fix(assets): land the transparent rasters that never reached main #96

Merged

coilysiren merged 4 commits from feat/sirens-deep-mark into main

2026-08-11 04:08:46 +00:00

coilyco-ops commented

2026-08-11 04:07:36 +00:00

Owner

Lands work that was already reviewed and merged, but into the wrong place.

What happened

#94 was stacked on feat/sirens-deep-mark rather than on main, as agreed. #84 merged that branch into main before #94 merged into it. So #94 landed on a branch that had already been consumed, and its content never reached main.

Both pull requests show as merged. The result is that main carries the mark with the raster problem #94 was written to fix.

What is on main right now

assets/sirens-deep.png and assets/sirens-deep-lockup.png bake the ink field into a full square tile. Sampled at the corner, both read (20, 17, 27, 255), fully opaque. They cannot drop onto any surface without dark corners, and they do not match the org marks in the website repository, three of which are transparent outside the ring.

What this brings over

The four stranded commits, unchanged and already reviewed:

the original transparency fix
its revert, which is what split the work into #94 in the first place
the re-application on the follow-up branch
the #94 merge commit

Net effect on main: both PNGs become transparent outside a circle of r 169.9, just inside the ring's outer edge at 171.5, with fully transparent pixels zeroed. The SVGs keep their opaque field and are untouched. docs/assets.md gains the six-line note on regenerating them.

Nothing new is authored here. If you would rather see a clean single-commit version instead of the revert-and-reapply pair, say so and I will squash it.

Same thing happened next door

coilysiren/website has the identical ordering problem, with more missing: its main has neither the water mark, the -lockup rename, nor the transparent rasters. Companion catch-up: coilysiren/website#108.

Lands work that was already reviewed and merged, but into the wrong place. ## What happened #94 was stacked on `feat/sirens-deep-mark` rather than on `main`, as agreed. #84 merged that branch into `main` **before** #94 merged into it. So #94 landed on a branch that had already been consumed, and its content never reached `main`. Both pull requests show as merged. The result is that `main` carries the mark with the raster problem #94 was written to fix. ## What is on main right now `assets/sirens-deep.png` and `assets/sirens-deep-lockup.png` bake the ink field into a full square tile. Sampled at the corner, both read `(20, 17, 27, 255)`, fully opaque. They cannot drop onto any surface without dark corners, and they do not match the org marks in the website repository, three of which are transparent outside the ring. ## What this brings over The four stranded commits, unchanged and already reviewed: - the original transparency fix - its revert, which is what split the work into #94 in the first place - the re-application on the follow-up branch - the #94 merge commit Net effect on `main`: both PNGs become transparent outside a circle of r 169.9, just inside the ring's outer edge at 171.5, with fully transparent pixels zeroed. The SVGs keep their opaque field and are untouched. `docs/assets.md` gains the six-line note on regenerating them. Nothing new is authored here. If you would rather see a clean single-commit version instead of the revert-and-reapply pair, say so and I will squash it. ## Same thing happened next door `coilysiren/website` has the identical ordering problem, with more missing: its `main` has neither the water mark, the `-lockup` rename, nor the transparent rasters. Companion catch-up: coilysiren/website#108.

coilyco-ops added 4 commits

2026-08-11 04:07:36 +00:00

fix(assets): make the mark rasters transparent outside the ring a1cba1a11c

Both PNGs baked the ink field into a full square tile, so the mark could
not drop onto any surface without dark corners. They are now transparent
outside a circle of r 169.9, just inside the ring's outer edge at 171.5,
which makes each render a coin.

This follows the convention already used by the org marks in the website
repository, where three of the four are masked this way.

The SVGs keep their opaque field so they still stand alone. docs/assets.md
records how to regenerate.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>

Revert "fix(assets): make the mark rasters transparent outside the ring" b8672f5c3a

Reverts a1cba1a so the raster background change can be reviewed on its own.
It returns on a follow-up branch stacked on this one, leaving this pull
request to cover the mark itself.

The rasters here still bake the ink field into a square tile, which does
not match the org marks in the website repository. That is the thing the
follow-up fixes, so land both.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>

fix(assets): make the mark rasters transparent outside the ring

ci / test (pull_request) Successful in 34s

Details

ci / publish-echo-image (pull_request) Has been skipped

Details

eb244b36e6

Both PNGs baked the ink field into a full square tile, so the mark could
not drop onto any surface without dark corners. They are now transparent
outside a circle of r 169.9, just inside the ring's outer edge at 171.5,
which makes each render a coin.

This follows the convention already used by the org marks in the website
repository, where three of the four are masked this way. gaming.png is the
exception there and was the wrong model to follow.

The SVGs keep their opaque field so they still stand alone. docs/assets.md
records how to regenerate.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>