index : cgit

#!/bin/sh

# The plain page hands over a repository's own bytes. A file with no better
# guess is served as text unless its bytes look binary, a directory answers
# with a bare listing of links, and a path that names nothing comes back as
# a 404 rather than an empty document. The blob page, which does the same
# for an object named by its id, is checked here too.

test_description='Check content on plain page'
. ./setup.sh

test_expect_success 'generate foo/plain/file-1' '
	cgit_ok_url "foo/plain/file-1" >tmp
'

test_expect_success 'a text file is served as text/plain' '
	head -n 1 tmp | grep "Status: 200 OK" &&
	sed "/^$/q" tmp | grep "^Content-Type: text/plain; charset=UTF-8$"
'

test_expect_success 'the body carries the blob byte for byte' '
	strip_headers <tmp >body &&
	git -C repos/foo show master:file-1 >expect &&
	test_cmp expect body
'

# Without html serving the bytes are the repository's, so the browser is
# told neither to guess a type of its own nor to load what they reference.
# A repository that serves html vouches for its files and gets neither.
test_expect_success 'a plain file is served under the untrusted headers' '
	cgit_ok_url "foo/plain/file-1" >tmp &&
	sed "/^$/q" tmp >headers &&
	test_grep "^X-Content-Type-Options: nosniff$" headers &&
	test_grep "^Content-Security-Policy: default-src .none.$" headers &&
	cgitrc_copy servingrc enable-html-serving=1 &&
	cgit_ok servingrc "url=foo/plain/file-1" >tmp &&
	sed "/^$/q" tmp >headers &&
	test_grep "^Content-Type: text/plain" headers &&
	test_grep ! "^X-Content-Type-Options:" headers &&
	test_grep ! "^Content-Security-Policy:" headers
'

test_expect_success 'a missing path is a 404' '
	cgit_url "foo/plain/no-such-file" >tmp &&
	head -n 1 tmp | grep "Status: 404"
'

test_expect_success 'a directory answers with a listing of links' '
	cgit_ok_url "foo/plain" >tmp &&
	test_grep "<li><a href=./foo/plain/file-1.>file-1</a></li>" tmp &&
	test_grep "<li><a href=./foo/plain/file-5.>file-5</a></li>" tmp
'

# A folder's listing leads back up a level, to the root's listing from a
# folder at the top.
test_expect_success 'a folder listing links its entries and its parent' '
	test_create_repo repos/folders &&
	(
		cd repos/folders &&
		mkdir -p dir/sub &&
		echo a >dir/a.txt &&
		echo b >dir/sub/b.txt &&
		git add dir &&
		git commit -m folders
	) &&
	cgitrc_repo foldersrc folders &&
	cgit_ok foldersrc "url=folders/plain/dir" >tmp &&
	test_grep "<title>/dir/</title>" tmp &&
	test_grep "<li><a href=./folders/plain/.>\.\./</a></li>" tmp &&
	test_grep "<li><a href=./folders/plain/dir/a.txt.>a.txt</a></li>" tmp &&
	test_grep "<li><a href=./folders/plain/dir/sub/.>sub</a></li>" tmp &&
	cgit_ok foldersrc "url=folders/plain/dir/sub" >tmp &&
	test_grep "<li><a href=./folders/plain/dir/.>\.\./</a></li>" tmp &&
	test_grep "<li><a href=./folders/plain/dir/sub/b.txt.>b.txt</a></li>" tmp
'

# A symlink is answered with its target as text. A submodule's commit lives
# in another repository, so its entry names that commit rather than linking
# a path here.
test_expect_success 'plain serves a link target and lists a submodule' '
	mkrepo repos/plainmod 1 &&
	(
		cd repos/plainmod &&
		sub=$(git rev-parse HEAD) &&
		git update-index --add --cacheinfo 160000,$sub,sub &&
		link=$(printf file-1 | git hash-object -w --stdin) &&
		git update-index --add --cacheinfo 120000,$link,link &&
		git commit -m "link and submodule"
	) &&
	cgitrc_repo plainmodrc plainmod &&
	cgit_ok plainmodrc "url=plainmod/plain/link" >tmp &&
	strip_headers <tmp >body &&
	printf file-1 >expect &&
	test_cmp expect body &&
	sub=$(git -C repos/plainmod rev-parse HEAD~1) &&
	cgit_ok plainmodrc "url=plainmod/plain/" >tmp &&
	test_grep "<li><span>sub</span> @ <span class=.ls-mod-hash. title=.$sub.>" tmp &&
	test_grep ! "href=./plainmod/plain/sub" tmp
'

test_expect_success 'a blob whose bytes look binary is served as octet-stream' '
	(
		cd repos/foo &&
		printf "\000\001\002" >blob.bin &&
		git add blob.bin &&
		git commit -m binary
	) &&
	cgit_ok_url "foo/plain/blob.bin" >tmp &&
	sed "/^$/q" tmp | grep "^Content-Type: application/octet-stream$"
'

