Honor the server's Content-Range when resuming a partial download (#198) (#428)

* Honor the server's Content-Range when resuming a partial download (#198)

A resumed download (Range: bytes=N-) may be answered with a 206 whose range
starts before N: block-aligned caches and CDNs routinely round the start down
to a block boundary, and RFC 7233 lets the server pick the range it returns.
httrack ignored the returned Content-Range and blindly appended the 206 body to
the bytes already on disk, so the overlapping bytes were duplicated and the file
grew by the overlap. With timing deciding which files get interrupted (and thus
resumed), this surfaced as a random subset of files corrupted on each run, each
a few bytes too large.

Resume at the server's crange_start instead: ftruncate the partial to that
offset and write the 206 body there (the in-memory branch keeps only that
prefix). When the returned range is unusable (a forward gap, no/garbage
Content-Range, or one that doesn't reach EOF) drop the partial and refetch the
whole file rather than stitch a corrupt one.

Reading the existing crange_start/crange_end/crange fields only, no ABI change.
Driven by tests/24_local-resume-overlap.test: pass 1 interrupts a download
mid-body, pass 2 resumes against a 206 that backs up 8 bytes, and the result
must be byte-identical to the same content fetched whole.

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

* Harden #198 fix: verify the truncate, assert the test hit the resume path

Two follow-ups from review of the resume fix.

If HTS_FTRUNCATE fails the partial could keep a stale tail (only when the
resource shrank between runs, sz > full, so the body write no longer covers
the old end). Check its return and, on failure, drop the partial and refetch
the whole file instead of writing a possibly-corrupt one.

The resume test only compared the resumed bytes against the whole file, which
also passes if httrack silently re-downloads the file with no Range (the bug
never fires). Mark when the server actually serves a resume 206 and assert pass
2 hit that path, so a full re-download fails the test instead of passing it.

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

* tests: run 24_local-resume-overlap under set -e

Follow the golden rule for shell scripts: start with set -e so a non-last
failure can't be masked. Guard the backgrounded-crawl kill/wait spots with
|| true so the expected SIGTERM exit doesn't abort the run.

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

---------

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

AGENTS.md

@@ -33,8 +33,9 @@ the operational checklist: toolchain, invariants, and how to ship a change.
 - Be terse. Comment the why, in English; translate French comments you touch.
 - Strip AI tells from prose (em-dash overuse, rule-of-three, filler, vague
   attributions). Ref: Wikipedia "Signs of AI writing". Claude Code: `/humanizer`.
-- Behavior change → add a test. Fast path: a hidden `httrack -#N` debug
-  subcommand (`htscoremain.c`) driven by a `tests/NN_*.test`, over a slow crawl.
+- Behavior change → add a test. Fast path: a hidden `httrack -#test=NAME` engine
+  self-test (registry in `htsselftest.c`; `-#test` lists them) driven by a
+  `tests/NN_*.test`, over a slow crawl.
 
 ## Review your change adversarially (strongly suggested)
 Before pushing, and when reviewing others, don't skim for bugs:

man/httrack.1

@@ -3,7 +3,7 @@
 .\"
 .\" This file is generated by man/makeman.sh; do not edit by hand.
 .\" SPDX-License-Identifier: GPL-3.0-or-later
-.TH httrack 1 "13 June 2026" "httrack website copier"
+.TH httrack 1 "26 June 2026" "httrack website copier"
 .SH NAME
 httrack \- offline browser : copy websites to a local directory
 .SH SYNOPSIS
@@ -313,12 +313,8 @@ debug HTTP headers in logfile (\-\-debug\-headers)
 .SS Guru options: (do NOT use if possible)
 .IP \-#X
 *use optimized engine (limited memory boundary checks) (\-\-fast\-engine)
-.IP \-#0
-filter test (\-#0 '*.gif' 'www.bar.com/foo.gif') (\-\-debug\-testfilters <param>)
-.IP \-#1
-simplify test (\-#1 ./foo/bar/../foobar)
-.IP \-#2
-type test (\-#2 /foo/bar.php)
+.IP \-#test
+list engine self\-tests (run one with \-#test=NAME [args])
 .IP \-#C
 cache list (\-#C '*.com/spider*.gif' (\-\-debug\-cache <param>)
 .IP \-#R

src/Makefile.am

@@ -56,7 +56,7 @@ whttrackrundir = $(bindir)
 whttrackrun_SCRIPTS = webhttrack
 
 libhttrack_la_SOURCES =  htscore.c htsparse.c htsback.c htscache.c \
-	htscache_selftest.c htsdns_selftest.c \
+	htscache_selftest.c htsdns_selftest.c htsselftest.c \
 	htscatchurl.c htsfilters.c htsftp.c htshash.c coucal/coucal.c \
 	htshelp.c htslib.c htscoremain.c \
 	htsname.c htsrobots.c htstools.c htswizard.c \
@@ -66,7 +66,7 @@ libhttrack_la_SOURCES =  htscore.c htsparse.c htsback.c htscache.c \
 	md5.c \
 	minizip/ioapi.c minizip/mztools.c minizip/unzip.c minizip/zip.c \
 	hts-indextmpl.h htsalias.h htsback.h htsbase.h htssafe.h \
-	htsbasenet.h htsbauth.h htscache.h htscache_selftest.h htsdns_selftest.h htscatchurl.h  \
+	htsbasenet.h htsbauth.h htscache.h htscache_selftest.h htsdns_selftest.h htsselftest.h htscatchurl.h  \
 	htsconfig.h htscore.h htsparse.h htscoremain.h htsdefines.h  \
 	htsfilters.h htsftp.h htsglobal.h htshash.h coucal/coucal.h \
 	htshelp.h htsindex.h htslib.h htsmd5.h \

src/htsback.c

@@ -57,7 +57,10 @@ Please visit our Website: http://www.httrack.com
 // DOS
 #include <process.h>            /* _beginthread, _endthread */
 #endif
+#include <io.h> /* _chsize_s */
+#define HTS_FTRUNCATE(fp, sz) _chsize_s(_fileno(fp), (sz))
 #else
+#define HTS_FTRUNCATE(fp, sz) ftruncate(fileno(fp), (sz))
 #endif
 
