Mgen+ 1mn Term (font resource for extensions)

日本語

Extension ID: shirokuma-library.font-mgenplus-1mn

A general-purpose font resource extension that supplies the Mgen+ 1mn Term font to other extensions. It has no UI and no commands of its own (contributes is empty, activationEvents is empty, and main is a no-op), and it does not install the font into the OS. Its sole purpose is to let a consuming extension load the bundled TTF from its own webview via vscode-resource.

Mgen+ 1mn Term is a modified version of Mgen+ 1mn, refitted for terminal use: the symbols a terminal counts as one column wide are brought under the width at which a renderer squashes them sideways. Ink wider than 1.15 cells is scaled down to exactly that width, and any ink that reaches past its cell is centred on the cell, so the overhang is the same on both sides and never exceeds 38 units of a 512-unit cell (about 7.5%). A symbol may therefore touch its neighbour slightly; what it never does is reach the squashing threshold. Advance widths are not part of that: the cell grid does not move, and this file's hmtx — every advance and every left side bearing, all 18,109 glyphs of the Regular face — is byte-identical to the #291 release that preceded this one (measured). The ratio itself — what it means, its 1.5 ceiling and why 1.15 — is defined once, in scripts/font-fit/manifest.json's note.

The two axes are scaled by different factors (since #293): the width goes to s, the height to s × 1.3 where the face allows it. Scaling ① down to the cell took its height to 588 units as well — below this face's own cap height of 748 (measured) — so the symbol read as small because it was short, and horizontally there was nothing left to give. The growth is clamped, per glyph, against a limit measured out of this face: the top and bottom edges of its own 漢 (U+6F22) — y = −78..843 in the Regular face, −95..847 in the Bold — and each glyph's stretch is capped so the stretched glyph stays inside them. What is bounded is the glyph's reach, not its height: the conversion keeps each glyph's vertical centre, so capping the height alone would still let a glyph whose centre sits off the kanji's push past the line. The statement that is true of these bytes is therefore a conditional one: the refit never pushes a glyph past that kanji's top or bottom, and a glyph already past it at equal scaling keeps the previous release's shape, unchanged to the unit (measured: 0 glyphs of either face newly reach outside). It is not claimed that nothing here reaches further above or below the line than 漢 — this typeface draws symbols taller than 漢 to begin with, and 141 of the 1001 refitted glyphs in the Regular face are outside its box at equal scaling already. The clamp's floor is 1.0, so not every refitted glyph gets taller: measured on the shipped bytes, 857 of the 1001 refitted glyphs grew in the Regular face and 899 of 1020 in the Bold — the rest had no room left at equal scaling and did not move a unit. 1.3 is a chosen value, arrived at by looking at five bakes; it is not derived from the cap-height ratio (748 / 588 = 1.272) it happens to sit near. Its meaning, its measured limit and the current default are defined in the same note. ASCII and Latin-1 are out of scope — a renderer does not rescale those, and refitting them would rework the body text — as are box drawing, block elements and private use; the exact scope is stated in font-packages/README.md. Nothing else about the outlines was touched. The full chain — which archive it came from, what did the converting, and the hash of what came out — is in "Provenance", below.

Bundled font

Two faces of one family, each a separate static TTF converted from the corresponding face of the same upstream distribution.

family is the measured name table value, not an assumption — see "family name" below. Each weight is declared as a single value, never a range: these are static fonts, and a range is a claim to cover the weights in it. A face declaring 100 700 is selected as the real thing for xterm's bold cells and painted at regular weight, so declaring the range suppresses synthetic bold instead of producing bold. That is exactly why the real 700 face exists here: with it declared, 700 is covered by glyphs that are actually bold rather than by a claim (#249).

Every weight above is spelled out rather than omitted. What an omission would cost differs by which declaration shape it sits in, and the two rules are stated once, for all packages, in font-packages/README.md — the canonical description of the contract.

See LICENSE.txt for the copyright notices and the full license text, and fonts/SIL_Open_Font_License_1.1.txt for the license text exactly as the distribution ships it.

Copyright and attribution (three parties)

Mgen+ fills the glyphs M+ OUTLINE FONTS does not cover with glyphs from Source Han Sans, so three parties are credited. Which source states what is kept separate below, because two of the three are named as copyright holders in so many words and the third is not.

