From 590367a6e50a98db5776c3bda7ef6a8756535662 Mon Sep 17 00:00:00 2001
From: Takase <20792268+takase1121@users.noreply.github.com>
Date: Sun, 16 Mar 2025 00:08:25 +0800
Subject: [PATCH] syntax: warn against malformed patterns and disable them
 (#2029)

(cherry picked from commit 3501f3d0bb6ee399ecb7cdf04cff8cf138b1ca02)
---
 data/core/syntax.lua    | 48 +++++++++++++++++++++++++++++++++++++++++
 data/core/tokenizer.lua |  1 +
 2 files changed, 49 insertions(+)

diff --git a/data/core/syntax.lua b/data/core/syntax.lua
index 9862fbd..2c6150c 100644
--- a/data/core/syntax.lua
+++ b/data/core/syntax.lua
@@ -1,4 +1,5 @@
 local common = require "core.common"
+local core   = require "core"
 
 local syntax = {}
 syntax.items = {}
@@ -6,10 +7,57 @@ syntax.items = {}
 local plain_text_syntax = { name = "Plain Text", patterns = {}, symbols = {} }
 
 
+---Checks whether the pattern / regex compiles correctly and matches something.
+---A pattern / regex must not match an empty string.
+---@param pattern_type "regex"|"pattern"
+---@param pattern string
+---@return boolean ok
+---@return string? error
+local function check_pattern(pattern_type, pattern)
+  local ok, err, mstart, mend
+  if pattern_type == "regex" then
+    ok, err = regex.compile(pattern)
+    if ok then
+      mstart, mend = regex.find_offsets(ok, "")
+      if mstart and mstart > mend then
+        ok, err = false, "Regex matches an empty string"
+      end
+    end
+  else
+    ok, mstart, mend = pcall(string.ufind, "", pattern)
+    if ok and mstart and mstart > mend then
+      ok, err = false, "Pattern matches an empty string"
+    elseif not ok then
+      err = mstart --[[@as string]]
+    end
+  end
+  return ok --[[@as boolean]], err
+end
+
 function syntax.add(t)
   if type(t.space_handling) ~= "boolean" then t.space_handling = true end
 
   if t.patterns then
+    -- do a sanity check on the patterns / regex to make sure they are actually correct
+    for i, pattern in ipairs(t.patterns) do
+      local p, ok, err, name = pattern.pattern or pattern.regex, nil, nil, nil
+      if type(p) == "table" then
+        for j = 1, 2 do
+          ok, err = check_pattern(pattern.pattern and "pattern" or "regex", p[j])
+          if not ok then name = string.format("#%d:%d <%s>", i, j, p[j]) end
+        end
+      elseif type(p) == "string" then
+        ok, err = check_pattern(pattern.pattern and "pattern" or "regex", p)
+        if not ok then name = string.format("#%d <%s>", i, p) end
+      else
+        ok, err, name = false, "Missing pattern or regex", "#"..i
+      end
+      if not ok then
+        pattern.disabled = true
+        core.warn("Malformed pattern %s in %s language plugin: %s", name, t.name, err)
+      end
+    end
+
     -- the rule %s+ gives us a performance gain for the tokenizer in lines with
     -- long amounts of consecutive spaces, can be disabled by plugins where it
     -- causes conflicts by declaring the table property: space_handling = false
diff --git a/data/core/tokenizer.lua b/data/core/tokenizer.lua
index 3382202..3168a4f 100644
--- a/data/core/tokenizer.lua
+++ b/data/core/tokenizer.lua
@@ -203,6 +203,7 @@ function tokenizer.tokenize(incoming_syntax, text, state, resume)
     local target, res = p.pattern or p.regex, { 1, offset - 1 }
     local p_idx = close and 2 or 1
     local code = type(target) == "table" and target[p_idx] or target
+    if p.disabled then return end
 
     if p.whole_line == nil then p.whole_line = { } end
     if p.whole_line[p_idx] == nil then