 #define VT_CLREOL       "\33[K"
@@ -3774,35 +3777,70 @@ void back_wait(struct_back * sback, httrackp * opt, cache_back * cache,
                     // xxc SI CHUNK VERIFIER QUE CA MARCHE??
                     if (back[i].r.statuscode == 206) {  // on nous envoie un morceau (la fin) coz une partie sur disque!
                       off_t sz = fsize_utf8(back[i].url_sav);
+                      /* RFC 7233: resume at the server's Content-Range start,
+                         not the offset we requested; a server may resume
+                         earlier and appending the overlap duplicates bytes
+                         (#198). */
+                      const LLint resume = back[i].r.crange_start;
+                      const hts_boolean range_ok =
+                          back[i].r.crange > 0 && resume >= 0 &&
+                          resume <= (LLint) sz &&
+                          back[i].r.crange_end + 1 == back[i].r.crange &&
+                          (back[i].r.totalsize < 0 ||
+                           back[i].r.totalsize ==
+                               back[i].r.crange_end - resume + 1);
 
 #if HDEBUG
                       printf("partial content: " LLintP " on disk..\n",
                              (LLint) sz);
 #endif
-                      if (sz >= 0) {
+                      if (sz >= 0 && range_ok) {
                         if (!is_hypertext_mime(opt, back[i].r.contenttype, back[i].url_sav)) {  // pas HTML
                           if (opt->getmode & HTS_GETMODE_NONHTML) {
                             filenote(&opt->state.strc, back[i].url_sav, NULL);  // noter fichier comme connu
                             file_notify(opt, back[i].url_adr, back[i].url_fil,
                                         back[i].url_sav, 0, 1,
                                         back[i].r.notmodified);
-                            back[i].r.out = FOPEN(fconv(catbuff, sizeof(catbuff), back[i].url_sav), "ab");       // append
+                            back[i].r.out =
+                                FOPEN(fconv(catbuff, sizeof(catbuff),
+                                            back[i].url_sav),
+                                      "r+b"); // resume in place
                             if (back[i].r.out && opt->cache != 0) {
-                              back[i].r.is_write = 1;   // écrire
-                              back[i].r.size = sz;      // déja écrit
-                              back[i].r.statuscode = HTTP_OK;   // Forcer 'OK'
+                              back[i].r.is_write = 1;
+                              back[i].r.size = resume; // bytes already on disk
+                              back[i].r.statuscode = HTTP_OK; // force 'OK'
                               if (back[i].r.totalsize >= 0)
-                                back[i].r.totalsize += sz;      // plus en fait
-                              fseek(back[i].r.out, 0, SEEK_END);        // à la fin
-                              /* create a temporary reference file in case of broken mirror */
-                              if (back_serialize_ref(opt, &back[i]) != 0) {
-                                hts_log_print(opt, LOG_WARNING,
-                                              "Could not create temporary reference file for %s%s",
-                                              back[i].url_adr, back[i].url_fil);
-                              }
+                                back[i].r.totalsize += resume; // -> full size
+                              // drop bytes past the resume point; a silent
+                              // failure could leave a stale tail, so on error
+                              // drop the partial and refetch the whole file
+                              if (HTS_FTRUNCATE(back[i].r.out,
+                                                (off_t) resume) != 0) {
+                                fclose(back[i].r.out);
+                                back[i].r.out = NULL;
+                                url_savename_refname_remove(
+                                    opt, back[i].url_adr, back[i].url_fil);
+                                UNLINK(back[i].url_sav);
+                                back[i].status = STATUS_READY;
+                                back_set_finished(sback, i);
+                                strcpybuff(back[i].r.msg,
+                                           "Can not truncate partial file, "
+                                           "restarting");
+                              } else {
+                                fseeko(back[i].r.out, (off_t) resume, SEEK_SET);
+                                /* create a temporary reference file in case of
+                                 * broken mirror */
+                                if (back_serialize_ref(opt, &back[i]) != 0) {
+                                  hts_log_print(opt, LOG_WARNING,
+                                                "Could not create temporary "
+                                                "reference file for %s%s",
+                                                back[i].url_adr,
+                                                back[i].url_fil);
+                                }
 #if HDEBUG
-                              printf("continue interrupted file\n");
+                                printf("continue interrupted file\n");
 #endif
+                              }
                             } else {    // On est dans la m**
                               back[i].status = STATUS_READY;    // terminé (voir plus loin)
                               back_set_finished(sback, i);
@@ -3814,17 +3852,18 @@ void back_wait(struct_back * sback, httrackp * opt, cache_back * cache,
                           FILE *fp =
                             FOPEN(fconv(catbuff, sizeof(catbuff), back[i].url_sav), "rb");
                           if (fp) {
-                            LLint alloc_mem = sz + 1;
+                            LLint alloc_mem = resume + 1;
 