Copyright holders, quoted verbatim

The bundled TTFs' own name table, nameID 0 (copyright) — identical in the Macintosh (platform 1, encoding 0, language 0x0) and Windows (platform 3, encoding 1, language 0x409) records, and byte-for-byte identical between the regular and the bold face (measured), so one quotation covers both. The conversion rewrites the family and style names but carries nameID 0 through byte-identically (re-measured on the placed files, #275), so this is still the machine-readable copyright field OFL §2 speaks of, unchanged from upstream:

[Source Han Sans]
Copyright © 2014, 2015 Adobe Systems Incorporated (http://www.adobe.com/), with Reserved Font Name 'Source'.

[M+ OUTLINE FONTS]
Copyright(c) 2015 M+ FONTS PROJECT

The distribution's own README (fonts/README_MgenPlus.txt, section 「ライセンスと著作権について」), which ships here unmodified, names the same two holders in Japanese:

・フォントデータに含まれる、源ノ角ゴシック由来の文字グリフの著作権は Adobe が所有しています。
・フォントデータに含まれる、M+ OUTLINE FONTS 由来の文字グリフの著作権は M+ FONTS PROJECT が所有しています。

Note the Reserved Font Name 'Source' (OFL §3) in the Adobe line. This package is a modified version, so that clause is squarely in scope — and it is satisfied, because the name the faces declare (Mgen+ 1mn Term) contains no reserved name. 'Source' is the only RFN declared anywhere in this package's license texts; the measurement is recorded under "Provenance", below. Nothing derived from this font may be distributed under a name containing 'Source'.

The third party: 自家製フォント工房 (Jikasei Font Studio)

Neither the name table nor the README's copyright section carries a verbatim copyright line for 自家製フォント工房. It is credited here as the author and distributor of the composed font — the work of selecting, merging and fitting the two upstream glyph sets into Mgen+ — which the distribution states in a separate section, 「■ 頒布元」 (fonts/README_MgenPlus.txt, line 90):

自家製フォント工房
http://jikasei.me/

Read that as an attribution of authorship and distribution, not as a quoted copyright notice.

Licenses, and a note on the links

The font is licensed under the SIL Open Font License, Version 1.1 (full text in LICENSE.txt, whose OFL body is byte-identical to the distribution's own copy at fonts/SIL_Open_Font_License_1.1.txt). The M+ OUTLINE FONTS licenses covering the upstream glyphs ship at fonts/mplus-TESTFLIGHT-059/LICENSE_E and .../LICENSE_J, under their original archive-internal path so their subject stays unambiguous.

Project links, for convenience: Mgen+ http://jikasei.me/font/mgenplus/, Source Han Sans https://github.com/adobe-fonts/source-han-sans, M+ FONTS PROJECT http://mplus-fonts.osdn.jp/. These are current addresses, not quotations from the distribution — its 「■ 改変元」 section cites http://mplus-fonts.sourceforge.jp/ and an Adobe store product page (http://store1.adobe.com/cfusion/store/html/index.cfm?store=OLS-JP&event=displayFontPackage&code=1967), both of which have since moved.

Scope limitation: it does not work in editor.fontFamily / terminal.integrated.fontFamily

A font supplied this way cannot be used by writing the font name Mgen+ 1mn Term into VS Code's settings (editor.fontFamily or terminal.integrated.fontFamily). This is a common misconception, so it is worth stating outright — it is also why this extension's displayName / description say it is "a resource for extension developers, not a font selectable from ordinary settings such as editor.fontFamily".

• To render with this font, the consuming extension must inject the @font-face itself into its own webview HTML/CSS. That @font-face definition exists only within the scope of that one webview; it is entirely absent from the CSS of VS Code's main window (the editor proper — Monaco — and the standard integrated terminal).
• editor.fontFamily is a setting for the editor that renders in the main window, so an @font-face injected for a webview never reaches it. Naming such a font behaves exactly like naming a font that does not exist: no error is raised, and VS Code silently falls through to the next fallback. terminal.integrated.fontFamily (VS Code's standard integrated terminal) is likewise rendered in the main window and is equally out of reach.
• This mechanism is therefore usable only when the consuming extension has its own webview and implements the borrowing steps itself. If what you want is this font in the editor or the standard terminal, the only way to get it there is to install the font into the OS.

