fix(markup): recognize issue references next to CJK punctuation by b84330808 · Pull Request #40097 · go-gitea/gitea

b84330808 · 2026-10-30T18:31:08Z

Issue references next to non-ASCII characters, common in Chinese and Japanese text, are ignored: 修复#1, （#2） or Closes #3， create no link or cross reference.

Like on GitHub, any non-ASCII character may now precede #123, !123 or owner/repo#123, and anything but a word character may follow it. Dead trailing-space handling in the reference and mention search is removed.

I used Claude Code to write this change.

References like "修复 go-gitea#1。" or "（go-gitea#2）" were neither rendered as links nor picked up as cross references or closing keywords, because the patterns only accepted ASCII delimiters around the reference. Assisted-by: Claude Code:claude-opus-5-5

wxiaoguang · 2026-10-30T18:34:47Z

Build the delimiter class from all non-ASCII \p{P} characters, so quotes like “go-gitea#4” and other full-width punctuation are covered too. ASCII punctuation keeps its existing rules ("/home/gitea/go-gitea#1234" is still not a reference). Co-Authored-By: Claude Opus 5.5 <[email protected]>

wxiaoguang · 2026-10-30T19:08:46Z

Why not simply use \p{P}?

b84330808 · 2026-10-31T05:06:42Z

Plain \p{P} also matches ASCII punctuation like /, #, & and ., so existing cases such as /home/gitea/#1234 start rendering as references (the current tests fail with it). So 03898af builds the class from the non-ASCII part of \p{P} plus U+3000: that covers 「」, “”, 。，、 and so on, and leaves the ASCII rules as they were.

wxiaoguang · 2026-10-31T08:08:56Z

Plain \p{P} also matches ASCII punctuation like /, #, & and ., so existing cases such as /home/gitea/#1234 start rendering as references (the current tests fail with it). So 03898af builds the class from the non-ASCII part of \p{P} plus U+3000: that covers 「」, “”, 。，、 and so on, and leaves the ASCII rules as they were.

OK, but the current implement is too complicated.

I guess the best approach is like this: keep the regexp as simple as possible, then use our own code to check the surrounding bytes. This approach should be flexible enough and clear enough.

…code The patterns now accept any non-ASCII character next to a reference, and validReferenceBoundaries checks it: punctuation or a space is fine ("修复 go-gitea#1。", "（go-gitea#2）"), a letter is not ("修复go-gitea#1"). This replaces the generated character class. Co-Authored-By: Claude Opus 5.5 <[email protected]>

b84330808 · 2026-11-01T02:34:59Z

Makes sense, simplified it: the patterns now just accept any non-ASCII character next to the reference, and a small validReferenceBoundaries check in code allows it only if it's punctuation or a space. So 修复 #1。 and （#2） link, 修复#1 still doesn't, and the generated class is gone.

* origin/main: (92 commits) feat(release): allow setting a release as the latest (go-gitea#40214) fix(repo): reject unparsable required approvals in branch protection (go-gitea#40213) ci: update giteabot to v1.5.1, label conflict-free backports (go-gitea#40247) fix(pull): enforce user blocks on pull request reviews (go-gitea#40194) feat(auth)!: remove support for obsolete OpenID 2.0 draft (go-gitea#40231) enhance(codeowners): support section / glob / file view display / skip review request (go-gitea#40243) fix: upgrade Alpine base packages in Docker images (go-gitea#40252) fix(deps): update module github.com/andybalholm/brotli to v1.2.7 [security] (go-gitea#40246) fix(perm): keep read access for restricted collaborators on public repos (go-gitea#40234) fix(build): pass `#` test patterns to `go test -run` unchanged (go-gitea#39977) perf: reduce git processes and queries on pull request view (go-gitea#40224) fix(dump): --file extension is ignored when --type is not set (go-gitea#40235) enhance: enable `DEFAULT_DELETE_BRANCH_AFTER_MERGE` by default for new repos (go-gitea#39975) fix(packages): return empty Chef dependencies instead of null (go-gitea#40237) fix(migrations): Do not fail Bitbucket Server migration when pull request activities cannot be loaded (go-gitea#40205) chore(deps): update go.yaml.in/yaml/v4 to v4.0.0 (go-gitea#40222) [skip ci] Updated translations via Crowdin fix: fix missing accessible name on static Fomantic combobox dropdowns (go-gitea#40163) fix(web): build issue popup request from parsed link (go-gitea#40207) fix(web): inline the site manifest (go-gitea#40198) ...

