[#138] Report links that escape repo directory

Problem: as in #138, when we see a local link, we are checking only
existance of referred file,
not checking that this file is a part of repo
and link will compatible with Github's renderer

Solution: manually count "nesting levels" of all local links,
checking that number of `".."`'s is always less
then number of real directories

src/Xrefcheck/Verify.hs

@@ -58,7 +58,7 @@ import Network.HTTP.Types.Header (hRetryAfter)
 import Network.HTTP.Types.Status (Status, statusCode, statusMessage)
 import System.Console.Pretty (Style (..), style)
 import System.Directory (doesDirectoryExist, doesFileExist)
-import System.FilePath (normalise, takeDirectory, (</>))
+import System.FilePath (takeDirectory, (</>), normalise, splitDirectories, makeRelative)
 import Text.ParserCombinators.ReadPrec qualified as ReadPrec (lift)
 import Text.Regex.TDFA.Text (Regex, regexec)
 import Text.URI (Authority (..), ParseExceptionBs, URI (..), mkURIBs)
@@ -119,6 +119,7 @@ instance Buildable a => Buildable (WithReferenceLoc a) where
 
 data VerifyError
   = LocalFileDoesNotExist FilePath
+  | LocalFileOutsideRepo FilePath
   | AnchorDoesNotExist Text [Anchor]
   | AmbiguousAnchorRef FilePath Text (NonEmpty Anchor)
   | ExternalResourceInvalidUri URIBS.URIParseError
@@ -138,6 +139,9 @@ instance Buildable VerifyError where
     LocalFileDoesNotExist file ->
       "⛀  File does not exist:\n   " +| file |+ "\n"
 
+    LocalFileOutsideRepo file ->
+      "⛀  Link targets a local file outside repository:\n   " +| file |+ "\n"
+
     AnchorDoesNotExist anchor similar ->
       "⛀  Anchor '" +| anchor |+ "' is not present" +|
       anchorHints similar
@@ -407,12 +411,29 @@ verifyReference
       _ -> Nothing
 
     checkRef mAnchor referredFile = verifying $ do
+      checkReferredFileIsInsideRepo referredFile
       checkReferredFileExists referredFile
       case M.lookup referredFile repoInfo of
         Nothing -> pass  -- no support for such file, can do nothing
         Just referredFileInfo -> whenJust mAnchor $
           checkAnchor referredFile (_fiAnchors referredFileInfo)
 
+    checkReferredFileIsInsideRepo file = unless
+      (noNegativeNesting $ makeRelative root file)
+      $ throwError (LocalFileOutsideRepo file)
+      where
+        -- | checks that relative filepath fully belongs to current directory
+        -- noNegativeNesting "a/../b" = True
+        -- noNegativeNesting "a/../../b" = False
+        noNegativeNesting path = all (>= 0) $ scanl
+          (\n dir -> n + nestingChange dir)
+          (0 :: Integer)
+          $ splitDirectories path
+
+        nestingChange ".." = -1
+        nestingChange "." = 0
+        nestingChange _ = 1
+
     checkReferredFileExists file = do
       let fileExists = readingSystem $ doesFileExist file
       let dirExists = readingSystem $ doesDirectoryExist file

tests/golden/check-local-references-outside-repo/check-local-references-outside-repo.bats

@@ -0,0 +1,27 @@
+#!/usr/bin/env bats
+
+# SPDX-FileCopyrightText: 2021 Serokell <https://serokell.io>
+#
+# SPDX-License-Identifier: MPL-2.0
+
+load '../helpers/bats-support/load'
+load '../helpers/bats-assert/load'
+load '../helpers'
+
+@test "We report all links that target files outside root" {
+  xrefcheck -r inner-directory\
+  | prepare > /tmp/check-notScanned.test || true
+
+  diff /tmp/check-notScanned.test expected.gold \
+    --ignore-space-change \
+    --ignore-blank-lines \
+    --new-file # treat absent files as empty
+
+  rm /tmp/check-notScanned.test
+}
+
+@test "When target is inside repo root, everything is good" {
+  run xrefcheck -r .\
+
+  assert_output --partial "All repository links are valid."
+}

tests/golden/check-local-references-outside-repo/expected.gold

@@ -0,0 +1,23 @@
+=== Invalid references found ===
+
+  ➥  In file inner-directory/check-local-references-outside-repo.md
+     bad reference (relative) at src:7:1-35:
+       - text: "Without anchor"
+       - link: ../link-target.md
+       - anchor: -
+
+     ⛀  Link targets a local file outside repository:
+        inner-directory/../link-target.md
+
+
+  ➥  In file inner-directory/check-local-references-outside-repo.md
+     bad reference (relative) at src:8:1-51:
+       - text: "With anchor"
+       - link: ../link-target.md
+       - anchor: link-target-anchor
+
+     ⛀  Link targets a local file outside repository:
+        inner-directory/../link-target.md
+
+
+Invalid references dumped, 2 in total.

tests/golden/check-local-references-outside-repo/inner-directory/check-local-references-outside-repo.md

@@ -0,0 +1,8 @@
+<!--
+ - SPDX-FileCopyrightText: 2021 Serokell <https://serokell.io>
+ -
+ - SPDX-License-Identifier: MPL-2.0
+ -->
+
+[Without anchor](../link-target.md)
+[With anchor](../link-target.md#link-target-anchor)

tests/golden/check-local-references-outside-repo/link-target.md

@@ -0,0 +1,7 @@
+<!--
+ - SPDX-FileCopyrightText: 2022 Serokell <https://serokell.io>
+ -
+ - SPDX-License-Identifier: MPL-2.0
+ -->
+
+# Link target anchor