                             if (back[i].r.totalsize >= 0)
                               alloc_mem += back[i].r.totalsize; // AJOUTER RESTANT!
                             if (deleteaddr(&back[i].r)
                                 && (back[i].r.adr =
                                     (char *) malloct((size_t) alloc_mem))) {
-                              back[i].r.size = sz;
+                              back[i].r.size = resume;
                               if (back[i].r.totalsize >= 0)
-                                back[i].r.totalsize += sz;      // plus en fait
-                              if ((fread(back[i].r.adr, 1, sz, fp)) != sz) {
+                                back[i].r.totalsize += resume; // -> full size
+                              if ((fread(back[i].r.adr, 1, (size_t) resume,
+                                         fp)) != (size_t) resume) {
                                 back[i].status = STATUS_READY;  // terminé (voir plus loin)
                                 back_set_finished(sback, i);
                                 strcpybuff(back[i].r.msg,
@@ -3842,14 +3881,30 @@ void back_wait(struct_back * sback, httrackp * opt, cache_back * cache,
                                          "No memory for partial file");
                             }
                             fclose(fp);
-                          } else {      // Argh.. 
+                          } else {                              // open failed
                             back[i].status = STATUS_READY;      // terminé (voir plus loin)
                             back_set_finished(sback, i);
                             strcpybuff(back[i].r.msg,
                                        "Can not open partial file");
                           }
                         }
-                      } else {  // Non trouvé??
+                      } else if (sz >=
+                                 0) { // unusable range -> restart whole file
+                        hts_log_print(opt, LOG_WARNING,
+                                      "Unusable partial-content range for %s%s "
+                                      "(have " LLintP " bytes, got " LLintP
+                                      "-" LLintP "/" LLintP "), restarting",
+                                      back[i].url_adr, back[i].url_fil,
+                                      (LLint) sz, back[i].r.crange_start,
+                                      back[i].r.crange_end, back[i].r.crange);
+                        url_savename_refname_remove(opt, back[i].url_adr,
+                                                    back[i].url_fil);
+                        UNLINK(back[i].url_sav);
+                        back[i].status = STATUS_READY;
+                        back_set_finished(sback, i);
+                        strcpybuff(back[i].r.msg,
+                                   "Unusable partial content, restarting");
+                      } else {                          // partial not found
                         back[i].status = STATUS_READY;  // terminé (voir plus loin)
                         back_set_finished(sback, i);
                         strcpybuff(back[i].r.msg, "Can not find partial file");

src/htscache.c

@@ -220,6 +220,25 @@ struct cache_back_zip_entry {
 	} \
 } while(0)
 
+/* A cache (new.zip) write failed: storage is gone (disk full / dropped share),
+   so the mirror is doomed too. Abort it via exit_xh, don't crash as assertf
+   did. */
+static void cache_zip_write_failed(httrackp *opt, cache_back *cache,
+                                   const char *what, int zErr) {
+  if (!cache->zipWriteFailed) {
+    cache->zipWriteFailed = HTS_TRUE;
+    if (check_fatal_io_errno()) {
+      hts_log_print(opt, LOG_ERROR,
+                    "Mirror aborted: disk full or filesystem problems");
+    } else {
+      hts_log_print(opt, LOG_ERROR,
+                    "Mirror aborted: cache write failed (%s): %s", what,
+                    hts_get_zerror(zErr));
+    }
+  }
+  opt->state.exit_xh = -1; /* fatal: stop the mirror, exit non-zero */
+}
+
 /* Ajout d'un fichier en cache */
 void cache_add(httrackp * opt, cache_back * cache, const htsblk * r,
                const char *url_adr, const char *url_fil, const char *url_save,
@@ -236,6 +255,10 @@ void cache_add(httrackp * opt, cache_back * cache, const htsblk * r,
   const char *url_save_suffix = url_save;
   int zErr;
 
+  /* already failed and aborting; don't touch the broken stream again */
+  if (cache->zipWriteFailed)
+    return;
+
   // robots.txt hack
   if (url_save == NULL) {
     dataincache = 0;            // testing links
@@ -346,9 +369,8 @@ void cache_add(httrackp * opt, cache_back * cache, const htsblk * r,
                                    */
                                   headers, (uInt) strlen(headers), NULL, 0, NULL,       /* comment */
                                   Z_DEFLATED, Z_DEFAULT_COMPRESSION)) != Z_OK) {
-    int zip_zipOpenNewFileInZip_failed = 0;
-
-    assertf(zip_zipOpenNewFileInZip_failed);
+    cache_zip_write_failed(opt, cache, "opening a cache entry", zErr);
+    return;
   }
 
   /* Write data in cache */
@@ -358,9 +380,8 @@ void cache_add(httrackp * opt, cache_back * cache, const htsblk * r,
         if ((zErr =
              zipWriteInFileInZip((zipFile) cache->zipOutput, r->adr,
                                  (int) r->size)) != Z_OK) {
-          int zip_zipWriteInFileInZip_failed = 0;
-
-          assertf(zip_zipWriteInFileInZip_failed);
+          cache_zip_write_failed(opt, cache, "writing to the cache", zErr);
+          return;
         }
       }
     } else {
@@ -381,9 +402,10 @@ void cache_add(httrackp * opt, cache_back * cache, const htsblk * r,
               if ((zErr =
                    zipWriteInFileInZip((zipFile) cache->zipOutput, buff,
                                        (int) nl)) != Z_OK) {
-                int zip_zipWriteInFileInZip_failed = 0;
-
-                assertf(zip_zipWriteInFileInZip_failed);
+                cache_zip_write_failed(opt, cache, "writing to the cache",
+                                       zErr);
+                fclose(fp);
+                return;
               }
             }
           } while(nl > 0);
@@ -397,16 +419,14 @@ void cache_add(httrackp * opt, cache_back * cache, const htsblk * r,
 
   /* Close */
   if ((zErr = zipCloseFileInZip((zipFile) cache->zipOutput)) != Z_OK) {
-    int zip_zipCloseFileInZip_failed = 0;
-
-    assertf(zip_zipCloseFileInZip_failed);
+    cache_zip_write_failed(opt, cache, "closing a cache entry", zErr);
+    return;
   }
 
   /* Flush */
   if ((zErr = zipFlush((zipFile) cache->zipOutput)) != 0) {
-    int zip_zipFlush_failed = 0;
-
-    assertf(zip_zipFlush_failed);
+    cache_zip_write_failed(opt, cache, "flushing the cache", zErr);
+    return;
   }
 }
 

src/htscache_selftest.c

@@ -47,6 +47,7 @@ Please visit our Website: http://www.httrack.com
 #include "htslib.h"
 #include "htszlib.h"
 
