Bound htsparse.c pointer-destination buffer writes (batch 15)

The makeindex_firstlink_, base, codebase and loc_ aliases in the HTML
parser are bare char* views onto HTS_URLMAXSIZE*2 caller arrays, so
strcpybuff degraded to a raw strcpy (htssafe.h pointer-dest branch).
Bound all five with strlcpybuff(..., HTS_URLMAXSIZE*2), the documented
capacity of every target (makeindex_firstlink/base/codebase/loc in
htscore.c, r->location aliasing loc).

Behavior-preserving: each source (tempo, lien, back[].r.location) is
itself an HTS_URLMAXSIZE*2 buffer, so its NUL-terminated contents are
<= cap-1 and copy identically; no truncation is reachable. htsparse.c
now has zero pointer-destination warnings; htsserver.c (5) is the last
file before the stub can be flipped to an error.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: Xavier Roche <roche@httrack.com>

src/htsalias.c

@@ -41,19 +41,24 @@ Please visit our Website: http://www.httrack.com
 
 #define _NOT_NULL(a) ( (a!=NULL) ? (a) : "" )
 
-// COPY OF cmdl_ins in htsmain.c
-// Insert a command in the argc/argv
-#define cmdl_ins(token,argc,argv,buff,ptr) \
-  { \
-  int i; \
-  for(i=argc;i>0;i--)\
-  argv[i]=argv[i-1];\
-  } \
-  argv[0]=(buff+ptr); \
-  strcpybuff(argv[0],token); \
-  ptr += (int) (strlen(argv[0])+1); \
+// COPY OF cmdl_ins in htscoremain.c
+/* Bytes left in x_argvblk from offset ptr. The offset can in principle outrun
+   the block (alias/doit.log expansion), so the copy aborts cleanly instead of
+   the subtraction wrapping to a huge unbounded size. */
+#define cmdl_room(bufsize, ptr)                                                \
+  ((ptr) < (size_t) (bufsize) ? (size_t) (bufsize) - (ptr) : 0)
+// Insert a command in the argc/argv (buff has total capacity bufsize)
+#define cmdl_ins(token, argc, argv, buff, bufsize, ptr)                        \
+  {                                                                            \
+    int i;                                                                     \
+    for (i = argc; i > 0; i--)                                                 \
+      argv[i] = argv[i - 1];                                                   \
+  }                                                                            \
+  argv[0] = (buff + ptr);                                                      \
+  strlcpybuff(argv[0], token, cmdl_room(bufsize, ptr));                        \
+  ptr += (int) (strlen(argv[0]) + 1);                                          \
   argc++
-// END OF COPY OF cmdl_ins in htsmain.c
+// END OF COPY OF cmdl_ins in htscoremain.c
 
 /*
   Aliases for command-line and config file definitions
@@ -468,7 +473,7 @@ const char *optalias_help(const char *token) {
 */
 /* Note: NOT utf-8 */
 int optinclude_file(const char *name, int *argc, char **argv, char *x_argvblk,
-                    int *x_ptr) {
+                    size_t x_argvblk_size, int *x_ptr) {
   FILE *fp;
 
   fp = fopen(name, "rb");
@@ -542,14 +547,15 @@ int optinclude_file(const char *name, int *argc, char **argv, char *x_argvblk,
               /* temporary argc: Number of parameters after minus insert_after_argc */
               insert_after_argc = (*argc) - insert_after;
               cmdl_ins((tmp_argv[2]), insert_after_argc, (argv + insert_after),
-                       x_argvblk, (*x_ptr));
+                       x_argvblk, x_argvblk_size, (*x_ptr));
               *argc = insert_after_argc + insert_after;
               insert_after++;
               /* Second one */
               if (return_argc > 1) {
                 insert_after_argc = (*argc) - insert_after;
                 cmdl_ins((tmp_argv[3]), insert_after_argc,
-                         (argv + insert_after), x_argvblk, (*x_ptr));
+                         (argv + insert_after), x_argvblk, x_argvblk_size,
+                         (*x_ptr));
                 *argc = insert_after_argc + insert_after;
                 insert_after++;
               }