How other extensions use this font (the convention)

Note: this is not an official VS Code mechanism for sharing fonts between extensions. No such official mechanism exists. What follows is an unofficial, homegrown convention, and it assumes the exact shape of this extension's ID and of the providedFont contract field in package.json. If this extension later changes the file layout, the ID, or the contract field, consumers will break silently.

The contract itself is the providedFont field in package.json:

"providedFont": {
  "family": "Mgen+ 1mn Term",
  "file": "fonts/MgenPlus1mnTerm-Regular.ttf",
  "weight": "400",
  "faces": [
    { "file": "fonts/MgenPlus1mnTerm-Regular.ttf", "weight": "400" },
    { "file": "fonts/MgenPlus1mnTerm-Bold.ttf", "weight": "700" }
  ],
  "license": "OFL-1.1"
}

faces lists every face of the family, in declaration order; a consumer that reads it injects one @font-face per entry. The flat file / weight pair is kept as a shorthand for the single 400 face and stays byte-equal to the first entry, so a consumer written against the older one-file contract keeps working unchanged (it simply gets no bold face and falls back to synthetic bold, as it did before this package shipped one).

How a consumer reads that — the two declaration shapes, which one wins, the normalization and degradation rules, and the borrowing steps with code — is font-packages/README.md in the same repository. That file is the single canonical description of the contract; this section records only what this package declares. tmux-opener's src/font-registry.ts / getHtml are a working implementation of the consumer side.

Provenance (upstream archive → conversion → output)

The two TTFs this package ships are conversion output, not the upstream bytes. Until #275 they were placed byte-for-byte as distributed, and the two-column sha256 table that rule comes with was the whole proof of provenance: hash the archive entry, hash the placed file, and the two columns agreeing said "identical to what upstream published". That proof is not available for a modified file — the thing it would be compared against is the upstream TTF, which this repository does not ship. It is replaced by a three-link chain, each link measured: where the input came from, what turned it into the output, and what the output hashes to. The chain is a different claim, not a cheaper version of the old one: it records how these bytes were made, and nothing in it can be machine-checked against upstream.

The accompanying license and README documents are still placed byte-for-byte and keep the two-column rule unchanged — they were never converted.

1. Upstream archive (the input)

• Distribution: Mgen+ 1.059.20150602, official archive mgenplus-1m-20150602.zip
• URL: https://ftp.iij.ad.jp/pub/osdn.jp/users/8/8592/mgenplus-1m-20150602.zip (IIJ mirror of OSDN; upstream project page http://jikasei.me/font/mgenplus/)
• Archive: 29,504,415 bytes, Last-Modified: Wed, 03 Jun 2015 14:17:54 GMT, sha256 fae75318098697f165d74c5c4297805dfcfa335e09b0c31175ed299663c4c009 (the archive itself is not committed)
• Extraction: python3 stdlib zipfile (this host has no unzip/7z/bsdtar). Which of the two candidate archives carried 1mn was not assumed: both users/8/8592 (-1m-) and users/8/8593 (-1-) were listed, and mgenplus-1mn-regular.ttf is in the -1m- one, so the 42.9MB archive was never needed.

The entry hashes below are the same values scripts/font-fit/manifest.json carries as srcSha256, and run.sh re-checks every one of them against the --src-dir it is given before it writes a single byte — a mixed-up input fails there rather than producing a plausible-looking output.

2. Conversion (what turned input into output)

• Pipeline: scripts/font-fit/run.sh in this repository, driven by scripts/font-fit/manifest.json (this package is one entry in it; membership of that file is what makes a package a converted one). fit.py does the refit, verify.py gates the result, dump-widths.mjs produces the cell-width table from @xterm/addon-unicode11's own provider — a hand-written width table is forbidden.

• Which converter, identified by content rather than by commit — introduced by issue #275, and pinned by the sha256 of each file that determines the output:

  pipeline file	sha256
  scripts/font-fit/fit.py	c634881d418a59d68f1892db4d35f4fc66780a71cff277448cbef91479470bf7
  scripts/font-fit/instance.py	86240ac88d1dfa4527ebaa1e7e97caf6c656e97cd04833d12b21149d2fefe2af
  scripts/font-fit/verify.py	5c7d58d6a8ff4e4d248990c4bc39bcf379faee2fec9303ea3966c6ce1c5d9cf5
  scripts/font-fit/dump-widths.mjs	4813198fba79edb3d3086aa07b8808183a362406b1a1da9c7d614c5b90633a03
  scripts/font-fit/run.sh	4966ccca260ded4b1d7aa94fa71bc9802956363e9a646c788166ee530ab8c55a
  scripts/font-fit/manifest.json	0fed27ae2614fd0cbe570cde156bb4ed5ff32fe7a0a7c814b97d71acaf49629b