…II characters GitHub links a reference after any non-ASCII character and before anything but a Unicode word character, so "修复go-gitea#1" and "go-gitea#1～go-gitea#5" are references there too. The word character set is now shared with heading anchors. Assisted-by: Claude Code:claude-opus-5-5

The renderer and the cross reference finder repeated the same search, including trailing-space handling that can never apply since the reference itself contains no whitespace. Assisted-by: Claude Code:claude-opus-5-5

…search A mention never contains whitespace, so spaceTrimmedPattern never matched and the search position always advanced to the mention's end. Assisted-by: Claude Code:claude-opus-5-5

silverwind · 2026-11-06T10:28:25Z

Pushed these changes and merged main:

Any non-ASCII character may precede a reference, so 修复#1 links, as it does on GitHub.
Only a Unicode word character after a reference blocks it, so symbols and emoji like #1～#5 or #1🎉 link too.
The word character set is shared with the heading anchors from fix(markup): make heading anchors match GitHub for whitespace and combining marks #40195 as util.IsWordRune.
Rendering and cross reference detection share one search loop.
Removed the dead trailing-space handling (spaceTrimmedPattern) from the reference and mention search.

Written by Claude (Opus 5.5).

* origin/main: (44 commits) fix(markup): recognize issue references next to CJK punctuation (go-gitea#40097) fix: add missing checks to several handlers and services (go-gitea#40120) fix(api): honor branch protection force push and allowlist options on create and edit (go-gitea#39990) fix(repo): mark repository as non-empty before git push returns (go-gitea#40102) [skip ci] Updated translations via Crowdin fix(web): handle TypeCastError when building form validation messages (go-gitea#40212) fix: stop discarding file close and test setup errors (go-gitea#39987) test(e2e): retry clipboard reads rejected by Firefox (go-gitea#40155) refactor: remove duplicated and dead Go code (go-gitea#40244) feat(release): allow setting a release as the latest (go-gitea#40214) fix(repo): reject unparsable required approvals in branch protection (go-gitea#40213) ci: update giteabot to v1.5.1, label conflict-free backports (go-gitea#40247) fix(pull): enforce user blocks on pull request reviews (go-gitea#40194) feat(auth)!: remove support for obsolete OpenID 2.0 draft (go-gitea#40231) enhance(codeowners): support section / glob / file view display / skip review request (go-gitea#40243) fix: upgrade Alpine base packages in Docker images (go-gitea#40252) fix(deps): update module github.com/andybalholm/brotli to v1.2.7 [security] (go-gitea#40246) fix(perm): keep read access for restricted collaborators on public repos (go-gitea#40234) fix(build): pass `#` test patterns to `go test -run` unchanged (go-gitea#39977) perf: reduce git processes and queries on pull request view (go-gitea#40224) ...

wxiaoguang · 2026-11-07T11:29:02Z

 	// issueNumericPattern matches string that references to a numeric issue, e.g. #1287
-	issueNumericPattern = regexp.MustCompile(`(?:\s|^|\(|\[|\'|\")([#!][0-9]+)(?:\s|$|\)|\]|\'|\"|[:;,.?!]\s|[:;,.?!]$)`)
+	// TODO: REFERENCE-REGEXP-PATTERN: findValidReferenceIndex checks a non-ASCII trailing character
+	issueNumericPattern = regexp.MustCompile(`(?:\s|^|\(|\[|\'|\"|[^\x00-\x7f])([#!][0-9]+)(?:\s|$|\)|\]|\'|\"|[:;,.?!]\s|[:;,.?!]$|[^\x00-\x7f])`)