src/htsalias.h

@@ -45,7 +45,7 @@ int optalias_find(const char *token);
 const char *optalias_help(const char *token);
 int optreal_find(const char *token);
 int optinclude_file(const char *name, int *argc, char **argv, char *x_argvblk,
-                    int *x_ptr);
+                    size_t x_argvblk_size, int *x_ptr);
 const char *optreal_value(int p);
 const char *optalias_value(int p);
 const char *opttype_value(int p);

src/htscoremain.c

@@ -69,23 +69,29 @@ Please visit our Website: http://www.httrack.com
 /* Resolver */
 extern int IPV6_resolver;
 
-// Add a command in the argc/argv
-#define cmdl_add(token,argc,argv,buff,ptr) \
-  argv[argc]=(buff+ptr); \
-  strcpybuff(argv[argc],token); \
-  ptr += (int) (strlen(argv[argc])+2); \
+/* Remaining room in the argv block; 0 once it is exhausted (alias expansion or
+   doit.log insertion can outrun the +32768 slack), so the copy aborts cleanly
+   instead of the subtraction wrapping to a huge unbounded size. */
+#define cmdl_room(bufsize, ptr)                                                \
+  ((ptr) < (size_t) (bufsize) ? (size_t) (bufsize) - (ptr) : 0)
+
+// Add a command in the argc/argv (buff has total capacity bufsize)
+#define cmdl_add(token, argc, argv, buff, bufsize, ptr)                        \
+  argv[argc] = (buff + ptr);                                                   \
+  strlcpybuff(argv[argc], token, cmdl_room(bufsize, ptr));                     \
+  ptr += (int) (strlen(argv[argc]) + 2);                                       \
   argc++
 
-// Insert a command in the argc/argv
-#define cmdl_ins(token,argc,argv,buff,ptr) \
-  { \
-  int i; \
-  for(i=argc;i>0;i--)\
-  argv[i]=argv[i-1];\
-  } \
-  argv[0]=(buff+ptr); \
-  strcpybuff(argv[0],token); \
-  ptr += (int) (strlen(argv[0])+2); \
+// Insert a command in the argc/argv (buff has total capacity bufsize)
+#define cmdl_ins(token, argc, argv, buff, bufsize, ptr)                        \
+  {                                                                            \
+    int i;                                                                     \
+    for (i = argc; i > 0; i--)                                                 \
+      argv[i] = argv[i - 1];                                                   \
+  }                                                                            \
+  argv[0] = (buff + ptr);                                                      \
+  strlcpybuff(argv[0], token, cmdl_room(bufsize, ptr));                        \
+  ptr += (int) (strlen(argv[0]) + 2);                                          \
   argc++
 
 #define htsmain_free() do { \
@@ -592,6 +598,7 @@ HTSEXT_API int hts_main2(int argc, char **argv, httrackp * opt) {
 static int hts_main_internal(int argc, char **argv, httrackp * opt) {
   char **x_argv = NULL;         // Patch pour argv et argc: en cas de récupération de ligne de commande
   char *x_argvblk = NULL;       // (reprise ou update)
+  size_t x_argvblk_size = 0;    // total capacity of x_argvblk
   int x_ptr = 0;                // offset
 
   //
@@ -669,7 +676,8 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
         *a = ' ';
       /* equivalent to "empty parameter" */
       if ((strcmp(argv[na], HTS_NOPARAM) == 0) || (strcmp(argv[na], HTS_NOPARAM2) == 0))        // (none)
-        strcpybuff(argv[na], "\"\"");
+        /* replacing "(none)"/"\"(none)\"" with "\"\"" always fits in place */
+        strlcpybuff(argv[na], "\"\"", strlen(argv[na]) + 1);
       if (strncmp(argv[na], "-&", 2) == 0)
         argv[na][1] = '%';
     }
@@ -691,6 +699,7 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
       htsmain_free();
       return -1;
     }
+    x_argvblk_size = (size_t) (current_size + 32768);
     x_argvblk[0] = '\0';
     x_ptr = 0;
 
