debian: 3.49.10-2, fix buildd FTBFS via quilt patch

3.49.10-1 failed to build on every Debian buildd: tests/28_local-pause
failed instead of skipping when python3 is absent (the local-server tests
need python3, which the minimal buildd chroot lacks). The fix is already
upstream, but the -1 orig is frozen in the archive, so carry it as a quilt
patch and bump to -2. Packaging-only; no upstream code change.

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

.github/workflows/ci.yml

@@ -61,6 +61,50 @@ jobs:
         if: failure()
         run: cat tests/test-suite.log 2>/dev/null || true
 
+  # Reproduce the Debian buildds: they build in a minimal chroot with no
+  # python3, so the local-server tests must SKIP (exit 77), not fail. GitHub
+  # runners ship python3, so every other job hides this path; here we remove it
+  # before `make check`. This is the guard that would have caught the 3.49.10-1
+  # FTBFS (28_local-pause failed instead of skipping when python3 was absent).
+  buildd-no-python3:
+    name: build (no python3, Debian buildd)
+    runs-on: ubuntu-24.04
+    steps:
+      - uses: actions/checkout@v6
+        with:
+          submodules: recursive
+
+      - name: Install build dependencies
+        run: |
+          set -euo pipefail
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends \
+            build-essential autoconf automake libtool autoconf-archive \
+            zlib1g-dev libssl-dev
+
+      - name: Configure
+        run: |
+          set -euo pipefail
+          autoreconf -fi
+          ./configure
+
+      - name: Build
+        run: make -j"$(nproc)"
+
+      - name: Test without python3
+        run: |
+          set -euo pipefail
+          # Hide every python3* so `command -v python3` fails like it does in the
+          # buildd chroot; masking with /bin/false would still resolve.
+          sudo find /usr/bin /usr/local/bin -maxdepth 1 -name 'python3*' \
+            -exec mv {} {}.hidden \;
+          ! command -v python3
+          make check
+
+      - name: Print the test log on failure
+        if: failure()
+        run: cat tests/test-suite.log 2>/dev/null || true
+
   # Portability: build and test on macOS (Darwin/clang) on a native runner --
   # no VM. The tree has no __APPLE__ branches, so Darwin exercises the
   # generic-Unix path on a second libc and kernel. brew's openssl@3 is keg-only,

debian/changelog

@@ -1,3 +1,12 @@
+httrack (3.49.10-2) unstable; urgency=medium
+
+  * Fix FTBFS: tests/28_local-pause failed instead of skipping when python3 is
+    absent (the local-server tests need python3, which the buildds lack). Add
+    patches/skip-local-pause-test-without-python3.patch to guard the test on
+    python3 up front, like its siblings, so it skips cleanly.
+
+ -- Xavier Roche <xavier@debian.org>  Sun, 28 Jun 2026 20:18:46 +0200
+
 httrack (3.49.10-1) unstable; urgency=medium
 
   * New upstream release: new download-pacing and URL-handling options plus a

debian/patches/series

@@ -0,0 +1 @@
+skip-local-pause-test-without-python3.patch

debian/patches/skip-local-pause-test-without-python3.patch

@@ -0,0 +1,29 @@
+Description: skip 28_local-pause when python3 is absent (fixes FTBFS on buildds)
+ The local-server tests skip (exit 77) when python3 is missing, but
+ 28_local-pause wrapped local-crawl.sh in a command substitution that swallowed
+ that skip. On the python3-less buildd chroot the test then ran serverless, both
+ crawls finished in 0s, and the 0s delta was reported as a failure. Guard the
+ test on python3 up front, like its siblings, so it skips cleanly.
+Author: Xavier Roche <roche@httrack.com>
+Origin: upstream, https://github.com/xroche/httrack/pull/445
+Applied-Upstream: https://github.com/xroche/httrack/commit/45279d7
+Forwarded: not-needed
+Last-Update: 2026-06-28
+diff --git a/tests/28_local-pause.test b/tests/28_local-pause.test
+index 8505a75..add95d1 100755
+--- a/tests/28_local-pause.test
++++ b/tests/28_local-pause.test
+@@ -9,6 +9,13 @@ set -e
+ 
+ : "${top_srcdir:=..}"
+ 
++# python3 runs the local server (mirror local-crawl.sh); skip when absent, else
++# run() swallows its exit-77 and the serverless 0s/0s crawl looks like a fail.
++command -v python3 >/dev/null || {
++    echo "python3 not found; skipping local crawl tests"
++    exit 77
++}
++
+ run() { # echoes the wall-clock seconds of one crawl
+     local t0 t1
+     t0=$(date +%s)