  Why not a commit SHA. The first draft of this section recorded one, and it was false within the same PR: the converter was then changed (a bold face was flagging itself Regular), the output moved, and the recorded SHA still pointed at the older converter — checking it out and re-running produced a different hash than the one recorded three lines below. That is not a slip that better care avoids: a git SHA is content-addressed, so a file cannot name the commit that contains it. Recording one means either splitting the TTFs, the checksums.json record and this table across two commits — the one coupling R13 exists to prevent — or leaving a placeholder for someone to substitute, which ships silently when they forget. Hashing the scripts has neither problem: the values are knowable while writing, they survive rebases and rewrites, and they identify the converter directly instead of via a commit that also carries unrelated changes. run.sh additionally appends -dirty to the commit it prints when the working tree differs from HEAD, so a provenance line generated from uncommitted edits can no longer be mistaken for one generated from a clean checkout.

• Measured tool versions (printed by run.sh at the start of the run, not copied from documentation):

  tool	version
  fontTools	4.63.0
  Python	3.14.4
  @xterm/addon-unicode11	0.10.0-beta.288

• Fixed environment: PYTHONHASHSEED=0, LC_ALL=C.UTF-8, and fontTools saved with recalcTimestamp=False (a recalculated head.modified would make every run produce different bytes and the output hashes below meaningless).

• The fitRatio the run was driven by: 1.15 (manifest.json's top-level default; this package declares no override). Ink wider than A × 1.15 — 588.8 units of the 512-unit cell — is scaled down to it; wider ink is what "scaled" counts below. The value's meaning and its 1.5 ceiling are defined in scripts/font-fit/manifest.json's note and not restated here.

• The vstretch the run was driven by: 1.3 (manifest.json's top-level default; this package declares no override), with the vertical limit measured from each face rather than declared — the top and bottom edges of 漢 (U+6F22), y = −78…843 in the Regular face and −95…847 in the Bold, as run.sh printed them. The clamp bound the stretch on 357 of the 1001 refitted glyphs in the Regular face and 284 of 1020 in the Bold; of those, 144 and 121 respectively had no room left at equal scaling and were left completely unmoved (the clamp's floor is 1.0). So the glyphs that actually grew taller are 857 of 1001 and 899 of 1020 — the vertical change is not universal, and the counts are the honest way to say so. Of the refitted glyphs, 141 (Regular) / 121 (Bold) already reached outside the 漢 box at equal scaling, and none of either face reaches outside it that did not already (measured on the shipped bytes against a vstretch = 1.0 bake of the same upstream). The horizontal side is untouched by this: hmtx is byte-identical to the pre-#293 release across all 18,109 (Regular) / 18,106 (Bold) glyphs (measured), and the widest in-scope ink is unchanged at 588.

• What the run changed, per face — measured, and reported by run.sh:

  face	glyphs refitted	of which moved only / scaled	already fitting
  Regular	1001	250 / 751	1344
  Bold	1020	232 / 788	1327

  "Moved only" grew and "scaled" shrank against the 0.96-ratio release: at 1.15 the room is wider than the cell, so a glyph whose ink already fits within 588.8 is re-centred at its original size rather than shrunk.

• All three gates passed on both faces (verify.py, run against the upstream file the converter actually read, at the declared ratio and stretch): contract ① — in-scope code points outside the declared band (ink ≤ A × 1.15, centred when it overhangs, advance == A): 0, with the widest in-scope ink landing on the declared room exactly (Regular 1024 → 588, Bold 1032 → 588); contract ② — body glyphs whose outlines changed: 0 of 13,101; contract ③ — refitted glyphs whose top and bottom edges differ from what the declared vstretch and the measured wall (this face's own 漢 y reach) imply: 0 of 1001 (Regular) and 0 of 1020 (Bold), checked as an equality in both directions, so bytes baked at a different stretch would fail whichever way they differed — and the half of that gate which recomputes no formula (③c) reports 0 glyphs newly outside the wall on both faces. Together those say the refit reached every glyph it was asked to reach, went exactly as far as the declaration and no further, and touched nothing else. In-scope is load-bearing rather than a hedge: the pipeline deliberately excludes ASCII / Latin-1, the BMP private use area, box drawing and block elements, and everything not width 1 (font-packages/README.md, "Terminal width conventions"), so a Latin-1 glyph overhanging its advance is outside this count by design.