@@ -712,7 +721,7 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
     //
     argv_url = 0;               /* pour comptage */
     //
-    cmdl_add(argv[0], x_argc, x_argv, x_argvblk, x_ptr);
+    cmdl_add(argv[0], x_argc, x_argv, x_argvblk, x_argvblk_size, x_ptr);
     na = 1;                     /* commencer après nom_prg */
     while(na < argc) {
       int result = 1;
@@ -733,9 +742,10 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
         }
 
         /* Copier */
-        cmdl_add(tmp_argv[0], x_argc, x_argv, x_argvblk, x_ptr);
+        cmdl_add(tmp_argv[0], x_argc, x_argv, x_argvblk, x_argvblk_size, x_ptr);
         if (tmp_argc > 1) {
-          cmdl_add(tmp_argv[1], x_argc, x_argv, x_argvblk, x_ptr);
+          cmdl_add(tmp_argv[1], x_argc, x_argv, x_argvblk, x_argvblk_size,
+                   x_ptr);
         }
 
         /* Compter URLs et détecter -i,-q.. */
@@ -807,7 +817,7 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
           char BIGSTK tempo[HTS_CDLMAXSIZE];
 
           strcpybuff(tempo, argv[na] + 1);
-          if (tempo[strlen(tempo) - 1] != '"') {
+          if (tempo[0] == '\0' || tempo[strlen(tempo) - 1] != '"') {
             char BIGSTK s[HTS_CDLMAXSIZE];
 
             sprintf(s, "Missing quote in %s", argv[na]);
@@ -816,7 +826,9 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
             return -1;
           }
           tempo[strlen(tempo) - 1] = '\0';
-          strcpybuff(argv[na], tempo);
+          /* tempo is argv[na] minus its surrounding quotes, so it fits in place
+           */
+          strlcpybuff(argv[na], tempo, strlen(argv[na]) + 1);
         }
 
         if (cmdl_opt(argv[na])) {       // option
@@ -917,18 +929,19 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
               (OPT_GET_BUFF(opt), OPT_GET_BUFF_SIZE(opt),
               StringBuff(opt->path_log),
                "hts-cache/doit.log"))) || (argv_url > 0)) {
-          if (!optinclude_file
-              (fconcat
-               (OPT_GET_BUFF(opt), OPT_GET_BUFF_SIZE(opt),
-               StringBuff(opt->path_log), HTS_HTTRACKRC),
-               &argc, argv, x_argvblk, &x_ptr))
-            if (!optinclude_file(HTS_HTTRACKRC, &argc, argv, x_argvblk, &x_ptr)) {
-              if (!optinclude_file
-                  (fconcat(OPT_GET_BUFF(opt), OPT_GET_BUFF_SIZE(opt),
-                  hts_gethome(), "/" HTS_HTTRACKRC),
-                   &argc, argv, x_argvblk, &x_ptr)) {
+          if (!optinclude_file(
+                  fconcat(OPT_GET_BUFF(opt), OPT_GET_BUFF_SIZE(opt),
+                          StringBuff(opt->path_log), HTS_HTTRACKRC),
+                  &argc, argv, x_argvblk, x_argvblk_size, &x_ptr))
+            if (!optinclude_file(HTS_HTTRACKRC, &argc, argv, x_argvblk,
+                                 x_argvblk_size, &x_ptr)) {
+              if (!optinclude_file(
+                      fconcat(OPT_GET_BUFF(opt), OPT_GET_BUFF_SIZE(opt),
+                              hts_gethome(), "/" HTS_HTTRACKRC),
+                      &argc, argv, x_argvblk, x_argvblk_size, &x_ptr)) {
 #ifdef HTS_HTTRACKCNF
-                optinclude_file(HTS_HTTRACKCNF, &argc, argv, x_argvblk, &x_ptr);
+                optinclude_file(HTS_HTTRACKCNF, &argc, argv, x_argvblk,
+                                x_argvblk_size, &x_ptr);
 #endif
               }
             }
@@ -981,7 +994,7 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
           if (strnotempty(lastp)) {
             insert_after_argc = argc - insert_after;
             cmdl_ins(lastp, insert_after_argc, (argv + insert_after), x_argvblk,
-                     x_ptr);
+                     x_argvblk_size, x_ptr);
             argc = insert_after_argc + insert_after;
             insert_after++;
           }
@@ -1101,7 +1114,7 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
       if (argv[i][0] == '-') {
         if (argv[i][1] == '-') {        // --xxx
           if ((strfield2(argv[i] + 2, "clean")) || (strfield2(argv[i] + 2, "tide"))) {  // nettoyer
-            strcpybuff(argv[i] + 1, "");
+            argv[i][1] = '\0';
             if (fexist
                 (fconcat
                  (OPT_GET_BUFF(opt), OPT_GET_BUFF_SIZE(opt), StringBuff(opt->path_log), "hts-log.txt")))
@@ -1210,7 +1223,8 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
             //
           } else if (strfield2(argv[i] + 2, "catchurl")) {      // capture d'URL via proxy temporaire!
             argv_url = 1;       // forcer a passer les parametres
-            strcpybuff(argv[i] + 1, "#P");
+            /* argv[i] is "--catchurl"; "#P" fits after its first char */
+            strlcpybuff(argv[i] + 1, "#P", strlen(argv[i] + 1) + 1);
             //
           } else if (strfield2(argv[i] + 2, "updatehttrack")) {
 #ifdef _WIN32
@@ -1538,7 +1552,7 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
         char BIGSTK tempo[HTS_CDLMAXSIZE + 256];
 