refactor: lazy init global regexp variables - #40269

* origin/main: fix(org): enforce unique team names in the database (go-gitea#40228) fix: truncate strings on character boundaries (go-gitea#40267) feat(packages): add private Terraform module registry (go-gitea#38040) fix(markup): recognize issue references next to CJK punctuation (go-gitea#40097) fix: add missing checks to several handlers and services (go-gitea#40120) fix(api): honor branch protection force push and allowlist options on create and edit (go-gitea#39990) fix(repo): mark repository as non-empty before git push returns (go-gitea#40102) [skip ci] Updated translations via Crowdin fix(web): handle TypeCastError when building form validation messages (go-gitea#40212) fix: stop discarding file close and test setup errors (go-gitea#39987) test(e2e): retry clipboard reads rejected by Firefox (go-gitea#40155) refactor: remove duplicated and dead Go code (go-gitea#40244) feat(release): allow setting a release as the latest (go-gitea#40214) fix(repo): reject unparsable required approvals in branch protection (go-gitea#40213) ci: update giteabot to v1.5.1, label conflict-free backports (go-gitea#40247) fix(pull): enforce user blocks on pull request reviews (go-gitea#40194) feat(auth)!: remove support for obsolete OpenID 2.0 draft (go-gitea#40231) enhance(codeowners): support section / glob / file view display / skip review request (go-gitea#40243) fix: upgrade Alpine base packages in Docker images (go-gitea#40252) fix(deps): update module github.com/andybalholm/brotli to v1.2.7 [security] (go-gitea#40246) Assisted-by: Claude Code:claude-opus-5-5 # Conflicts: # services/release/release_test.go

* origin/main: (66 commits) chore: add report requirements (go-gitea#40262) refactor: lazy init global regexp variables (go-gitea#40269) fix(org): enforce unique team names in the database (go-gitea#40228) fix: truncate strings on character boundaries (go-gitea#40267) feat(packages): add private Terraform module registry (go-gitea#38040) fix(markup): recognize issue references next to CJK punctuation (go-gitea#40097) fix: add missing checks to several handlers and services (go-gitea#40120) fix(api): honor branch protection force push and allowlist options on create and edit (go-gitea#39990) fix(repo): mark repository as non-empty before git push returns (go-gitea#40102) [skip ci] Updated translations via Crowdin fix(web): handle TypeCastError when building form validation messages (go-gitea#40212) fix: stop discarding file close and test setup errors (go-gitea#39987) test(e2e): retry clipboard reads rejected by Firefox (go-gitea#40155) refactor: remove duplicated and dead Go code (go-gitea#40244) feat(release): allow setting a release as the latest (go-gitea#40214) fix(repo): reject unparsable required approvals in branch protection (go-gitea#40213) ci: update giteabot to v1.5.1, label conflict-free backports (go-gitea#40247) fix(pull): enforce user blocks on pull request reviews (go-gitea#40194) feat(auth)!: remove support for obsolete OpenID 2.0 draft (go-gitea#40231) enhance(codeowners): support section / glob / file view display / skip review request (go-gitea#40243) ... # Conflicts: # tests/e2e/webauthn.test.ts

* origin/main: (84 commits) fix(web): don't block README images with `REQUIRE_SIGNIN_VIEW = expensive` (go-gitea#40264) enhance(actions): create commit status for workflow_dispatch runs (go-gitea#39961) fix(repo): do not reuse pull mirror credentials for a new host (go-gitea#39991) chore: add report requirements (go-gitea#40262) refactor: lazy init global regexp variables (go-gitea#40269) fix(org): enforce unique team names in the database (go-gitea#40228) fix: truncate strings on character boundaries (go-gitea#40267) feat(packages): add private Terraform module registry (go-gitea#38040) fix(markup): recognize issue references next to CJK punctuation (go-gitea#40097) fix: add missing checks to several handlers and services (go-gitea#40120) fix(api): honor branch protection force push and allowlist options on create and edit (go-gitea#39990) fix(repo): mark repository as non-empty before git push returns (go-gitea#40102) [skip ci] Updated translations via Crowdin fix(web): handle TypeCastError when building form validation messages (go-gitea#40212) fix: stop discarding file close and test setup errors (go-gitea#39987) test(e2e): retry clipboard reads rejected by Firefox (go-gitea#40155) refactor: remove duplicated and dead Go code (go-gitea#40244) feat(release): allow setting a release as the latest (go-gitea#40214) fix(repo): reject unparsable required approvals in branch protection (go-gitea#40213) ci: update giteabot to v1.5.1, label conflict-free backports (go-gitea#40247) ...

