From eae88a7473475511220767b7f2ed792da82a79d6 Mon Sep 17 00:00:00 2001
From: Daniel Stenberg <daniel@haxx.se>
Date: Wed, 12 Aug 2026 12:11:37 +0200
Subject: [PATCH] urlapi: avoid dedotdotify() if possible

The dedotdotify() function that removes ./ and ../ sequences from paths
juggles memory and is slow. Now needs_dedotdot() is called first to
determine if the removal process is necessary and otherwise avoids doing
it. Avoids unnecessary memory operations.

Adjusted unit test 1395 accordingly because now a lot of input strings
return NULL for "no change necessary".

Suggested-by: Max Dymond

Closes #22557
---
 lib/urlapi.c          | 29 ++++++++++++++++++---
 tests/unit/unit1395.c | 60 +++++++++++++++++++++----------------------
 2 files changed, 56 insertions(+), 33 deletions(-)

diff --git a/lib/urlapi.c b/lib/urlapi.c
index 1e61fbcae5..1b498e95d0 100644
--- a/lib/urlapi.c
+++ b/lib/urlapi.c
@@ -750,6 +750,29 @@ static bool is_dot(const char **str, size_t *clen)
 
 #define ISSLASH(x) ((x) == '/')
 
+/* prescan the string to see if it needs work */
+static bool needs_dedotdot(const char *p, size_t pn)
+{
+  /* a single byte path cannot be cleaned up */
+  if(pn < 2)
+    return FALSE;
+  while(pn) {
+    if(is_dot(&p, &pn)) {
+      /* "./" or dot before end of string */
+      if(!pn || ISSLASH(*p))
+        return TRUE;
+      /* "../" or ".." before end of string */
+      else if(is_dot(&p, &pn) && (!pn || ISSLASH(*p)))
+        return TRUE;
+    }
+    else {
+      p++;
+      pn--;
+    }
+  }
+  return FALSE;
+}
+
 /*
  * dedotdotify()
  *
@@ -761,7 +784,8 @@ static bool is_dot(const char **str, size_t *clen)
  *
  * RETURNS
  *
- * Zero for success and 'out' set to an allocated dedotdotified string.
+ * Zero for success and 'out' set to an allocated string (or NULL if there's
+ * nothing to do).
  *
  * @unittest 1395
  */
@@ -776,8 +800,7 @@ UNITTEST int dedotdotify(const char *input, size_t clen, char **outp)
   size_t dlen = clen;
 
   *outp = NULL;
-  /* a single byte path cannot be cleaned up */
-  if(clen < 2)
+  if(!needs_dedotdot(input, clen))
     return 0;
 
   curlx_dyn_init(&out, clen + 1);
diff --git a/tests/unit/unit1395.c b/tests/unit/unit1395.c
index a286303977..58945b59a2 100644
--- a/tests/unit/unit1395.c
+++ b/tests/unit/unit1395.c
@@ -39,33 +39,33 @@ static CURLcode test_unit1395(const char *arg)
     { "/%2f%2e%2e%2f/../a", "/a" },
     { "/%2f%2e%2e%2f/../", "/" },
     { "/%2f%2e%2e%2f/.", "/%2f%2e%2e%2f/" },
-    { "/%2f%2e%2e%2f/", "/%2f%2e%2e%2f/" },
-    { "/%2f%2e%2e%2f", "/%2f%2e%2e%2f" },
-    { "/%2f%2e%2e%2", "/%2f%2e%2e%2" },
-    { "/%2f%2e%2e%", "/%2f%2e%2e%" },
+    { "/%2f%2e%2e%2f/", NULL },
+    { "/%2f%2e%2e%2f", NULL },
+    { "/%2f%2e%2e%2", NULL },
+    { "/%2f%2e%2e%", NULL },
     { "/%2f%2e%2e", "/%2f%2e%2e" },
-    { "/%2f%2e%2", "/%2f%2e%2" },
-    { "/%2f%2e%", "/%2f%2e%" },
+    { "/%2f%2e%2", NULL },
+    { "/%2f%2e%", NULL },
     { "/%2f%2e", "/%2f%2e" },
-    { "/%2f%2", "/%2f%2" },
-    { "/%2f%", "/%2f%" },
-    { "/%2f", "/%2f" },
-    { "/%2", "/%2" },
+    { "/%2f%2", NULL },
+    { "/%2f%", NULL },
+    { "/%2f", NULL },
+    { "/%2", NULL },
     { "%2f%2e%2e%2f/../a", "%2f%2e%2e%2f/a" },
     { "%2f%2e%2e%2f/../", "%2f%2e%2e%2f/" },
     { "%2f%2e%2e%2f/.", "%2f%2e%2e%2f/" },
-    { "%2f%2e%2e%2f/", "%2f%2e%2e%2f/" },
-    { "%2f%2e%2e%2f", "%2f%2e%2e%2f" },
-    { "%2f%2e%2e%2", "%2f%2e%2e%2" },
-    { "%2f%2e%2e%", "%2f%2e%2e%" },
+    { "%2f%2e%2e%2f/", NULL },
+    { "%2f%2e%2e%2f", NULL },
+    { "%2f%2e%2e%2", NULL },
+    { "%2f%2e%2e%", NULL },
     { "%2f%2e%2e", "%2f%2e%2e" },
-    { "%2f%2e%2", "%2f%2e%2" },
-    { "%2f%2e%", "%2f%2e%" },
+    { "%2f%2e%2", NULL },
+    { "%2f%2e%", NULL },
     { "%2f%2e", "%2f%2e" },
-    { "%2f%2", "%2f%2" },
-    { "%2f%", "%2f%" },
-    { "%2f", "%2f" },
-    { "%2", "%2" },
+    { "%2f%2", NULL },
+    { "%2f%", NULL },
+    { "%2f", NULL },
+    { "%2", NULL },
     { "%", NULL },
     { "2", NULL },
     { "e", NULL },
@@ -108,12 +108,12 @@ static CURLcode test_unit1395(const char *arg)
     { "/1/./%2e.", "/" },
     { "/1/./../2", "/2" },
     { "/hello/1/./../2", "/hello/2" },
-    { "test/this", "test/this" },
+    { "test/this", NULL },
     { "test/this/../now", "test/now" },
     { "/1../moo../foo", "/1../moo../foo" },
     { "/../../moo", "/moo" },
     { "/../../moo?", "/moo?" },
-    { "/123?", "/123?" },
+    { "/123?", NULL },
     { "/", NULL },
     { "", NULL },
     { "/.../", "/.../" },
@@ -131,16 +131,16 @@ static CURLcode test_unit1395(const char *arg)
     { "/a/%2E%2e/b", "/b" },
     { "/a/%2e./b", "/b" },
     { "/a/.%2e/b", "/b" },
-    { "/%2f..%2f", "/%2f..%2f" },
+    { "/%2f..%2f", NULL },
     { "/a/b/.", "/a/b/" },
     { "/a/b/..", "/a/" },
-    { "well-known", "well-known" },
-    { ".well-known", ".well-known" },
-    { "..well-known", "..well-known" },
-    { "...well-known", "...well-known" },
-    { "....well-known", "....well-known" },
-    { "%2ewell-known", "%2ewell-known" },
-    { "%2Ewell-known", "%2Ewell-known" },
+    { "well-known", NULL },
+    { ".well-known", NULL },
+    { "..well-known", NULL },
+    { "...well-known", NULL },
+    { "....well-known", NULL },
+    { "%2ewell-known", NULL },
+    { "%2Ewell-known", NULL },
     { "../.well-known", ".well-known" },
   };