nandi/frq · deps.edn

nandi/frqpublic⑂ Fork 0

⑂ 57a93bbc4ec783ac1772cfadcc4c573d3f30971c

deps.edn · 100 lines · 5.2 KBClojure    Blame HistoryRaw
A freeq client in jolt, as glimmer components on Vidya   1{:paths ["src"]
3 ;; glimmer owns the reactive half (ratom, components, reconciler);
Paint frq with jvui, and take the last Rust out of the window   4 ;; glimmer-jvui paints it as jvui, which is dvui's shape written in jolt on
5 ;; SDL3 — no shared object of its own, and no Rust anywhere under the UI.
6 ;; That is the whole reason for the switch off glimmer-vidya: libvidya meant
7 ;; jolt-native's Cargo workspace, and a Cargo workspace meant a build.
Show the list and the conversation at once on a wide window   9 ;; Our fork of glimmer, until the reconciler fixes in it land upstream. A
Keep a row's actions from wandering when its line gets an id   10 ;; component unmounted between a cell firing and the queued render running
11 ;; used to render anyway, into widgets the backend had already freed and
12 ;; handed out again — which took the message list apart in a busy channel.
Show the list and the conversation at once on a wide window   13 ;; And replacing a native node with one of another tag left every watcher
14 ;; under it subscribed, so swapping the phone layout for the split one and
15 ;; back shredded the chat: one component's props painted onto another's
16 ;; widget.
Paint frq with jvui, and take the last Rust out of the window   17 :deps {jolt-lang/glimmer  {:git/url "https://gitlab.com/nandithebull/glimmer"
18                            :git/sha "399df371c790d690fb6e4560c3d4d7f838502857"}
19        nandi/glimmer-jvui {:git/url "https://gitlab.com/nandithebull/jolt-native"
Take the jvui that fills the window and has glyphs for the icons   20                            :git/sha "7d907d008b66abc03da83dc36411cf13f3085a35"
Paint frq with jvui, and take the last Rust out of the window   21                            :deps/root "glimmer-backends/glimmer-jvui"}}
A freeq client in jolt, as glimmer components on Vidya   22
Paint frq with jvui, and take the last Rust out of the window   23 ;; What the UI loads is SDL, and jvui declares it itself — SDL3, SDL3_ttf
24 ;; and SDL3_image, all :optional, so the layout tests run on a machine with
25 ;; none of them. Nothing of jolt-native's Cargo workspace is loaded by the
26 ;; window any more; `just tui` still takes libjolttui, which is the one
27 ;; object left and only for the terminal.
Take both native libraries from one place   28
Switch frq.av onto the jolt media plane, and stop shipping libjoltmoq   29 ;; Calls used to be the one thing this client could not do in jolt. The media
30 ;; half is `frq.av.plane` now — MoQ over QUIC from libmoq_ffi, Opus from
31 ;; libopus, H.264 from openh264 through a shim, and the devices from V4L2
32 ;; and ALSA. Signaling was always IRC and still lives in src/frq/av.clj.
Take both native libraries from one place   33
Build the desktop libraries rather than fetching a release of them   34 ;;   just run
Paint frq with jvui, and take the last Rust out of the window   35 :jolt/native [;; The transport half of a call: MoQ over QUIC, the one
Switch frq.av onto the jolt media plane, and stop shipping libjoltmoq   36               ;; piece with no C implementation anywhere. Fetched from
37               ;; moq-ffi's release rather than built — building it means a
38               ;; 1062-crate workspace.
Bind libmoq_ffi from the object's own metadata, not from its header   39
Switch frq.av onto the jolt media plane, and stop shipping libjoltmoq   40               ;; The codecs are NOT in it. moq-ffi's `audio` and `video`
41               ;; features are off in every Linux artifact upstream ships,
42               ;; so Opus and H.264 come from their own C libraries below,
43               ;; which is the whole point of doing this over FFI rather
44               ;; than over a Rust facade.
Fetch libmoq_ffi again, and keep the codecs out of Rust   45
46               ;; No :darwin: :systems is x86_64-linux and aarch64-linux, so
47               ;; a .dylib named here would name a file nothing fetches.
Bind libmoq_ffi from the object's own metadata, not from its header   48               {:name "moq_ffi"
Bind the C libraries a call actually needs   49                :linux ["libmoq_ffi.so"]}
51               ;; The codecs, as C libraries rather than as somebody's
52               ;; bindings to them. Opus first; openh264 beside it when the
53               ;; video half lands.
55               ;; The SONAME, not the bare .so: a `libopus.so` is the -dev
56               ;; symlink and is not what a runtime closure carries. The
57               ;; loader looks these up by name in one directory, which the
58               ;; flake's `nativeAll` is.
59               {:name "opus"
60                :linux ["libopus.so.0"]
61                :darwin ["libopus.0.dylib"]}
63               ;; H.264, one step removed. openh264's C API is a vtable —
64               ;; `ISVCEncoder` is `const ISVCEncoderVtbl*` — and jolt.ffi
65               ;; can only call a literal C symbol, never a function
66               ;; pointer. c/frq_h264.c walks the vtable and exports flat
67               ;; symbols; this is that. libopenh264 itself comes along as
68               ;; its DT_NEEDED and is never named here.