src/htsbasiccharsets.sh

@@ -3,12 +3,12 @@
 
 # Change this to download files
 if false; then
-    echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/ISO8859/8859-*.TXT" | lftp
-    echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/PC/CP*.TXT" | lftp
-    echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WINDOWS/CP*.TXT" | lftp
-    echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/EBCDIC/CP*.TXT" | lftp
-    echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/VENDORS/MISC/CP*.TXT" | lftp
-    echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/VENDORS/MISC/KOI8*.TXT" | lftp
+    echo "mget https://www.unicode.org/Public/MAPPINGS/ISO8859/8859-*.TXT" | lftp
+    echo "mget https://www.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/PC/CP*.TXT" | lftp
+    echo "mget https://www.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WINDOWS/CP*.TXT" | lftp
+    echo "mget https://www.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/EBCDIC/CP*.TXT" | lftp
+    echo "mget https://www.unicode.org/Public/MAPPINGS/VENDORS/MISC/CP*.TXT" | lftp
+    echo "mget https://www.unicode.org/Public/MAPPINGS/VENDORS/MISC/KOI8*.TXT" | lftp
     rm -f CP932.TXT CP936.TXT CP949.TXT CP950.TXT
 fi
 

src/htsencoding.c

@@ -30,12 +30,14 @@ Please visit our Website: http://www.httrack.com
 /* Author: Xavier Roche                                         */
 /* ------------------------------------------------------------ */
 
+#include <stdint.h>
+
 #include "htscharset.h"
 #include "htsencoding.h"
 #include "htssafe.h"
 
-/* static int decode_entity(const unsigned int hash, const size_t len);
-*/
+/* static int decode_entity(const uint64_t hash, const size_t len);
+ */
 #include "htsentities.h"
 
 /* hexadecimal conversion */
@@ -50,30 +52,31 @@ static int get_hex_value(char c) {
     return -1;
 }
 
-/* Numerical Recipes,
-   see <http://en.wikipedia.org/wiki/Linear_congruential_generator> */
-#define HASH_PRIME ( 1664525 )
-#define HASH_CONST ( 1013904223 )
-#define HASH_ADD(HASH, C) do {                  \
-    (HASH) *= HASH_PRIME;                       \
-    (HASH) += HASH_CONST;                       \
-    (HASH) += (C);                              \
-  } while(0)
+/* 64-bit FNV-1a; must match htsentities.sh, which keys the entity table on it.
+ */
+#define HASH_INIT 0xcbf29ce484222325ULL
+#define HASH_PRIME 0x100000001b3ULL
+#define HASH_ADD(HASH, C)                                                      \
+  do {                                                                         \
+    (HASH) ^= (unsigned char) (C);                                             \
+    (HASH) *= HASH_PRIME;                                                      \
+  } while (0)
 
 int hts_unescapeEntitiesWithCharset(const char *src, char *dest, const size_t max, const char *charset) {
   size_t i, j, ampStart, ampStartDest;
   int uc;
   int hex;
-  unsigned int hash;
+  uint64_t hash;
 
   assertf(max != 0);
-  for(i = 0, j = 0, ampStart = (size_t) -1, ampStartDest = 0,
-        uc = -1, hex = 0, hash = 0 ; src[i] != '\0' ; i++) {
+  for (i = 0, j = 0, ampStart = (size_t) -1, ampStartDest = 0, uc = -1, hex = 0,
+      hash = HASH_INIT;
+       src[i] != '\0'; i++) {
     /* start of entity */
     if (src[i] == '&') {
       ampStart = i;
       ampStartDest = j;
-      hash = 0;
+      hash = HASH_INIT;
       uc = -1;
     }
     /* inside a potential entity */
@@ -174,14 +177,11 @@ int hts_unescapeEntitiesWithCharset(const char *src, char *dest, const size_t ma
       }
       /* alphanumerical entity */
       else {
-        /* alphanum and not too far ('&thetasym;' is the longest) */
-        if (i <= ampStart + 10 &&
-            (
-             (src[i] >= '0' && src[i] <= '9')
-             || (src[i] >= 'A' && src[i] <= 'Z')
-             || (src[i] >= 'a' && src[i] <= 'z')
-             )
-            ) {
+        /* alphanum, capped at the longest name
+         * '&CounterClockwiseContourIntegral;' (31) */
+        if (i <= ampStart + 31 && ((src[i] >= '0' && src[i] <= '9') ||
+                                   (src[i] >= 'A' && src[i] <= 'Z') ||
+                                   (src[i] >= 'a' && src[i] <= 'z'))) {
           /* compute hash */
           HASH_ADD(hash, (unsigned char) src[i]);
         } else {