+#include <errno.h>
 #include <stdio.h>
 #include <string.h>
 
@@ -316,6 +317,136 @@ static int disk_fallback_selftest(httrackp *opt) {
   return fail;
 }
 
+typedef struct {
+  size_t budget;  /**< bytes allowed through before writes start failing */
+  int fail_errno; /**< errno set on the failing write (ENOSPC, EIO, ...) */
+  int writes;     /**< zwrite call count, to detect re-entry into the stream */
+} writefail_inject;
+
+/* zwrite that copies until the budget runs out, then fails with inj->fail_errno
+   (the #174/#219 condition). Counts calls so the test can prove a flagged cache
+   never re-enters the stream. */
+static uLong selftest_failing_zwrite(voidpf opaque, voidpf stream,
+                                     const void *buf, uLong size) {
+  writefail_inject *inj = (writefail_inject *) opaque;
+
+  inj->writes++;
+  if (inj->budget >= (size_t) size) {
+    inj->budget -= (size_t) size;
+    return (uLong) fwrite(buf, 1, (size_t) size, (FILE *) stream);
+  }
+  errno = inj->fail_errno;
+  return 0; /* short write -> the minizip op returns an error */
+}
+
+/* Open a ZIP whose writes fail past inj->budget, so cache_add() hits an error.
+ */
+static zipFile selftest_open_failing_zip(const char *path,
+                                         writefail_inject *inj) {
+  zlib_filefunc_def ff;
+
+  fill_fopen_filefunc(&ff); /* real fopen/read/seek/close; ignores opaque */
+  ff.zwrite_file = selftest_failing_zwrite;
+  ff.opaque = inj;
+  return zipOpen2(path, APPEND_STATUS_CREATE, NULL, &ff);
+}
+
+/* Store one octet-stream body into `cache` (all-in-cache, body in the ZIP). */
+static void writefail_store(httrackp *opt, cache_back *cache, const char *fil,
+                            const char *body, size_t body_len) {
+  htsblk r;
+  char locbuf[4];
+  char *bodycopy = malloct(body_len);
+
+  hts_init_htsblk(&r);
+  r.statuscode = 200;
+  r.size = (LLint) body_len;
+  strcpybuff(r.msg, "OK");
+  strcpybuff(r.contenttype, "application/octet-stream");
+  locbuf[0] = '\0';
+  r.location = locbuf;
+  r.is_write = 0;
+  memcpy(bodycopy, body, body_len);
+  r.adr = bodycopy;
+  cache_add(opt, cache, &r, "example.com", fil, "example.com/blob.bin", 1,
+            NULL);
+  freet(bodycopy);
+}
+
+/* #174/#219: a failing cache write used to crash via assertf(); it must instead
+   stop the mirror (exit_xh = -1) without crashing. Assert that, plus the cache
+   is flagged and a sibling write doesn't re-enter the broken stream. */
+int cache_write_failure_selftest(httrackp *opt, const char *dir) {
+  int fail = 0;
+  char path[HTS_URLMAXSIZE];
+  /* incompressible + big, so deflate flushes (and fails) mid-write, before
+   * close */
+  static const size_t body_len = 256 * 1024;
+  char *body = malloct(body_len);
+  int phase;
+
+  gen_body(body, body_len, 1 /* incompressible */);
+  fconcat(path, sizeof(path), dir, "/wfail.zip");
+
+  /* phase 0: fail on the body write, fatal errno (ENOSPC, the disk-full
+     branch). phase 1: fail on the open, non-fatal errno (EIO, dropped-share
+     branch). Both must abort the mirror. */
+  for (phase = 0; phase < 2; phase++) {
+    cache_back cache;
+    writefail_inject inj;
+    int writes_after_fail;
+
+    inj.budget = (phase == 0) ? 4096 : 0;
+    inj.fail_errno = (phase == 0) ? ENOSPC : EIO;
+    inj.writes = 0;
+    memset(&cache, 0, sizeof(cache));
+    cache.type = 1;
+    cache.log = stderr;
+    cache.errlog = stderr;
+    cache.hashtable = coucal_new(0);
+    cache.zipOutput = selftest_open_failing_zip(path, &inj);
+    if (cache.zipOutput == NULL) {
+      fprintf(stderr, "cache-writefail: could not open injected ZIP\n");
+      fail++;
+      continue;
+    }
+
+    opt->state.exit_xh = 0; /* clear; the failing write must set it to -1 */
+    writefail_store(opt, &cache, "/blob.bin", body, body_len);
+    if (!cache.zipWriteFailed) {
+      fprintf(stderr, "cache-writefail: phase %d: write error not caught\n",
+              phase);
+      fail++;
+    }
+    if (opt->state.exit_xh != -1) {
+      fprintf(stderr,
+              "cache-writefail: phase %d: mirror not aborted (exit_xh=%d)\n",
+              phase, opt->state.exit_xh);
+      fail++;
+    }
+
+    /* a flagged cache must no-op a sibling write: no further backend write */
+    writes_after_fail = inj.writes;
+    writefail_store(opt, &cache, "/blob2.bin", body, 16);
+    if (inj.writes != writes_after_fail) {
+      fprintf(stderr,
+              "cache-writefail: phase %d: sibling write re-entered the broken "
+              "stream (%d extra backend writes)\n",
+              phase, inj.writes - writes_after_fail);
+      fail++;
+    }
+
+    if (cache.zipOutput != NULL) {
+      zipClose(cache.zipOutput,
+               NULL); /* best-effort; may fail on the backend */
+      cache.zipOutput = NULL;
+    }
+  }
+
+  freet(body);
+  return fail;
+}
+
 int cache_selftests(httrackp *opt, const char *dir) {
   int failures = 0;
   cache_back cache;

src/htscache_selftest.h

@@ -52,6 +52,10 @@ int cache_selftests(httrackp *opt, const char *dir);
    committed file, never by the test). Returns the failed-check count. */
 int cache_golden_selftest(httrackp *opt, const char *dir, int regen);
 