69               {:name "frqh264"
70                :linux ["libfrqh264.so"]
71                :darwin ["libfrqh264.dylib"]}
73               ;; Audio devices. No :darwin — CoreAudio is the other side of
74               ;; that door and is not this library.
76               ;; V4L2 is deliberately absent from this list: the camera is
77               ;; ioctls against libc and the kernel, so there is nothing to
78               ;; load. See frq.capture.v4l2.
79               {:name "asound"
80                :linux ["libasound.so.2"]}]
A freeq client in jolt, as glimmer components on Vidya   81
Paint the same screens into a terminal   82 :aliases {:frq {:main-opts ["-m" "frq.app"]}
A freeq client in jolt, as glimmer components on Vidya   83
Paint the same screens into a terminal   84           ;; The same screens in a terminal. glimmer-tui is the other backend
85           ;; for the same reconciler — libjolttui's tree ABI is libvidya's
86           ;; with cells under it — so `frq.tui` requires it after `frq.app`
87           ;; and renders the hiccup that is already written.
Merge origin/main, and let one jolt-native answer for both backends   89           ;; A path rather than a sha, like glimmer-vidya's would be if this
90           ;; alias were the way in: `just tui` builds jolt-native out of the
91           ;; flake and hands both Jolt halves over with -Sdeps, so this is for
92           ;; a checkout beside the tree and nothing else reads it.
Paint the same screens into a terminal   93
94           ;;   just tui
95           :tui {:extra-deps {nandi/glimmer-tui
Retire the release-bumping script   96                              {:local/root "../jolt-native/glimmer-backends/glimmer-tui"}}
Paint the same screens into a terminal   97                 :main-opts ["-m" "frq.tui"]}}
99 :tasks {frq "jolt -M:frq"
100         tui "jolt -M:tui"}}

A freeq client in jolt, as glimmer components on Vidya 4719d6f nandi 20d ago	1	{:paths ["src"]
	2
	3	;; glimmer owns the reactive half (ratom, components, reconciler);
Paint frq with jvui, and take the last Rust out of the window 77963dc nandi 9d ago	4	;; glimmer-jvui paints it as jvui, which is dvui's shape written in jolt on
	5	;; SDL3 — no shared object of its own, and no Rust anywhere under the UI.
	6	;; That is the whole reason for the switch off glimmer-vidya: libvidya meant
	7	;; jolt-native's Cargo workspace, and a Cargo workspace meant a build.
	8	;;
Show the list and the conversation at once on a wide window 10a1bd0 nandi 19d ago	9	;; Our fork of glimmer, until the reconciler fixes in it land upstream. A
Keep a row's actions from wandering when its line gets an id 5b5e90e nandi 20d ago	10	;; component unmounted between a cell firing and the queued render running
	11	;; used to render anyway, into widgets the backend had already freed and
	12	;; handed out again — which took the message list apart in a busy channel.
Show the list and the conversation at once on a wide window 10a1bd0 nandi 19d ago	13	;; And replacing a native node with one of another tag left every watcher
	14	;; under it subscribed, so swapping the phone layout for the split one and
	15	;; back shredded the chat: one component's props painted onto another's
	16	;; widget.
Paint frq with jvui, and take the last Rust out of the window 77963dc nandi 9d ago	17	:deps {jolt-lang/glimmer {:git/url "https://gitlab.com/nandithebull/glimmer"
	18	:git/sha "399df371c790d690fb6e4560c3d4d7f838502857"}
	19	nandi/glimmer-jvui {:git/url "https://gitlab.com/nandithebull/jolt-native"
Take the jvui that fills the window and has glyphs for the icons 57a93bb nandi 9d ago	20	:git/sha "7d907d008b66abc03da83dc36411cf13f3085a35"
Paint frq with jvui, and take the last Rust out of the window 77963dc nandi 9d ago	21	:deps/root "glimmer-backends/glimmer-jvui"}}
A freeq client in jolt, as glimmer components on Vidya 4719d6f nandi 20d ago	22
Paint frq with jvui, and take the last Rust out of the window 77963dc nandi 9d ago	23	;; What the UI loads is SDL, and jvui declares it itself — SDL3, SDL3_ttf
	24	;; and SDL3_image, all :optional, so the layout tests run on a machine with
	25	;; none of them. Nothing of jolt-native's Cargo workspace is loaded by the
	26	;; window any more; `just tui` still takes libjolttui, which is the one
	27	;; object left and only for the terminal.
Take both native libraries from one place a66d1c1 nandi 19d ago	28	;;
Switch frq.av onto the jolt media plane, and stop shipping libjoltmoq b33f288 nandi 9d ago	29	;; Calls used to be the one thing this client could not do in jolt. The media
	30	;; half is `frq.av.plane` now — MoQ over QUIC from libmoq_ffi, Opus from
	31	;; libopus, H.264 from openh264 through a shim, and the devices from V4L2
	32	;; and ALSA. Signaling was always IRC and still lives in src/frq/av.clj.