• The family and style names in the name table are rewritten (Mgen+ 1mn → Mgen+ 1mn Term); nameID 0, the copyright field, is carried through byte-identically (re-measured on the output, quoted above).

• Reproducibility, as measured, not as promised: re-running the pipeline on the same machine, from the same input, with the tool versions above, produced byte-identical output — the sha256s below. Nothing is claimed about a different fontTools or Python: that was not tested, and neither was the effect of environment variables beyond the three fixed above.

3. Output (what is shipped)

Measured on the placed files, and recorded in scripts/font-fit/checksums.json as the same values — from #276 npm test compares the two, so a TTF and its record can only move together. Since #291 the record carries the fitRatio beside each sha256 and since #293 the vstretch as well, and npm test compares both against the manifest too.

Reserved Font Name check (measured)

OFL 1.1 §3 forbids a Modified Version from carrying a Reserved Font Name, so a converted package must establish which names are reserved before choosing its own. Measured over every license text this package ships:

Superseded (#288). This grep is no longer the procedure. Run scripts/font-fit/check-license.py instead — the steps are in font-packages/README.md ("Reserved Font Name check"). The command below cannot reach the name table, where a package can declare a reserved name that appears in no file at all (measured on Source Han Code JP, #285); it is also blind to license documents that are not UTF-8. The table below is kept as the record this grep produced, not as instructions: it predates the name-table stage, so any fonts/*.ttf#0:nameID N row the tool reports is missing from it. Adding those rows is #289's work.

grep -rniI 'reserved font name' \
  font-packages/font-mgenplus-1mn/LICENSE.txt \
  font-packages/font-mgenplus-1mn/fonts/

Two flags and the path list are all load-bearing, and each of them was got wrong once in #275:

• -i — OFL prose writes the phrase title-case (Reserved Font Name) but a section heading may shout it, and a case-sensitive grep silently drops those lines. That is how the first version of this table came to omit LICENSE.txt:62, the very line stating this package's conclusion.
• The path list, not --include='*.txt' — the M+ license documents this font derives from ship as LICENSE_E / LICENSE_J, with no extension, so a *.txt filter reads right past them while the prose above claims to have searched every license text. Naming LICENSE.txt plus the whole fonts/ directory covers them. (They contain no reserved font name — measured, 0 hits each — so the verdict was never in doubt; the search that produced it was.)
• -I — fonts/ also holds the TTFs, and without it grep reports Binary file … matches and clutters a table that is supposed to be reproducible line for line.

A verdict is only as good as the search that produced it, so the command is written down and the table below lists every line it returns, not just the interesting ones:

The name-table rows the grep could not reach (#289, measured with scripts/font-fit/check-license.py). These are not corrections to the table above — the name they declare is the one it already records — they are evidence the text-only procedure was structurally blind to, added here because the superseded note above says they were missing:

The tool's own verdict on the shipped bytes, verbatim: 'Mgen+ 1mn Term': どの予約名も含まない → OK （照合した予約名 9 件）, exit 0. Its text hit lines carry different line numbers from the table above because LICENSE.txt has been rewritten since that grep was run; the table stays as the record of what that command returned, which is what it is for.

'Source' is the only Reserved Font Name declared here, and neither Mgen+ nor 1mn nor Term contains it, so Mgen+ 1mn Term is a permitted name for this modified version. (For contrast, font-plemoljp-console declares "PlemolJP" as reserved — which is one reason that package is not converted at all.)

Accompanying documents (byte-for-byte, two-column rule)

Both columns are measured, in that order: the left one comes off the extracted file in a scratchpad directory before anything was copied into the repository, the right one off the committed file afterwards. The rule — and why a single self-referential column would verify nothing — is stated once for every package in font-packages/README.md.