+/* #174/#219: assert a failing cache write aborts the mirror cleanly instead of
+   crashing. Returns the failed-check count. */
+int cache_write_failure_selftest(httrackp *opt, const char *dir);
+
 #endif
 
 #endif

src/htscore.h

@@ -214,6 +214,8 @@ struct cache_back {
   cache_back_zip_entry *zipEntries;
   int zipEntriesOffs;
   int zipEntriesCapa;
+  hts_boolean
+      zipWriteFailed; /**< a cache write failed; stop touching the stream */
 };
 
 #ifndef HTS_DEF_FWSTRUCT_hash_struct

src/htshelp.c

@@ -646,9 +646,7 @@ void help(const char *app, int more) {
   infomsg("");
   infomsg("Guru options: (do NOT use if possible)");
   infomsg(" #X *use optimized engine (limited memory boundary checks)");
-  infomsg(" #0  filter test (-#0 '*.gif' 'www.bar.com/foo.gif')");
-  infomsg(" #1  simplify test (-#1 ./foo/bar/../foobar)");
-  infomsg(" #2  type test (-#2 /foo/bar.php)");
+  infomsg(" #test  list engine self-tests (run one with -#test=NAME [args])");
   infomsg(" #C  cache list (-#C '*.com/spider*.gif'");
   infomsg(" #R  cache repair (damaged cache)");
   infomsg(" #d  debug parser");

src/htslib.c

@@ -4177,9 +4177,10 @@ HTSEXT_API hts_boolean get_httptype_sized(httrackp *opt, char *s, size_t ssize,
     /* Check html -> text/html */
     const char *a = fil + strlen(fil) - 1;
 
-    while((*a != '.') && (*a != '/') && (a > fil))
+    /* a < fil when fil is empty: bound before dereferencing */
+    while ((a > fil) && (*a != '.') && (*a != '/'))
       a--;
-    if (*a == '.' && strlen(a) < 32) {
+    if (a >= fil && *a == '.' && strlen(a) < 32) {
       int j = 0;
 
       a++;

src/htsselftest.h

@@ -0,0 +1,52 @@
+/* ------------------------------------------------------------ */
+/*
+HTTrack Website Copier, Offline Browser for Windows and Unix
+Copyright (C) 2026 Xavier Roche and other contributors
+
+SPDX-License-Identifier: GPL-3.0-or-later
+
+This program is free software: you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation, either version 3 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License
+along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+Ethical use: we kindly ask that you NOT use this software to harvest email
+addresses or to collect any other private information about people. Doing so
+would dishonor our work and waste the many hours we have spent on it.
+
+Please visit our Website: http://www.httrack.com
+*/
+
+/* ------------------------------------------------------------ */
+/* File: htsselftest.h                                          */
+/*       named dispatch for the hidden engine self-tests        */
+/* Author: Xavier Roche                                         */
+/* ------------------------------------------------------------ */
+
+#ifndef HTSSELFTEST_DEFH
+#define HTSSELFTEST_DEFH
+
+#ifdef HTS_INTERNAL_BYTECODE
+
+#ifndef HTS_DEF_FWSTRUCT_httrackp
+#define HTS_DEF_FWSTRUCT_httrackp
+typedef struct httrackp httrackp;
+#endif
+
+/* Run engine self-test `name` over the positional args argv[0..argc-1], or list
+   the available tests when name is NULL, empty, or "list". Prints the result;
+   returns the process exit code (0 == success). The caller owns option cleanup.
+   Reached through the hidden `httrack -#test[=NAME ...]` subcommand. */
+int hts_selftest(httrackp *opt, const char *name, int argc, char **argv);
+
+#endif
+
+#endif

tests/01_engine-cache-golden.test

@@ -4,7 +4,7 @@
 # POSIX /bin/sh on some platforms (e.g. macOS), so avoid bashisms and GNU-only
 # tool flags despite the #!/bin/bash above.
 
-# Golden cache-format regression test (driven by 'httrack -#B <dir>').
+# Golden cache-format regression test (driven by 'httrack -#test=cache-golden <dir>').
 #
 # 01_engine-cache.test writes the cache with the same build it reads back (a
 # round-trip), so it cannot catch a read-path or ZIP-format regression where
@@ -13,7 +13,7 @@
 # byte-exact.
 #
 # Regenerate the fixture after a deliberate format change with
-# 'httrack -#B <dir> regen', then copy <dir>/hts-cache/new.zip over the
+# 'httrack -#test=cache-golden <dir> regen', then copy <dir>/hts-cache/new.zip over the
 # committed file.
 
 set -eu
@@ -37,11 +37,11 @@ trap 'rm -rf "$dir"' EXIT
 mkdir -p "$dir/hts-cache"
 cp "$fixture/hts-cache/new.zip" "$dir/hts-cache/new.zip"
 
-out=$(httrack -#B "$dir")
+out=$(httrack -#test=cache-golden "$dir")
 