src/htsentities.sh

@@ -1,75 +1,92 @@
 #!/bin/bash
 #
+# Regenerate htsentities.h from the WHATWG named character references.
 
-src=html40.txt
-url=http://www.w3.org/TR/1998/REC-html40-19980424/html40.txt
+set -euo pipefail
+
+src=entities.json
+url=https://html.spec.whatwg.org/entities.json
 dest=htsentities.h
 
-(
-    cat <<EOF
-/*
-  -- ${dest} --
-  FILE GENERATED BY $0, DO NOT MODIFY
+# 64-bit FNV-1a of $1, printed as a C constant. Must match the hash in
+# htsencoding.c. The offset basis is stored as its wrapped (signed) bit pattern;
+# bash arithmetic is 64-bit two's complement, so the result is bit-exact.
+fnv1a() {
+    local s=$1 i c h=$((0xcbf29ce484222325))
+    for ((i = 0; i < ${#s}; i++)); do
+        printf -v c '%d' "'${s:i:1}"
+        h=$(((h ^ (c & 0xff)) * 0x100000001b3))
+    done
+    printf '0x%016xULL' "$h"
+}
 
-  We compute the LCG hash
-  (see <http://en.wikipedia.org/wiki/Linear_congruential_generator>)
-  for each entity. We should in theory check using strncmp() that we
-  actually have the correct entity, but this is actually statistically
-  not needed.
+if [ ! -f "$src" ]; then
+    curl -fsS "$url" -o "$src"
+fi
 
-  We may want to do better, but we expect the hash function to be uniform, and
-  let the compiler be smart enough to optimize the switch (for example by
-  checking in log2() intervals)
-  
-  This code has been generated using the evil $0 script.
-*/
+# Keep ';'-terminated single-codepoint names; the ~93 multi-codepoint refs can't
+# fit decode_entity's single-codepoint return and are skipped (left verbatim).
+pairs=$(jq -r '
+    to_entries
+    | map(select((.key | endswith(";")) and (.value.codepoints | length == 1)))
+    | sort_by(.key)
+    | .[] | "\(.key | ltrimstr("&") | rtrimstr(";"))\t\(.value.codepoints[0])"' "$src")
 