…itea#40097) Issue references next to non-ASCII characters, common in Chinese and Japanese text, are ignored: `修复go-gitea#1`, `（go-gitea#2）` or `Closes go-gitea#3，` create no link or cross reference. Like on GitHub, any non-ASCII character may now precede `go-gitea#123`, `!123` or `owner/repo#123`, and anything but a word character may follow it. Dead trailing-space handling in the reference and mention search is removed. I used Claude Code to write this change. --------- Co-authored-by: Claude Opus 5.5 <[email protected]> Co-authored-by: silverwind <[email protected]> Co-authored-by: wxiaoguang <[email protected]>

github-actions Bot added topic/content-rendering Changes how certain filetypes are displayed type/bug labels Oct 30, 2026

GiteaBot added the lgtm/need 2 This PR needs two approvals by maintainers to be considered for merging. label Oct 30, 2026

silverwind added 4 commits November 6, 2026 11:11

refactor(references): share one search loop for issue references

5093b34

The renderer and the cross reference finder repeated the same search, including trailing-space handling that can never apply since the reference itself contains no whitespace. Assisted-by: Claude Code:claude-opus-5-5

refactor(references): drop dead trailing-space handling from mention …

02200d3

…search A mention never contains whitespace, so spaceTrimmedPattern never matched and the search position always advanced to the mention's end. Assisted-by: Claude Code:claude-opus-5-5

silverwind approved these changes Nov 6, 2026

View reviewed changes

GiteaBot added lgtm/need 1 This PR needs approval from one additional maintainer to be merged. and removed lgtm/need 2 This PR needs two approvals by maintainers to be considered for merging. labels Nov 6, 2026

wxiaoguang added 3 commits November 7, 2026 17:23

Merge branch 'main' into fix-cjk-issue-refs

94f2624

IsWordRune is not a general function

4440b8e

comment

db27352

wxiaoguang approved these changes Nov 7, 2026

View reviewed changes

GiteaBot added lgtm/done This PR has enough approvals to get merged. There are no important open reservations anymore. and removed lgtm/need 1 This PR needs approval from one additional maintainer to be merged. labels Nov 7, 2026

silverwind enabled auto-merge November 7, 2026 10:03

silverwind added this pull request to the merge queue Nov 7, 2026

Merged via the queue into go-gitea:main with commit 7efb5c0 Nov 7, 2026
24 checks passed

GiteaBot added this to the 29.0.0 milestone Nov 7, 2026

wxiaoguang reviewed Nov 7, 2026

View reviewed changes

Navigation Menu

Repository navigation

fix(markup): recognize issue references next to CJK punctuation - #40097

fix(markup): recognize issue references next to CJK punctuation#40097
silverwind merged 10 commits into
go-gitea:mainfrom
b84330808:fix-cjk-issue-refs

b84330808 commented Oct 30, 2026 •

edited by silverwind

Loading

Uh oh!

wxiaoguang commented Oct 30, 2026

Uh oh!

wxiaoguang commented Oct 30, 2026

Uh oh!

b84330808 commented Oct 31, 2026

Uh oh!

wxiaoguang commented Oct 31, 2026

Uh oh!

b84330808 commented Nov 1, 2026

Uh oh!

silverwind commented Nov 6, 2026

Uh oh!

Uh oh!

wxiaoguang Nov 7, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

b84330808 commented Oct 30, 2026 • edited by silverwind Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

wxiaoguang commented Oct 30, 2026

Uh oh!

wxiaoguang commented Oct 30, 2026

Uh oh!

b84330808 commented Oct 31, 2026

Uh oh!

wxiaoguang commented Oct 31, 2026

Uh oh!

b84330808 commented Nov 1, 2026

Uh oh!

silverwind commented Nov 6, 2026

Uh oh!

Uh oh!

wxiaoguang Nov 7, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

b84330808 commented Oct 30, 2026 •

edited by silverwind

Loading