They ship at their original archive-internal paths: the OFL 1.1 text that covers Mgen+ itself, the distribution's own README_MgenPlus.txt (which states the version and what the font is composed of), and the upstream M+ OUTLINE FONTS licenses that Mgen+ derives from, kept under their mplus-TESTFLIGHT-059/ directory rather than flattened, so their subject stays unambiguous. They describe the upstream font; this README is where the modification is described.

family name (measured, not assumed)

Re-measured in #275 on the converted files, since those are what ships: the conversion rewrites the name table, so a family string carried over from the pre-conversion measurement would have been a claim about bytes this package no longer contains. Read from the name table directly (nameID 1 / 2 / 4 / 6, and nameID 16 / 17 where present) — every face this package ships is measured, because the whole borrowing scheme rests on the two of them naming the same typographic family:

nameID 16 is the typographic family and it is identical across the two faces — upstream designed them as one family and the conversion preserves that, which is what makes pairing them under a single CSS font-family legitimate (and their metrics agreement, below, an expected property rather than a coincidence). The injection does not depend on it — @font-face declares a family name rather than reading one, and CSS font matching never consults a file's internal names; font-packages/README.md states that mechanism in full. This table is evidence about the fonts, not about the mechanism.

Mgen+ 1mn Term is the value this extension's providedFont.family carries, and it must stay identical to the consumer-side registry entry (tmux-opener's FONT_REGISTRY). It is recorded here as a measurement, not as a name chosen and then assumed to have been applied. A face added or replaced here gets its own column, measured — the table is the record of that measurement, not a substitute for making it (#249).

Note for consumers: the pre-conversion name Mgen+ 1mn is not what these files declare any more. tmux-opener keeps it working by registering it as an alias and injecting an alias-named @font-face over the same files, so an existing settings.json renders with the converted bytes; that is a consumer-side compatibility layer, not a property of this package.

A note on fontconfig, which the pre-#275 version of this table was taken from: fc-scan reports spacing as dual (90), not mono (100), for these faces — normal for a CJK font, and precisely why fontconfig's spacing is not the monospace evidence. The evidence is test/font-metrics.test.js, below.

Monospace metrics (measured)

npm test in the repository root (test/font-metrics.test.js) covers both files and asserts this on every run:

• unitsPerEm 1024; every ASCII printable codepoint 0x21–0x7E is present in cmap and shares one advance A = 512.
• The fullwidth samples 一 / 漢 / あ / ん / ア each advance 1024 = 2A.
• The two faces agree with each other: same unitsPerEm, same A (#249). Each face being monospaced on its own is not enough — the terminal measures its cell from one family and paints both faces into it, so bold at a different scale would drift out of column while every per-file check still passed.

The values above are hmtx advances of the default instance, which for these static faces is the whole font. These particular values are unchanged by the conversion — but plenty of other advances are not, and the distinction is worth stating exactly, because a looser version of this sentence shipped in #275 claiming the conversion never touches an advance at all, which is false:

• A width-1 code point's advance is set to the cell width, deliberately, and for nearly all of them that is a change. That is half the point of the refit: upstream draws nearly all of these at the full-width advance 1024, and the conversion sets them to the cell width. Measured on the shipped files against the upstream ones, 997 glyphs in the regular face and 1016 in the bold went 1024 → 512, and 1024 → 512 is the only transition either file contains. Both numbers were re-measured on the #291 bytes and are unchanged — fitRatio moves ink, never advances, so the hmtx advances are byte-identical to the 0.96-ratio release. The hmtx table as a whole is not: each entry is (advanceWidth, lsb), and moving a refitted glyph's ink moves its lsb, so against that release 875 entries in the regular face and 935 in the bold differ while 0 advances do (measured with fontTools, #292).
• A body glyph's advance is not changed, and neither is its outline — that is contract ② in "Provenance", verified glyph by glyph.
• The cell grid does not move: A stays 512 and full-width glyphs stay 1024, which is why the numbers in this section still hold and why converted text sets in the same columns.

Conflating the second and third points with "no advance changed" is the mistake that produced the false sentence. Contract ② is about the glyphs the refit did not select; a selected glyph changes in outline, and in advance too whenever upstream drew it full-width — which is nearly always, but not always. Measured: 1001 glyphs were refitted in the regular face and 997 of those changed advance (1020 and 1016 in the bold); the 4 per face left over — ↑ U+2191, ↓ U+2193, ↕ U+2195, plus ȿ U+023F in the regular and Ϡ U+03E0 in the bold — are glyphs upstream already drew one column wide, so the refit re-centred their ink and left the advance where it was.