         strcpybuff(tempo, argv[na] + 1);
-        if (tempo[strlen(tempo) - 1] != '"') {
+        if (tempo[0] == '\0' || tempo[strlen(tempo) - 1] != '"') {
           char s[HTS_CDLMAXSIZE + 256];
 
           sprintf(s, "Missing quote in %s", argv[na]);
@@ -1547,7 +1561,9 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
           return -1;
         }
         tempo[strlen(tempo) - 1] = '\0';
-        strcpybuff(argv[na], tempo);
+        /* tempo is argv[na] minus its surrounding quotes, so it fits in place
+         */
+        strlcpybuff(argv[na], tempo, strlen(argv[na]) + 1);
       }
 
       if (cmdl_opt(argv[na])) { // option
@@ -3206,7 +3222,7 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
         if (urlSize < HTS_URLMAXSIZE) {
           ensureUrlCapacity(url, url_sz, capa);
           if (strnotempty(url))
-            strcatbuff(url, " ");       // espace de séparation
+            strlcatbuff(url, " ", url_sz); // separator space
           append_escape_spc_url(unescape_http_unharm(catbuff, sizeof(catbuff), argv[na], 1), url, url_sz);
         }
       }                         // if argv=- etc. 

src/htsparse.c

@@ -617,13 +617,15 @@ int htsparse(htsmoduleStruct * str, htsmoduleStructExtended * stre) {
                                              "index.html")) == 0) {
                         detect_title = 1;       // ok détecté pour cette page!
                         makeindex_links++;      // un de plus
-                        strcpybuff(makeindex_firstlink, tempo);
+                        strlcpybuff(makeindex_firstlink, tempo,
+                                    HTS_URLMAXSIZE * 2);
                         //
 
                         /* Hack */
                         if (opt->mimehtml) {
-                          strcpybuff(makeindex_firstlink,
-                                     "cid:primary/primary");
+                          strlcpybuff(makeindex_firstlink,
+                                      "cid:primary/primary",
+                                      HTS_URLMAXSIZE * 2);
                         }
 