 # Match the exact success line: the read must have found and verified every
-# entry, not merely failed to enter the mode (a bad -#B falls through to the
-# usage screen, which also exits non-zero but never prints this).
+# entry, not merely failed to enter the mode (a renamed/removed test prints the
+# registry to stderr, which also exits non-zero but never prints this).
 test "$out" = "cache-golden: OK" || {
     echo "expected 'cache-golden: OK', got: $out" >&2
     exit 1

tests/01_engine-cache-writefail.test

@@ -0,0 +1,24 @@
+#!/bin/bash
+#
+# Keep this POSIX-portable: the harness runs it via $(BASH), which is a plain
+# POSIX /bin/sh on some platforms (e.g. macOS), so avoid bashisms and GNU-only
+# tool flags despite the #!/bin/bash above.
+
+# Cache write-failure handling (httrack -#test=cache-writefail <dir>). #174/#219.
+# A failing new.zip write (disk full) used to crash the process via assertf; it
+# must instead stop the mirror with a fatal error (exit_xh=-1), no crash. The
+# self-test asserts that; reverting the fix makes -#test=cache-writefail abort (SIGABRT) and fail.
+
+set -eu
+
+dir=$(mktemp -d)
+trap 'rm -rf "$dir"' EXIT
+
+out=$(httrack -#test=cache-writefail "$dir")
+
+# Match the exact success line (error logs also go to stdout); a renamed/removed
+# test prints the registry to stderr, which exits non-zero but never prints this.
+printf '%s\n' "$out" | grep -qx "cache-writefail: OK" || {
+    echo "expected 'cache-writefail: OK', got: $out" >&2
+    exit 1
+}

tests/01_engine-cache.test

@@ -4,7 +4,7 @@
 # POSIX /bin/sh on some platforms (e.g. macOS), so avoid bashisms and GNU-only
 # tool flags despite the #!/bin/bash above.
 
-# Cache create/read/update logic (driven by 'httrack -#A <dir>').
+# Cache create/read/update logic (driven by 'httrack -#test=cache <dir>').
 #
 # The in-process self-test stores several hand-crafted edge entries (normal
 # HTML, an empty redirect with a near-limit location, a non-HTML body kept via
@@ -20,13 +20,13 @@ set -eu
 dir=$(mktemp -d)
 trap 'rm -rf "$dir"' EXIT
 
-# Like the other -# debug modes, a trailing token (the working directory) is
-# required; a bare '-#A' falls through to the usage screen.
-out=$(httrack -#A "$dir")
+# The working directory is a required argument; without it the test prints a
+# usage line to stderr and returns non-zero.
+out=$(httrack -#test=cache "$dir")
 
 # Match the exact success line, so the test cannot pass for an unrelated reason
-# (e.g. the -#A mode being gone and falling through to the usage screen, which
-# also exits non-zero but never prints this).
+# (e.g. the cache test being gone, which prints the registry to stderr but
+# never prints this line).
 test "$out" = "cache-selftest: OK" || {
     echo "expected 'cache-selftest: OK', got: $out" >&2
     exit 1

tests/01_engine-charset.test

@@ -4,13 +4,13 @@
 set -euo pipefail
 
 # charset -> UTF-8 conversion (hts_convertStringToUTF8).
-# -#3 <charset> <string> prints the string re-decoded from <charset> as UTF-8.
+# -#test=charset <charset> <string> prints the string re-decoded from <charset> as UTF-8.
 conv() {
-    test "$(httrack -O /dev/null -#3 "$1" "$2")" == "$3" || exit 1
+    test "$(httrack -O /dev/null -#test=charset "$1" "$2")" == "$3" || exit 1
 }
 # crash probe: malformed input must exit cleanly, not abort.
 runs() {
-    httrack -O /dev/null -#3 "$1" "$2" >/dev/null 2>&1 || exit 1
+    httrack -O /dev/null -#test=charset "$1" "$2" >/dev/null 2>&1 || exit 1
 }
 
 # the source bytes below are UTF-8 (this file is UTF-8); "café" is 0x63 61 66 C3 A9.
@@ -31,7 +31,7 @@ conv 'us-ascii' 'hello' 'hello'
 # unknown charset: ASCII passes through unchanged, but non-ASCII input cannot be
 # decoded and yields empty output (an error is printed to stderr).
 conv 'no-such-charset-xyz' 'abc' 'abc'
-test "$(httrack -O /dev/null -#3 'no-such-charset-xyz' 'café' 2>/dev/null)" == "" || exit 1
+test "$(httrack -O /dev/null -#test=charset 'no-such-charset-xyz' 'café' 2>/dev/null)" == "" || exit 1
 
 # malformed UTF-8 (lone continuation byte, truncated lead byte) must not crash
 runs 'utf-8' $'\x80'

tests/01_engine-cookies.test

@@ -1,14 +1,15 @@
 #!/bin/bash
 #
 # Issue #151 guard: the request Cookie header must be bare RFC 6265 name=value
-# pairs, no $Version/$Path attributes. Driven by the 'httrack -#Q' selftest.
+# pairs, no $Version/$Path attributes. Driven by the 'httrack -#test=cookies' selftest.
 
 set -eu
 
-# A trailing token is required; a bare '-#Q' falls through to the usage screen.
-out=$(httrack -#Q run)
+# 'run' is an ignored placeholder argument.
+out=$(httrack -#test=cookies run)
 
-# Exact-match the success line so a fall-through to usage can't pass the test.
+# Exact-match the success line so a renamed/removed test (it prints the registry
+# to stderr) can't pass.
 test "$out" = "cookie-header: OK" || {
     echo "expected 'cookie-header: OK', got: $out" >&2
     exit 1

tests/01_engine-copyopt.test

@@ -2,15 +2,16 @@
 #
 # Regression guard for the unsigned-enum sentinel trap: copy_htsopt's
 # `if (from->X > -1)` guard is always false for unsigned hts_boolean fields, so
-# they silently stop being copied. Driven by the in-process 'httrack -#9' test.
+# they silently stop being copied. Driven by the in-process 'httrack -#test=copyopt' test.
 # Keep POSIX-portable (harness runs it via $(BASH), a plain /bin/sh on macOS).
 
 set -eu
 
-# A trailing token is required; a bare '-#9' falls through to the usage screen.
-out=$(httrack -#9 run)
+# 'run' is an ignored placeholder argument.
+out=$(httrack -#test=copyopt run)
 
