docs(diff): say that files over the one-file cap take no line comment

commit: de918e26676ba21609d469826188eae32c7509f1
author: brice · 2d ago · 2026-10-20 08:19 UTC
committer: brice · 2d ago · 2026-10-20 09:16 UTC
parent: 4af5505
signature: GPG VerifiedCD88F9E564E0102646FB6077A1A07C7C7C3E349D · arch-gpg Verified for the committer email

2 files · +5 -4

M docs/merge-requests.md +1 -1


@@ -67,7 +67,7 @@ In the web merge box a mode that is not possible is disabled with its reason, an
67	67
68	68		The merge request, commit and compare pages list every changed file. Generated files (lockfiles, `vendor/`, `node_modules/`, minified or generated code) and large files (over 2,000 changed lines) are collapsed: the header and the counts show, the lines do not. "load diff" fetches that one file in place, in full, and you can comment on its lines. Without scripts the link reloads the page with that file loaded. After the first 25 files of a merge request, the other files load as you scroll.
69	69
70		-	A commit page reads the hunks of its first 300 files with the page; the other files are listed collapsed and load one at a time. A file that cannot be shown in the browser, over 8 MiB of text or over 50,000 changed lines when loaded alone, has no "load diff": the page links to its raw blob instead.
	70	+	A commit page reads the hunks of its first 300 files with the page; the other files are listed collapsed and load one at a time. A file that cannot be shown in the browser, over 8 MiB of text or over 50,000 changed lines when loaded alone, has no "load diff": the page links to its raw blob instead, and its lines take no comment.
71	71
72	72		## Notes
73	73


@@ -67,7 +67,7 @@ In the web merge box a mode that is not possible is disabled with its reason, an
67			67
68		The merge request, commit and compare pages list every changed file. Generated files (lockfiles, `vendor/`, `node_modules/`, minified or generated code) and large files (over 2,000 changed lines) are collapsed: the header and the counts show, the lines do not. "load diff" fetches that one file in place, in full, and you can comment on its lines. Without scripts the link reloads the page with that file loaded. After the first 25 files of a merge request, the other files load as you scroll.	68		The merge request, commit and compare pages list every changed file. Generated files (lockfiles, `vendor/`, `node_modules/`, minified or generated code) and large files (over 2,000 changed lines) are collapsed: the header and the counts show, the lines do not. "load diff" fetches that one file in place, in full, and you can comment on its lines. Without scripts the link reloads the page with that file loaded. After the first 25 files of a merge request, the other files load as you scroll.
69			69
70	-	A commit page reads the hunks of its first 300 files with the page; the other files are listed collapsed and load one at a time. A file that cannot be shown in the browser, over 8 MiB of text or over 50,000 changed lines when loaded alone, has no "load diff": the page links to its raw blob instead.	70	+	A commit page reads the hunks of its first 300 files with the page; the other files are listed collapsed and load one at a time. A file that cannot be shown in the browser, over 8 MiB of text or over 50,000 changed lines when loaded alone, has no "load diff": the page links to its raw blob instead, and its lines take no comment.
71			71
72		## Notes	72		## Notes
73			73

M internal/usecase/mergerequests.go +4 -3


@@ -780,9 +780,10 @@ func (m *MergeRequests) checkLine(ctx context.Context, mr domain.MergeRequest, p
780	780		// file's hunks within the default total limits would reject a comment on a
781	781		// file the earlier ones left no budget for. So a first read lists the files
782	782		// with no hunks (From past the end: cheap, git is stopped at once), and a
783		-	// second loads the one file in full (a Full window: no line cap, the byte
784		-	// guard only, generated files included), as the web view does when it loads
785		-	// a collapsed file: any line the page can show can be commented on.
	783	+	// second loads the one file in full (a Full window: the byte guard and
	784	+	// domain.MaxOneFileLines only, generated files included), as the web view
	785	+	// does when it loads a collapsed file: any line the page can show can be
	786	+	// commented on, and a file over those limits takes no line comment.
786	787		list := domain.DefaultDiffLimits()
787	788		list.MaxFiles, list.From = 0, math.MaxInt
788	789		_, diff, err := m.git.DiffCommits(ctx, mr.Repo, base, mr.HeadID, list)

 	// file's hunks within the default total limits would reject a comment on a
 	// file the earlier ones left no budget for. So a first read lists the files
 	// with no hunks (From past the end: cheap, git is stopped at once), and a
-	// second loads the one file in full (a Full window: no line cap, the byte
+	// second loads the one file in full (a Full window: the byte guard and
-	// guard only, generated files included), as the web view does when it loads
+	// domain.MaxOneFileLines only, generated files included), as the web view
-	// a collapsed file: any line the page can show can be commented on.
+	// does when it loads a collapsed file: any line the page can show can be
+	// commented on, and a file over those limits takes no line comment.
 	list := domain.DefaultDiffLimits()
 	list.MaxFiles, list.From = 0, math.MaxInt
 	_, diff, err := m.git.DiffCommits(ctx, mr.Repo, base, mr.HeadID, list)