                         if ((b == a) || (a == NULL) || (b == NULL)) {   // pas de titre
@@ -2319,12 +2321,12 @@ int htsparse(htsmoduleStruct * str, htsmoduleStructExtended * stre) {
                       switch (p_type) {
                       case 2:{
                           //if (*lien!='/') strcatbuff(base,"/");
-                          strcpybuff(base, lien);
+                          strlcpybuff(base, lien, HTS_URLMAXSIZE * 2);
                         }
                         break;  // base
                       case -2:{
                           //if (*lien!='/') strcatbuff(codebase,"/");
-                          strcpybuff(codebase, lien);
+                          strlcpybuff(codebase, lien, HTS_URLMAXSIZE * 2);
                         }
                         break;  // base
                       }
@@ -4397,7 +4399,7 @@ int hts_mirror_wait_for_next_file(htsmoduleStruct * str,
       memcpy(r, &(back[b].r), sizeof(htsblk));
       r->location = stre->loc_; // ne PAS copier location!! adresse, pas de buffer
       if (back[b].r.location)
-        strcpybuff(r->location, back[b].r.location);
+        strlcpybuff(r->location, back[b].r.location, HTS_URLMAXSIZE * 2);
       back[b].r.adr = NULL;     // ne pas faire de desalloc ensuite
 
       // libérer emplacement backing

tests/01_engine-cmdline.test

@@ -30,6 +30,17 @@ run() {
     RC=$?
 }
 
+# crawl using exactly the given args as the only URL(s), no implicit primary URL;
+# leaves the exit status in RC
+run_only() {
+    local out="$1"
+    shift
+    rm -rf "$out"
+    mkdir -p "$out"
+    httrack -O "$out" --quiet -n "$@" >"$out/.log" 2>&1
+    RC=$?
+}
+
 # assert the value was accepted: clean exit and the fixture was mirrored
 accepted() {
     { test "$RC" -eq 0 && test -n "$(find "$1" -type f -path '*/index.html' -print -quit)"; } ||
@@ -68,4 +79,15 @@ refused "#152: over-cap -F not refused cleanly"
 run "$tmp/ov-l" --user-agent "$over"
 refused "#152: over-cap --user-agent not refused cleanly"
 
+# Quote handling on the sole URL (run_only, so the quoted arg is the only URL and
+# can't be masked by an implicit one). A fully "-quoted URL has its surrounding
+# quotes stripped in place and is mirrored; a dangling opening quote, and a lone
+# quote (empty after the opening "), are refused cleanly and never crash.
+run_only "$tmp/q-ok" "\"file://$tmp/index.html\""
+accepted "$tmp/q-ok" "quoted URL not stripped/mirrored"
+run_only "$tmp/q-bad" '"foo'
+refused "dangling-quote argument not refused cleanly"
+run_only "$tmp/q-lone" '"'
+refused "lone-quote argument not refused cleanly"
+
 exit 0