-# Exact-match the success line so a fall-through to usage can't pass the test.
+# Exact-match the success line so a renamed/removed test (it prints the registry
+# to stderr) can't pass.
 test "$out" = "copy-htsopt: OK" || {
     echo "expected 'copy-htsopt: OK', got: $out" >&2
     exit 1

tests/01_engine-dns.test

@@ -5,9 +5,8 @@ set -euo pipefail
 
 # DNS resolver/cache self-test: a mock getaddrinfo (no network) checks address
 # family, single-address selection, the -@i4/-@i6 family filter, and cache reuse.
-# The trailing token is required, like the other -# selftests, so a bare command
-# line isn't treated as "no arguments" and routed to the usage screen.
-out=$(httrack -#D run)
+# 'run' is an ignored placeholder argument.
+out=$(httrack -#test=dns run)
 
 test "$out" = "dns-selftest: OK" || {
     echo "expected 'dns-selftest: OK', got: $out" >&2

tests/01_engine-entities.test

@@ -4,13 +4,13 @@
 set -euo pipefail
 
 # HTML entity unescaping (hts_unescapeEntitiesWithCharset).
-# -#6 <string> prints the string with entities decoded (UTF-8 output).
+# -#test=entities <string> prints the string with entities decoded (UTF-8 output).
 ent() {
-    test "$(httrack -O /dev/null -#6 "$1")" == "$2" || exit 1
+    test "$(httrack -O /dev/null -#test=entities "$1")" == "$2" || exit 1
 }
 # crash probe: malformed input must exit cleanly, not abort.
 runs() {
-    httrack -O /dev/null -#6 "$1" >/dev/null 2>&1 || exit 1
+    httrack -O /dev/null -#test=entities "$1" >/dev/null 2>&1 || exit 1
 }
 
 # named entities

tests/01_engine-filter.test

@@ -4,13 +4,13 @@
 set -euo pipefail
 
 # wildcard filter engine (strjoker), the core of +/- include/exclude rules.
-# -#0 <filter> <string> prints "<string> does match <filter>" or "... does NOT match ...".
+# -#test=filter <filter> <string> prints "<string> does match <filter>" or "... does NOT match ...".
 
 match() {
-    test "$(httrack -O /dev/null -#0 "$1" "$2")" == "$2 does match $1" || exit 1
+    test "$(httrack -O /dev/null -#test=filter "$1" "$2")" == "$2 does match $1" || exit 1
 }
 nomatch() {
-    test "$(httrack -O /dev/null -#0 "$1" "$2")" == "$2 does NOT match $1" || exit 1
+    test "$(httrack -O /dev/null -#test=filter "$1" "$2")" == "$2 does NOT match $1" || exit 1
 }
 
 # bare star matches everything

tests/01_engine-hashtable.test

@@ -3,5 +3,7 @@
 
 set -euo pipefail
 
-# httrack internal hashtable autotest on 100K keys
-httrack -#7 100000
+# httrack internal hashtable autotest on 100K keys. Assert the success line (on
+# stderr) so a misrouted registry entry can't pass on exit code alone.
+out=$(httrack -#test=hashtable 100000 2>&1)
+printf '%s\n' "$out" | grep -q "all hashtable tests were successful!" || exit 1

tests/01_engine-idna.test

@@ -3,13 +3,13 @@
 
 set -euo pipefail
 
-# IDNA / punycode encode (-#4) and decode (-#5). This code has a CVE history,
+# IDNA / punycode encode (-#test=idna-encode) and decode (-#test=idna-decode). This code has a CVE history,
 # so the edge cases below cover passthrough, round-trips, and malformed input.
 
-enc() { test "$(httrack -O /dev/null -#4 "$1")" == "$2" || exit 1; }
-dec() { test "$(httrack -O /dev/null -#5 "$1")" == "$2" || exit 1; }
+enc() { test "$(httrack -O /dev/null -#test=idna-encode "$1")" == "$2" || exit 1; }
+dec() { test "$(httrack -O /dev/null -#test=idna-decode "$1")" == "$2" || exit 1; }
 # crash probe: malformed ACE input must exit cleanly, not abort.
-runs() { httrack -O /dev/null -#5 "$1" >/dev/null 2>&1 || exit 1; }
+runs() { httrack -O /dev/null -#test=idna-decode "$1" >/dev/null 2>&1 || exit 1; }
 
 # encode
 enc 'www.café.com' 'www.xn--caf-dma.com'

tests/01_engine-mime.test

@@ -4,13 +4,13 @@
 set -euo pipefail
 
 # MIME type guessing from extension (get_httptype / give_mimext).
-# -#2 <path> prints "<path> is '<mime>'" then "and its local type is '.<ext>'".
+# -#test=mime <path> prints "<path> is '<mime>'" then "and its local type is '.<ext>'".
 
 mime() {
-    test "$(httrack -O /dev/null -#2 "$1" | head -1)" == "$1 is '$2'" || exit 1
+    test "$(httrack -O /dev/null -#test=mime "$1" | head -1)" == "$1 is '$2'" || exit 1
 }
 unknown() {
-    test "$(httrack -O /dev/null -#2 "$1" | head -1)" == "$1 is of an unknown MIME type" || exit 1
+    test "$(httrack -O /dev/null -#test=mime "$1" | head -1)" == "$1 is of an unknown MIME type" || exit 1
 }
 
 mime '/a/b.html' 'text/html'

tests/01_engine-relative.test

@@ -8,7 +8,7 @@ set -euo pipefail
 # relative path from <curr>'s directory to <link>
 rel() {
     local got
-    got=$(httrack -O /dev/null -#l "$1" "$2")
+    got=$(httrack -O /dev/null -#test=relative "$1" "$2")
     test "$got" == "relative=$3" ||
         {
             echo "FAIL rel($1, $2): got '$got' want 'relative=$3'"
@@ -19,7 +19,7 @@ rel() {
 # resolve <link> against origin <adr>/<fil> -> adr=.. fil=..
 ident() {
     local got
-    got=$(httrack -O /dev/null -#i "$1" "$2" "$3")
+    got=$(httrack -O /dev/null -#test=resolve "$1" "$2" "$3")
     test "$got" == "$4" ||
         {
             echo "FAIL ident($1, $2, $3): got '$got' want '$4'"

tests/01_engine-savename.test

@@ -3,11 +3,11 @@
 
 set -euo pipefail
 
-# Local save-name extension resolution (url_savename via -#N <fil> <content-type>).
+# Local save-name extension resolution (url_savename via -#test=savename <fil> <content-type>).
 # Asserts on the basename of "savename: <path>".
 