# The blob page hands over one object's bytes by its id alone, under the
# same untrusted headers as plain, with nothing in a tree to name a type by.
test_expect_success 'the blob page serves an object by its id' '
	blob=$(git -C repos/foo rev-parse master:file-1) &&
	cgit_ok_url "foo/blob/&id=$blob" >tmp &&
	sed "/^$/q" tmp >headers &&
	test_grep "^Content-Type: text/plain; charset=UTF-8$" headers &&
	test_grep "^X-Content-Type-Options: nosniff$" headers &&
	test_grep "^Content-Security-Policy: default-src .none.$" headers &&
	strip_headers <tmp >body &&
	git -C repos/foo cat-file blob $blob >expect &&
	test_cmp expect body
'

test_expect_success 'the blob page serves binary bytes as octet-stream' '
	blob=$(printf "\\000\\001" | git -C repos/foo hash-object -w --stdin) &&
	cgit_ok_url "foo/blob/&id=$blob" >tmp &&
	sed "/^$/q" tmp >headers &&
	test_grep "^Content-Type: application/octet-stream$" headers
'

test_expect_success 'the blob page refuses an id that is no object id' '
	cgit_url "foo/blob/&id=xyz" >tmp &&
	test_grep "^Status: 404" tmp &&
	test_grep "<div class=.error.>Bad object id: xyz</div>" tmp
'

# RFC 7303 names the XML types by an xml subtype or a +xml suffix, and those
# take a charset like text does. A type that only spells xml inside a word,
# as the Office formats do, is binary and gets none.
test_expect_success 'only text and xml types carry a charset' '
	(
		cd repos/foo &&
		printf "PK" >doc.docx &&
		printf "<svg/>" >pic.svg &&
		git add doc.docx pic.svg &&
		git commit -m types
	) &&
	docx=application/vnd.openxmlformats-officedocument. &&
	docx=${docx}wordprocessingml.document &&
	cgitrc_copy typesrc enable-html-serving=1 "mimetype.docx=$docx" \
		mimetype.svg=image/svg+xml &&
	cgit_ok typesrc "url=foo/plain/doc.docx" >tmp &&
	sed "/^$/q" tmp | grep "^Content-Type: $docx$" &&
	cgit_ok typesrc "url=foo/plain/pic.svg" >tmp &&
	sed "/^$/q" tmp | grep "^Content-Type: image/svg+xml; charset=UTF-8$"
'

# Without html serving a text or application type gives way to the sniffed
# one, except PDF, which a browser shows without running the site's code.
test_expect_success 'only PDF keeps an application type without html serving' '
	test_create_repo repos/types &&
	(
		cd repos/types &&
		printf "%%PDF-1.4" >doc.pdf &&
		printf "alert(1)" >run.js &&
		git add doc.pdf run.js &&
		git commit -m types
	) &&
	cgitrc_repo pdfrc types mimetype.pdf=application/pdf mimetype.js=application/javascript &&
	cgit_ok pdfrc "url=types/plain/doc.pdf" >tmp &&
	sed "/^$/q" tmp | grep "^Content-Type: application/pdf$" &&
	cgit_ok pdfrc "url=types/plain/run.js" >tmp &&
	sed "/^$/q" tmp | grep "^Content-Type: text/plain; charset=UTF-8$"
'

# A tree can carry two entries of one name, which only a repository with
# fsck off accepts, and the plain view answers the first alone.
test_expect_success 'plain answers once for duplicate tree entries' '
	test_create_repo repos/dup &&
	(
		cd repos/dup &&
		blob=$(printf PNGDATA | git hash-object -w --stdin) &&
		inner=$(printf "100644 blob %s\\tinside\\n" $blob | git mktree) &&
		top=$(printf "100644 blob %s\\tx.png\\n040000 tree %s\\tx.png\\n" $blob $inner | git mktree --missing) &&
		c=$(echo dup | git commit-tree $top) &&
		git update-ref refs/heads/master $c
	) &&
	cgitrc_repo duprc dup mimetype.png=image/png &&
	cgit_ok duprc "url=dup/plain/x.png" >tmp &&
	test $(grep -c "^Status:" tmp) = 1 &&
	test $(grep -c "^Content-Type:" tmp) = 1 &&
	grep "^Content-Type: image/png" tmp
'

# A header line cannot carry a control byte, and a web server refuses the
# whole response over one, so a file name holding one is sent with a stand-in.
# The commit joins the dup repository the test above built.
test_expect_success 'a control character in a file name stays out of the headers' '
	(
		cd repos/dup &&
		blob=$(printf x | git hash-object -w --stdin) &&
		top=$(printf "100644 blob %s\\tctl\\001name.txt\\n" $blob | git mktree) &&
		c=$(echo ctl | git commit-tree $top) &&
		git update-ref refs/heads/ctl $c
	) &&
	cgit_ok duprc "url=dup/plain/ctl%01name.txt&h=ctl" >tmp &&
	grep "^Content-Disposition: inline; filename=.ctl?name.txt.$" tmp
'

test_done