointer in a
til_setting_t once described and shown to the user.
Management of this pointer is largely nonexistant from the
libtil perspective. It's simply initialized to NULL when a new
setting is added, and never accessed again. 100% the caller's
responsibility.
This works fine since libtil/til_settings_t only accumulates
til_setting_t entries and never removes them except when
discarding an entire til_settings_t wholesale.
|
|
This resulted in a NULL ptr deref, simply treating as invalid
since what's the point of handling a composition devoid of any
layers - it's probably a mistake.
|
|
attempt at making these more clear
|
|
Now that til_setting_t.desc is not only a thing, but a thing that
is intended to be refreshed regularly in the course of things
like GUI interactive settings construction, it's not really
appropriate to try even act like this these are const anymore.
|
|
Mechanically replaced ad-hoc til_module_t.destroy_context()
invocations with helper calls.
|
|
|
|
Always only capitalize the first letter, never capitalize like
titles.
|
|
Originally the thinking was that rototiller modules would become
dlopen()ed shared objects, and that it would make sense to let
them be licensed differently.
At this time only some modules I have written were gplv3, Phil's
modules are all gplv2, and I'm not inclined to pivot towards a
dlopen model.
So this commit drops the license field from til_module_t,
relicenses my v3 code to v2, and adds a gplv2 LICENSE file to the
source root dir. As of now rototiller+libtil and all its modules
are simply gplv2, and anything linking in libtil must use a gplv2
compatible license - the expectation is that you just use gplv2.
|
|
This is helpful for forcing underlying setup methods to
redescribe their settings, regardless of what a til_settings_t's
internal state is.
|
|
When integrating libtil into higher order application/gui
toolkits, it's desirable to passthru any unhandled arguments to
the toolkit's initializer.
The specific impetus for adding this is glimmer, which uses
gtk/glib where g_application_run() wants an argv but won't know
what to do with the libtil-recognized args.
|
|
The existing iterative *_setup() interface only described
settings not found, quietly accepting usable settings already
present in the til_settings_t.
This worked fine for the existing interactive text setup thing,
but it's especially problematic for providing a GUI setup
frontend.
This commit makes it so the *_setup() methods always describe
undescribed settings they recognize, leaving the setup frontend
loop calling into the *_setup() methods to both apply the
description validation if wanted and actually tie the description
to respective setting returned by the _setup() methods as being
related to the returned description.
A new helper called til_settings_get_and_describe_value() has
been introduced primarily for use of module setup methods to
simplify this nonsense, replacing the til_settings_get_value()
calls and surrounding logic, but retaining the til_setting_desc_t
definitions largely verbatim.
This also results in discarding of some ad-hoc
til_setting_desc_check() calls, now that there's a centralized
place where settings become "described" (setup_interactively in
the case of rototiller).
Now a GUI frontend (like glimmer) would just provide its own
setup_interactively() equivalent for constructing its widgets for
a given *_setup() method's chain of returned descs. Whereas in
the past this wasn't really feasible unless there was never going
to be pre-supplied settings.
I suspect the til_setting_desc_check() integration into
setup_interactively() needs more work, but I think this is good
enough for now and I'm out of spare time for the moment.
|
|
Currently this was done rather late for vestigial pre-libtil reasons;
it used to be a local function for specifically "create rendering
threads" purpose.
But it's rather awkward now to see such an initializer called late
after myriad other til_* API is being used, and there's nothing
gauranteeing til_init() will continue to strictly create rendering
threads.
Nothing is actually changing in what til_init() does here, it's just a
cosmetic movement of the call site and s/librototiller/libtil/ in the
error message.
|
|
Just removing some copy pasta from the error paths, nothing
functionally different.
|
|
The existing simulation would always accumulate, eventually
filling the volume with density.
This adds a decay to diminish the density, with the default less
quickly filling the volume vs. before.
|
|
These are making it into the settings strings, it's benign only
because regexps aren't currently being enforced. Fix it up
anyways.
|
|
This is totally opt-in for libtil callers, but is a step
towards enabling uniform cli invocations across frontends.
The help side of this is particularly janky, but since what's
appropriate there is directly related to the args parsing it
seems appropriate to bring along. The janky part is the
implicit output formatting assumptions being made, as-is it
doesn't really lend itself well to being augmented into broader
frontend help output. Alas, this is rototiller playground, so
let's just go easy and assume frontends will largely spit out
whatever this provides - or completely replace it if appropriate.
|
|
Largely mechanical rename of librototiller -> libtil, but
introducing a til_ prefix to all librototiller (now libtil)
functions and types where a rototiller prefix was absent.