tests/01_engine-rcfile.test

@@ -0,0 +1,91 @@
+#!/bin/bash
+#
+
+# Config-file alias loading (no network). A .httrackrc in the working directory
+# is read by optinclude_file(), whose cmdl_ins macro inserts each alias-expanded
+# token into the x_argvblk block. That macro used to copy with an unbounded
+# strcpy on a bare char*; it is now bounded (strlcpybuff + cmdl_room over the
+# block capacity). Two properties are checked:
+#   1. The bound does not truncate: a long user-agent alias reaches doit.log
+#      intact. user-agent expands to two tokens (-F <value>), so it exercises
+#      both cmdl_ins insertions.
+#   2. The bound holds under exhaustion: a pathological .httrackrc whose alias
+#      expansions overflow the block aborts cleanly through the htssafe bounds
+#      check (a message naming htsalias.c) instead of overrunning the heap. The
+#      unbounded version segfaulted here.
+
+# set -e with the intentional-nonzero httrack runs guarded explicitly (the
+# crawls below are expected to fail/abort and their status is inspected by hand).
+set -euo pipefail
+
+# Resolve httrack to an absolute path before we cd: PATH may hold a build-relative
+# entry that would not resolve from the temp directory.
+bin=$(command -v httrack) || {
+    echo "FAIL: httrack not found on PATH"
+    exit 1
+}
+case "$bin" in
+/*) ;;
+*) bin="$(cd "$(dirname "$bin")" && pwd)/$(basename "$bin")" ;;
+esac
+
+tmp=$(mktemp -d "${TMPDIR:-/tmp}/httrack_rcfile.XXXXXX") || exit 1
+trap 'rm -rf "$tmp"' EXIT HUP INT QUIT PIPE TERM
+
+# --- 1. alias token survives the bound intact -------------------------------
+d1="$tmp/intact"
+mkdir -p "$d1"
+echo '<html><body>hello</body></html>' >"$d1/index.html"
+
+# optinclude_file() lowercases each config line, so the marker is lowercase to
+# survive the comparison verbatim.
+marker='zzz_rcfile_marker_0123456789_abcdefghijklmnopqrstuvwxyz_intact'
+printf 'user-agent=%s\n' "$marker" >"$d1/.httrackrc"
+
+# Run with no -O so the working-directory .httrackrc is loaded (an -O path makes
+# the engine skip the rc files). Output lands in the temp dir. Guard the run so a
+# nonzero exit is captured for the assertion instead of tripping set -e.
+rc=0
+(cd "$d1" && "$bin" "file://$d1/index.html" --quiet -n >.log 2>&1) || rc=$?
+
+test "$rc" -eq 0 || {
+    echo "FAIL: rc-file crawl exited $rc"
+    exit 1
+}
+test -f "$d1/hts-cache/doit.log" || {
+    echo "FAIL: doit.log not written (rc file not processed)"
+    exit 1
+}
+# A truncated copy would cut the token; require the full -F value.
+grep -q -- "-F $marker" "$d1/hts-cache/doit.log" || {
+    echo "FAIL: user-agent alias missing or truncated in doit.log"
+    head -1 "$d1/hts-cache/doit.log"
+    exit 1
+}
+
+# --- 2. block exhaustion aborts through the bound, not the heap -------------
+d2="$tmp/exhaust"
+mkdir -p "$d2"
+echo '<html><body>hi</body></html>' >"$d2/index.html"
+
+# Each line inserts ~two tokens of ~200 bytes; 400 lines overflow the block's
+# fixed slack (current_size + 32768) many times over, deterministically.
+val=$(printf 'a%.0s' $(seq 1 200))
+for _ in $(seq 1 400); do
+    printf 'user-agent=%s\n' "$val"
+done >"$d2/.httrackrc"
+
+# The process aborts (httrack turns the fatal signal into exit 134 either way),
+# so the exit code does not distinguish the bounded abort from a heap overflow;
+# the stderr diagnostic does. The htssafe bounds check names the offending file.
+# Expected to fail, so the nonzero exit is swallowed; only the log is inspected.
+(cd "$d2" && "$bin" "file://$d2/index.html" --quiet -n >.log 2>&1) || true
+
+grep -Eq "overflow while copying.*htsalias\.c" "$d2/.log" || {
+    echo "FAIL: exhausted rc file did not abort through the htsalias.c bound"
+    echo "(an unbounded copy would overrun the heap here)"
+    tail -3 "$d2/.log"
+    exit 1
+}
+
+exit 0

tests/Makefile.am

@@ -25,6 +25,7 @@ TESTS = \
 	01_engine-idna.test \
 	01_engine-mime.test \
 	01_engine-parse.test \
+	01_engine-rcfile.test \
 	01_engine-simplify.test \
 	01_engine-strsafe.test \
 	02_manpage-regen.test \

tests/Makefile.in

@@ -499,6 +499,7 @@ TESTS = \
 	01_engine-idna.test \
 	01_engine-mime.test \
 	01_engine-parse.test \
+	01_engine-rcfile.test \
 	01_engine-simplify.test \
 	01_engine-strsafe.test \
 	02_manpage-regen.test \