-static int decode_entity(const unsigned int hash, const size_t len) {
+# Skipped multi-codepoint names, kept to prove none aliases an emitted hash.
+skipped=$(jq -r '
+    to_entries
+    | map(select((.key | endswith(";")) and (.value.codepoints | length > 1)))
+    | .[] | .key | ltrimstr("&") | rtrimstr(";")' "$src")
+
+cases=""
+emit_hashes=""
+while IFS=$'\t' read -r name cp; do
+    hash=$(fnv1a "$name")
+    cases+="    /* $name */"$'\n'
+    cases+="  case $hash:"$'\n'
+    cases+="    if (len == ${#name}) {"$'\n'
+    cases+="      return $cp;"$'\n'
+    cases+="    }"$'\n'
+    cases+="    break;"$'\n'
+    emit_hashes+="$hash"$'\n'
+done <<<"$pairs"
+
+skip_hashes=""
+while IFS= read -r name; do
+    [ -n "$name" ] && skip_hashes+="$(fnv1a "$name")"$'\n'
+done <<<"$skipped"
+
+# The switch keys on the hash alone, so the dispatch is correct only while every
+# emitted name hashes uniquely; prove it here, no runtime name compare needed.
+dups=$(printf '%s' "$emit_hashes" | sort | uniq -d || true)
+if [ -n "$dups" ]; then
+    echo "FATAL: two entity names share a hash (duplicate switch case); change the hash:" >&2
+    echo "$dups" >&2
+    exit 1
+fi
+# A skipped name colliding with an emitted hash would mis-decode instead of
+# staying verbatim; forbid that too.
+aliased=$(comm -12 <(printf '%s' "$emit_hashes" | sort -u) <(printf '%s' "$skip_hashes" | sort -u) || true)
+if [ -n "$aliased" ]; then
+    echo "FATAL: a skipped multi-codepoint name aliases an emitted hash:" >&2
+    echo "$aliased" >&2
+    exit 1
+fi
+
+cat >"$dest" <<EOF
+/* GENERATED by $0 from the WHATWG named character references
+   (${url}). DO NOT EDIT.
+   Dispatch keys on a 64-bit FNV-1a hash of the entity name; the generator
+   aborts on any hash collision, so no runtime name compare is needed. */
+
+#include <stdint.h>
+
+static int decode_entity(const uint64_t hash, const size_t len) {
   switch(hash) {
-EOF
-    (
-        if test -f ${src}; then
-            cat ${src}
-        else
-            GET "${url}"
-        fi
-    ) |
-        grep -E '^<!ENTITY [a-zA-Z0-9_]' |
-        sed \
-            -e 's/<!ENTITY //' -e "s/[[:space:]][[:space:]]*/ /g" \
-            -e 's/-->$//' \
-            -e 's/\([^ ]*\) CDATA "&#\([^\"]*\);" -- \(.*\)/\1 \2 \3/' |
-        (
-            read -r A
-            while test -n "$A"; do
-                ent="${A%% *}"
-                code=$(echo "$A" | cut -f2 -d' ')
-                # compute hash
-                hash=0
-                i=0
-                a=1664525
-                c=1013904223
-                m="$((1 << 32))"
-                while test "$i" -lt ${#ent}; do
-                    d="$(echo -n "${ent:${i}:1}" | hexdump -v -e '/1 "%d"')"
-                    hash="$((((hash * a) % (m) + d + c) % (m)))"
-                    i=$((i + 1))
-                done
-                echo -e "    /* $A */"
-                echo -e "  case ${hash}u:"
-                echo -e "    if (len == ${#ent} /* && strncmp(ent, \"${ent}\") == 0 */) {"
-                echo -e "      return ${code};"
-                echo -e "    }"
-                echo -e "    break;"
-
-                # next
-                read -r A
-            done
-        )
-    cat <<EOF
-  }
+${cases}  }
   /* unknown */
   return -1;
 }
 EOF
-) >${dest}
+
+echo "wrote $dest ($(grep -c '^  case ' "$dest") entities)" >&2

tests/01_engine-entities.test

@@ -18,6 +18,21 @@ ent '&' '&'
 ent '&lt;&gt;' '<>'
 ent '&eacute;' 'é'
 
+# HTML5 names from the WHATWG set
+ent '&hellip;' '…'
+ent '&bigcup;' '⋃'
+# longest name (31 chars) exercises the name-length cap
+ent '&CounterClockwiseContourIntegral;' '∳'
+# astral codepoint -> 4-byte UTF-8
+ent '&Aopf;' '𝔸'
+# multi-codepoint refs are skipped at generation, so left verbatim
+ent '&fjlig;' '&fjlig;'
+
+# common HTML4 names still decode (regression guard against accidental drops)
+ent '&copy;&reg;&trade;' '©®™'
+ent '&mdash;&ndash;' '—–'
+ent '&alpha;&beta;' 'αβ'
+
 # numeric: decimal and hex
 ent '&#65;&#66;' 'AB'
 ent '&#x41;' 'A'

tests/28_local-pause.test

@@ -9,6 +9,13 @@ set -e
 
 : "${top_srcdir:=..}"
 
+# python3 runs the local server (mirror local-crawl.sh); skip when absent, else
+# run() swallows its exit-77 and the serverless 0s/0s crawl looks like a fail.
+command -v python3 >/dev/null || {
+    echo "python3 not found; skipping local crawl tests"
+    exit 77
+}
+
 run() { # echoes the wall-clock seconds of one crawl
     local t0 t1
     t0=$(date +%s)