Repository navigation
fix(theme): restore symbol coverage lost when M+ 1p was repointed - #259
Merged
Merged
Conversation
d93e7bf repointed `M+ 1p Fallback` from fonts/mplus-1p-regular-fallback.ttf to GEM_FONTS_DIR/mplus1p-regular-fallback.ttf so that hard line breaks keep the body font's line height. The two copies are not nested — neither is a superset of the other: local copy 6726 codepoints, 1243 of them in no other font in the chain bundled copy 6982 codepoints, 1504 of them absent from the local copy The bundled copy adds U+000A, U+200B, U+202F, U+2060 and a wider CJK range. The local copy carries IPA extensions and a set of dingbats, among them U+2718 HEAVY BALLOT X. Repointing the entry silently dropped that 1243 and left the file in fonts/ shipped but unreferenced. 66df4a2 fixed the list markers by moving them onto glyphs JetBrainsMono maps directly, which is the right fix there because the marker has its own font family. Document body text has no such escape: its chain is body -> M+ 1p Fallback -> Noto Sans -> Noto Emoji, and a spec that writes U+2718 in a table now gets WARNING: Could not locate the character `✘' (✘) in the following fonts: body, M+ 1p Fallback, Noto Sans, Noto Emoji which is fatal for any build using --failure-level=WARN, as riscv-cheri does. U+2714 on the same line still resolves, because the bundled copy happens to map it — so the pair renders inconsistently even when the build is allowed to pass. Register the local copy under a second name and append it to the end of the fallback chain. Nothing is removed and nothing is reordered: the bundled copy still comes first and still decides line feed and CJK, and the added entry is reachable only for codepoints no earlier font provides. Verified in ghcr.io/riscv/riscv-docs-base-container-image with a document containing a U+2714/U+2718 table, a `+` hard-break block and three nested list levels, built with the same flags riscv-cheri uses: * before: 1 warning, exit 1; after: 0 warnings, exit 0 * U+2718 moves from a missing-glyph box to mplus-1p-regular; U+2714 is unchanged, still served by the bundled copy * hard-break baseline advances stay at 13.980/13.979pt, identical before and after and unchanged from 66df4a2 * list markers still render in JetBrainsMono-Regular; no new font is pulled into the marker chain * of 28 text spans, only the 3 U+2718 spans differ Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Signed-off-by: Bill Traynor <wmat@riscv.org>
This was referenced Sep 1, 2026
Tariq Kurd (tariqkurd-repo)
pushed a commit
to riscv/riscv-cheri
that referenced
this pull request
Sep 2, 2026
Supersedes #1204. ## Why The PDF builds are failing on any PR that picks up the `docs-resources` bump to `d93e7bf` ([failing run](https://github.com/riscv/riscv-cheri/actions/runs/33456702369/job/99698095218?pr=1204)). All four PDF targets fail with `Error 1` and no error message, only warnings: ``` asciidoctor: WARNING: Could not locate the character `⚫' (⚫) in the following fonts: JetBrainsMono, M+ 1p Fallback, Noto Sans, Noto Emoji asciidoctor: WARNING: Could not locate the character `✘' (✘) in the following fonts: body, M+ 1p Fallback, Noto Sans, Noto Emoji ``` There is no error — the Makefile passes `--failure-level=WARN`, which promotes these to build failures. `d93e7bf` repointed the `M+ 1p Fallback` theme entry from the copy of M+ 1p in `docs-resources/fonts` to the copy bundled with Asciidoctor PDF, to fix hard-break line height. The two copies have disjoint coverage — the bundled one maps `U+000A` and more CJK, the local one maps ~1240 codepoints the bundled one lacks — so that change removed the only source of `U+26AB` and `U+2718` in the chain. ## What this picks up * **riscv/docs-resources#258** — moves the `disc` and `circle` list markers off `U+26AB` onto `U+25CF`, which JetBrainsMono maps directly, so the marker resolves in its own font instead of entering the fallback chain. * **riscv/docs-resources#259** — appends the local copy of M+ 1p to the end of the fallback chain, restoring `U+2718`. Nothing is removed or reordered: the bundled copy still comes first and still decides line feed and CJK. `U+2718` is used in the privileged integration table in `src/cheri/riscv-priv-integration.adoc`. `U+2714` on the same rows kept resolving, because the bundled copy happens to map it, so the pair rendered inconsistently even before the build began to fail. ## Verification Run locally against `ghcr.io/riscv/riscv-docs-base-container-image:latest`, the image CI uses: * `make -j$(nproc)` — the same command the workflow runs — exits **0** with **zero warnings**, building all 15 artifacts, including the four PDFs that failed in the run above. * Since the Makefile passes `--failure-level=WARN`, exit 0 is itself proof that no warning was emitted. * In the resulting 186-page `riscv-cheri.pdf`, all 19 check/cross glyphs render in `mplus-1p-regular` (13 `✔`, 6 `✘`). The six `✘` were missing-glyph boxes before. ## Note on #1204 #1204 bumps to `d93e7bf`, the commit that introduced the regression, so it should be closed rather than merged. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Signed-off-by: Bill Traynor <wmat@riscv.org> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
d93e7bfrepointed theM+ 1p Fallbackcatalog entry fromfonts/mplus-1p-regular-fallback.ttftoGEM_FONTS_DIR/mplus1p-regular-fallback.ttf, so that hard line breaks keep the body font's line height.The two copies are not nested — neither is a superset of the other:
fonts/copyThe bundled copy adds
U+000A,U+200B,U+202F,U+2060and a wider CJK range — all good reasons to keep it first. The local copy carries IPA extensions and a set of dingbats includingU+2718 HEAVY BALLOT X. Repointing the entry silently dropped those 1243 codepoints and leftfonts/mplus-1p-regular-fallback.ttfshipped but unreferenced.66df4a2fixed the list markers by moving them onto glyphs JetBrainsMono maps directly. That is the right fix there, because the marker has its own font family. Document body text has no such escape — its chain isbody → M+ 1p Fallback → Noto Sans → Noto Emoji— so a spec that writes✘in a table now gets:which is fatal for any build using
--failure-level=WARN. This is currently breakingriscv-cheriCI on every PR that picks up the submodule bump (e.g. riscv/riscv-cheri#1204). Note✔on the same table line still resolves, because the bundled copy happens to map it — so the pair renders inconsistently even where the build is allowed to pass.Change
Register the local copy under a second catalog name and append it to the end of the fallback chain.
Nothing is removed and nothing is reordered. The bundled copy still comes first and still decides line feed and CJK; the added entry is reachable only for codepoints no earlier font in the chain provides. Placing it last is deliberate — inserting it mid-chain would pull
☺ ♠ ♣ ♥ ♦away from Noto Sans/Noto Emoji.Verification
Built in
ghcr.io/riscv/riscv-docs-base-container-image:latestwith a document containing a✔/✘table, a+hard-break block and three nested list levels, using the same flagsriscv-cheripasses:✘moves from a missing-glyph box tomplus-1p-regular;✔is unchanged, still served by the bundled copy66df4a2JetBrainsMono-Regular; no new font enters the marker chain✘spans differ🤖 Generated with Claude Code