This is just a step towards a more libized librototiller, and til
is just a nicer to type/read prefix than rototiller_.
|
|
|
|
This manifests in the current unconfigured rtv glimmer shows, since
the default is a the "none" module when no settings are applied.
But it turns out this isn't just a glimmer problem, "none" is advertised
in the settings as a blanking alternative to snow. So it's actually
broken in rototiller as well.
This fixes it by detecting the nil "none" module's lack of any
prepare_frame or render_fragment methods, and open coding the blanker
with a fb_fragment_zero() inline.
|
|
When an fb has been resized, all the pages can be forced to go
through a backend free->alloc cycle by calling this function.
This process gets queued and realized as the pages pass through
the inactive state.
This does mean that whatever pages are already queued as ready
still get displayed, so it's not exactly without
racing/flickering at the fringes on window growth.
It's possible to do something with the pages in the ready queue
to combat this, but for now I'm just leaving them be to keep it
simple. The ready pages could be stretched or something...
|
|
Adds fb_t.n_pages member, we need the number of pages for queueing
rebuilds.
|
|
This introduces a previous pointer to _fb_page_t, but it's
only used on the inactive list.
This is a preparatory commit to facilitate adding inactive pages
at the head and consuming inactive pages off the tail, without
having to always iterate the list for the new tail.
Rebuilding pages on resize needs to happen somewhere, and the
inactive list seems like the right place to do it. And to do it
with a simple per-fb pages-needing-rebuild counter requires
turning the inactive list into a FIFO queue.
|
|
It's desirable to be able to cancel a rendering loop thread
used in librototiller callers, which will necessarily use these
functions.
When the cancellation originates from a GUI thread, which then
joins on the rendering loop thread being cancelled, this blocks
the GUI thread until the rendering loop thread realizes the
cancellation and exits.
But the GUI thread is likely party to the machinery of consuming
fb pages being produced by librototiller, throwing them
on-screen, and making previously visible pages available for
reuse.
The rendering loop thread blocks on waiting for that making pages
available step, to have space for rendering a new frame into.
This creates a circular dependency, and when the GUI thread is
stuck in the cancel+join of the render loop thread, this
establishes potential for deadlock when there's no extra page
already available, and the GUI thread can't service its event
loop to flip a page.
Thread cancellation is the escape hatch for such situations,
which is why cancellation can't be disabled across the critical
section of librototiller calls and reenabled+polled at the top of
the render loop thread. We *need* cancellation to wake up and
break out of all the cancellation points within that critical
section, so they don't deadlock in this scenario. They need to
wake up, cleanup any held locks, and the calling thread exit so
the join may finish and life goes on.
Hence, put these unlocks in cleanup handlers so callers can
cancel them without leaking held locks.
|
|
This is a wrapper around threads_wait_idle(), so the caller
can easily synchronize with outstanding work before tearing
things down or otherwise rejiggering the world.
|
|
and ignore NULL parameters as benign
|
|
None of the existing fb_ops_t implementations need this, but due
to how GTK+ works, the GTK+ frontend using librototiller will likely
want to wire up calling fb_flip() on the fb from behind fb_ops.
|
|
This is a first approximation of separating the core modules and
threaded rendering from the cli-centric rototiller program and
its sdl+drm video backends.
Unfortunately this seemed to require switching over to libtool
archives (.la) to permit consolidating the per-lib and
per-module .a files into the librototiller.a and linking just
with librototiller.a to depend on the aggregate of
libs+modules+librototiller-glue in a simple fashion.
If an alternative to .la comes up I will switch over to it,
using libtool really slows down the build process.
Those are implementation/build system details though. What's
important in these changes is establishing something resembling a
librototiller API boundary, enabling creating alternative
frontends which vendor this tree as a submodule and link just to
librototiller.{la,a} for all the modules+threaded rendering of
them, while providing their own fb_ops_t for outputting into, and
their own settings applicators for driving the modules setup.
|
|
These modules are meta modules, and the only place this
information is presented currently is in the rtv module captions
overlaying the visual output of unrelated modules.
So it's rather misleading to put the meta module's author and
license on-screen when what's being shown is arguably just a tiny
fraction of the meta module's contribution.
Rather than bother with constructing license and author lists at
runtime from the modules incorporated by these meta modules,
let's instead adopt a policy of meta modules omit any declaration
of license or authorship outside of the source. This is a simple
solution for now, it can be revisited later if necessary.
Changing the .author member of rototiller_module_t to an
.authors() function pointer wouldn't be difficult. But it does
open up something of a can of worms when considering recursive
dependencies and needing to construct unique authors and licenses
lists from things like nested meta modules. Obviously there
can't be infinite recursion as that would manifest in the
rendering path as well, but what I'm more concerned about is
properly handling potentialy quite long lists. It's already
annoying when rtv has to deal with a long settings string, which
I believe currently is just truncated. The same would have to be
done with long authors/licenses I guess.