Take both native libraries from one place a66d1c1 nandi 19d ago	33	;;
Build the desktop libraries rather than fetching a release of them 0b81161 nandi 15d ago	34	;; just run
Paint frq with jvui, and take the last Rust out of the window 77963dc nandi 9d ago	35	:jolt/native [;; The transport half of a call: MoQ over QUIC, the one
Switch frq.av onto the jolt media plane, and stop shipping libjoltmoq b33f288 nandi 9d ago	36	;; piece with no C implementation anywhere. Fetched from
	37	;; moq-ffi's release rather than built — building it means a
	38	;; 1062-crate workspace.
Bind libmoq_ffi from the object's own metadata, not from its header 9f081a1 nandi 9d ago	39	;;
Switch frq.av onto the jolt media plane, and stop shipping libjoltmoq b33f288 nandi 9d ago	40	;; The codecs are NOT in it. moq-ffi's `audio` and `video`
	41	;; features are off in every Linux artifact upstream ships,
	42	;; so Opus and H.264 come from their own C libraries below,
	43	;; which is the whole point of doing this over FFI rather
	44	;; than over a Rust facade.
Fetch libmoq_ffi again, and keep the codecs out of Rust f518938 nandi 9d ago	45	;;
	46	;; No :darwin: :systems is x86_64-linux and aarch64-linux, so
	47	;; a .dylib named here would name a file nothing fetches.
Bind libmoq_ffi from the object's own metadata, not from its header 9f081a1 nandi 9d ago	48	{:name "moq_ffi"
Bind the C libraries a call actually needs 87e5be9 nandi 9d ago	49	:linux ["libmoq_ffi.so"]}
	50
	51	;; The codecs, as C libraries rather than as somebody's
	52	;; bindings to them. Opus first; openh264 beside it when the
	53	;; video half lands.
	54	;;
	55	;; The SONAME, not the bare .so: a `libopus.so` is the -dev
	56	;; symlink and is not what a runtime closure carries. The
	57	;; loader looks these up by name in one directory, which the
	58	;; flake's `nativeAll` is.
	59	{:name "opus"
	60	:linux ["libopus.so.0"]
	61	:darwin ["libopus.0.dylib"]}
	62
	63	;; H.264, one step removed. openh264's C API is a vtable —
	64	;; `ISVCEncoder` is `const ISVCEncoderVtbl*` — and jolt.ffi
	65	;; can only call a literal C symbol, never a function
	66	;; pointer. c/frq_h264.c walks the vtable and exports flat
	67	;; symbols; this is that. libopenh264 itself comes along as
	68	;; its DT_NEEDED and is never named here.
	69	{:name "frqh264"
	70	:linux ["libfrqh264.so"]
	71	:darwin ["libfrqh264.dylib"]}
	72
	73	;; Audio devices. No :darwin — CoreAudio is the other side of
	74	;; that door and is not this library.
	75	;;
	76	;; V4L2 is deliberately absent from this list: the camera is
	77	;; ioctls against libc and the kernel, so there is nothing to
	78	;; load. See frq.capture.v4l2.
	79	{:name "asound"
	80	:linux ["libasound.so.2"]}]
A freeq client in jolt, as glimmer components on Vidya 4719d6f nandi 20d ago	81
Paint the same screens into a terminal ab83b42 nandi 17d ago	82	:aliases {:frq {:main-opts ["-m" "frq.app"]}
A freeq client in jolt, as glimmer components on Vidya 4719d6f nandi 20d ago	83
Paint the same screens into a terminal ab83b42 nandi 17d ago	84	;; The same screens in a terminal. glimmer-tui is the other backend
	85	;; for the same reconciler — libjolttui's tree ABI is libvidya's
	86	;; with cells under it — so `frq.tui` requires it after `frq.app`
	87	;; and renders the hiccup that is already written.
	88	;;
Merge origin/main, and let one jolt-native answer for both backends 1e8b590 nandi 16d ago	89	;; A path rather than a sha, like glimmer-vidya's would be if this
	90	;; alias were the way in: `just tui` builds jolt-native out of the
	91	;; flake and hands both Jolt halves over with -Sdeps, so this is for
	92	;; a checkout beside the tree and nothing else reads it.
Paint the same screens into a terminal ab83b42 nandi 17d ago	93	;;
	94	;; just tui
	95	:tui {:extra-deps {nandi/glimmer-tui
Retire the release-bumping script 4f38d68 nandi 11d ago	96	{:local/root "../jolt-native/glimmer-backends/glimmer-tui"}}
Paint the same screens into a terminal ab83b42 nandi 17d ago	97	:main-opts ["-m" "frq.tui"]}}
	98
	99	:tasks {frq "jolt -M:frq"
	100	tui "jolt -M:tui"}}