Repository navigation
Fullbleed 2.6.0a2: static components and SVG styling
Pre-release
This opt-in Python alpha brings stable 2.5.27's SVG styling fixes to component authoring. Document class rules, inherited color and named-part opacity now reach supported inline SVG descendants. Styled HTML reproduces the document PDF; unstyled HTML retains classes for host CSS.
Install in a virtual environment:
python -m pip install --pre "fullbleed==2.6.0a2"The existing experimental API is preserved: reusable styles, named parts, all 23 Bootstrap 5.0.0 component families, and to_html(include_styles=True/False). Both PDF and HTML remain static; Bootstrap JavaScript is not added. The API stays under fullbleed.ui.experimental. Older alpha/preview downloads remain unchanged, and ordinary pip resolution was verified to select stable 2.5.27. This release does not publish a Rust alpha or change Node/.NET engine versions.
- Download the runnable Bootstrap kit
- Component and styling guide
- Reusable component recipes
- Report PDF and static HTML catalog
Validation: 922 local Python tests, 1,404 native tests and 11 doctests pass. All 76 retained component PDFs match between local, Linux CI and Windows CI. The new SVG/component checks pass in PDF and browser HTML, including a negative control that reproduces the old alpha's paint bug. All 15 public platform wheels and the source archive match the validated release artifacts; a fresh public installation passes the package/contract checks, first document, 29-PDF kit and three guide blocks. Typing checks reject deliberate errors, and 24 editor completion requests pass.
The 54 SVG fixtures retain 648 matching PDFs across qualified local, release and public Windows builds. Native previews match; independent PDFium previews differ only in small order-number text below the SVG on Linux. The complete 1,662-case CSS corpus retains 1,619 PASS, 25 FAIL and 18 REFERENCE-DISPUTED verdicts. All verdicts match both baselines. Scores match stable except for oversized table pagination: that case's PDF and all three-page artifacts match published alpha 1, and its raster pixels match the reference. Relative to alpha 1, the single changed case matches the already reviewed stable translucent-rectangle perimeter. Both differences are retained in the evidence archive. The comparator's failure remains visible.
These are bounded regression checks, not exhaustive Bootstrap/browser parity, accessibility or PDF conformance. The dedicated PDF/VT validator remains unconfigured. Independent author feedback and stable API promotion remain open.