In any case, I think it's probably fine to just leave authorship
and license ambiguous when a meta module is shown in rtv. It's
certainly preferable to vcaputo@pengaru.com getting credit for
everything shown in the three meta modules currently implemented,
or more specifically, the two shown in rtv; compose and montage.
Note this required making rtv tolerante of NULL .license and
.author rototiller_module_t members.
|
|
A lot of errors were being conflated as ENOMEM due to the lazy
use of NULL pointer returns for errors.
This commit reworks a handful of those return paths to instead
return an errno-style int, storing the results on success at a
supplied result pointer.
It's kind of ugly, and I make some assumptions about libdrm
setting errno on failure - it too uses this lazy API of returning
NULL pointers on failure. Hopefully errno is always set by an
underlying ioctl failing.
The SDL error API is also pretty gross, being cross-platform it
defines its own error codes so I try vaguely map these to errno
values.
I'm considering this a first approximation at fixing this up, but
there are probably bugs as I did it real fast and nasty.
It at least seems to all still work OK here in the non-error
paths I tested. So it doesn't seem more broken than before at a
glance.
|
|
Some omissions, nothing in these is public outside of what's
explicitly plumbed out via fb_ops_t.
|
|
Minor cosmetic consistency fixup
|
|
Just a fun little swarm based loosely on 80s-era boids
It would be interesting to make stuff like the # of particles
and the weights runtime configurable, or exposed as knobs.
Using a Z-buffer for occlusions and perhaps shading by depth might
make a significant improvement on the visual quality. It might also
be interesting to draw the particles as lines connecting their current
position with their previous, instead as pixels. Or fat pixels like
stars...
|
|
This is a quick and dirty jab at normalizing the plasma size to
be independent of the frame size.
I kind of hate this module as-is, it's tempting to discard all
the fixed point stuff and just redo it using floats.. the plasma
itself isn't that attractive as-is either.
But I have other things to work on currently, just wanted to make
it so the plasma doesn't look like a solid color in the montage
tile.
|
|
- move LUT initialization to context create
- minor syntactic changes
|
|
- switch puddle_sample() to 0..1 coordinates to avoid
some pointless extra arithmetic on every pixel
- avoid redundant ->w multiplies in puddle_sample()
- avoid multiplies in inner loops of drizzle_render_fragment()
by accumulating coordinates w/addition instead
I noticed full-screen 'compose' was struggling to keep a full
frame rate on my laptop when testing with the new 'plato' layer.
valgrind profiles showed drizzle as the big hog, mostly the
puddle_sample() function. These changes help but it's still not
great, getting much better will likely become invasive and
crufty. It would be nice to cache the vertical lerp results and
reuse them across puddle_sample() calls when valid, that might be
a useful TODO.
The runner-up is spiro, prolly some low-hanging fruit there as
well, I haven't looked yet.
|
|
now: "drizzle:stars:spiro:plato"
|
|
plato implements very simple software-rendered 3D models of
the five convex regular polyhedra / Platonic solids
Some TODO items:
- procedurally generate vertices at runtime
- add hidden surface removal setting (Z-buffer?)
- add flat shaded rendering setting
- add gouraud shading, maybe phong too?
- show dual polyhedra
This was more about slapping together a minimal 3D wireframe
software renderer than anything to do with polyhedra, convex
regular polyhedra just seemed like an excellent substrate since
they're so simple to model.
|
|
In case some code path creates module contexts and renders without
applying settings, it's important to ensure there are defaults.
As-is this would have crashed compose because compose_layers would
have been NULL, and compose_create_context() assumed compose_layers
always contained something useful.
Montage would have been an example of this, though for other reasons
montage has had compose disabled so I don't think anything currently
would have triggered this.
|
|
The threaded rendering backend isn't reentrant and compose could
hypothetically have montage as a layer triggering infinite
recursion.
For now use a big hammer and block compose module from montage.
|
|
Once upon a time montage had to skip stars and pixbouce because
they crashed. That has since been corrected, but the commit which
removed the skips didn't remove the lookups.
This removes the leftover lookups.
|
|
This module needs some love, but it already always supplies 1 to
the open-coded rendering for the tiles. It never should have been
supplying the real num_cpus to module.create_context(), then only
supplying 1 for rendering.
|
|
--module=compose,layers=first:second:third:...
this draws the named modules in the order listed, overdrawing the
output of the previous layers in a cumulative fashion.
|