 name() {
-    out="$(httrack -O /dev/null -#N "$1" "$2" | sed -n 's/^savename: //p')"
+    out="$(httrack -O /dev/null -#test=savename "$1" "$2" | sed -n 's/^savename: //p')"
     test "${out##*/}" == "$3" || {
         echo "FAIL: '$1' '$2' -> '$out' (want '$3')"
         exit 1

tests/01_engine-selftest-dispatch.test

@@ -0,0 +1,17 @@
+#!/bin/bash
+#
+# The -#test dispatch itself: a bare -#test lists the registry, and an unknown
+# name errors (non-zero, diagnostic) instead of silently passing.
+
+set -eu
+
+# Bare -#test lists known tests (printed to stderr).
+list=$(httrack -#test 2>&1)
+printf '%s\n' "$list" | grep -q "filter" || exit 1
+printf '%s\n' "$list" | grep -q "cache-writefail" || exit 1
+
+# Unknown name: non-zero exit + diagnostic, and no test result line.
+rc=0
+err=$(httrack -#test=bogus 2>&1) || rc=$?
+test "$rc" -ne 0 || exit 1
+printf '%s\n' "$err" | grep -q "Unknown self-test" || exit 1

tests/01_engine-simplify.test

@@ -5,7 +5,7 @@ set -euo pipefail
 
 # path simplify engine (fil_simplifie): collapses ./ and ../ segments.
 simp() {
-    test "$(httrack -O /dev/null -#1 "$1")" == "simplified=$2" || exit 1
+    test "$(httrack -O /dev/null -#test=simplify "$1")" == "simplified=$2" || exit 1
 }
 
 simp './foo/bar/' 'foo/bar/'

tests/01_engine-strsafe.test

@@ -3,23 +3,22 @@
 
 set -euo pipefail
 
-# htssafe.h bounded string operations (driven by 'httrack -#8').
+# htssafe.h bounded string operations (driven by 'httrack -#test=strsafe').
 
 # Success path: every bounded op (strcpybuff/strcatbuff/strncatbuff/strlcpybuff)
-# must behave correctly. Like the other -# debug modes, a trailing token is
-# required (a bare '-#8' falls through to the usage screen).
+# must behave correctly. 'run' selects the success path (vs the overflow modes).
 rc=0
-out=$(httrack -#8 run) || rc=$?
+out=$(httrack -#test=strsafe run) || rc=$?
 test "$rc" -eq 0 || exit 1
 test "$out" == "strsafe: OK" || exit 1
 
 # Overflow path: an over-capacity write into a sized buffer must be caught by
 # the bounded macro and abort the process, not be silently truncated/completed.
 # Assert the htssafe abort signature specifically, so the test cannot pass for
-# an unrelated reason (e.g. the -#8 mode being gone and falling through to the
-# usage screen, which also exits non-zero).
+# an unrelated reason (e.g. the strsafe test being gone, which prints the
+# registry to stderr and also exits non-zero).
 # the bounded macro aborts (non-zero exit), so don't let set -e trip on it
-err=$(httrack -#8 overflow "this string is far too long for the buffer" 2>&1) || true
+err=$(httrack -#test=strsafe overflow "this string is far too long for the buffer" 2>&1) || true
 case "$err" in
 *"strsafe: NOT aborted"*)
     echo "over-capacity write was NOT caught" >&2
@@ -36,7 +35,7 @@ esac
 # capacity (4 bytes into a 4-byte buffer), so this also pins the boundary: a
 # '<=' off-by-one in the capacity check would let it through (and print "NOT
 # aborted"). Match the specific htsbuff abort message, not just any assert.
-err=$(httrack -#8 overflow-buff "abcd" 2>&1) || true
+err=$(httrack -#test=strsafe overflow-buff "abcd" 2>&1) || true
 case "$err" in
 *"strsafe: NOT aborted"*)
     echo "htsbuff over-capacity write was NOT caught" >&2

tests/22_local-broken-size.test

@@ -0,0 +1,17 @@
+#!/bin/bash
+# Issues #32/#41: a Content-Length that disagrees with the body warns "bogus
+# state (broken size)" and skips the cache; -%B (tolerant) accepts it.
+
+: "${top_srcdir:=..}"
+
+# Default: warn, but the file is still written.
+bash "$top_srcdir/tests/local-crawl.sh" --errors 0 \
+    --found 'size/oversize.bin' \
+    --log-found 'bogus state \(broken size' \
+    httrack 'BASEURL/size/index.html'
+
+# -%B (tolerant): no warning, file written.
+bash "$top_srcdir/tests/local-crawl.sh" --errors 0 \
+    --found 'size/oversize.bin' \
+    --log-not-found 'bogus state' \
+    httrack 'BASEURL/size/index.html' '-%B'

tests/23_local-errpage.test

@@ -0,0 +1,19 @@
+#!/bin/bash
+# Issue #17: with "no error pages" (-o0), 4xx/5xx bodies must not be written;
+# a genuine 0-byte 200 stays. Default (-o1) writes the error page. (#17's purge
+# half also does not reproduce; the purge path is not exercised here.)
+set -e
+
+: "${top_srcdir:=..}"
+
+# -o0: 404 suppressed, good page and the legit 0-byte 200 kept.
+bash "$top_srcdir/tests/local-crawl.sh" --errors 1 \
+    --found 'errpage/good.html' \
+    --found 'errpage/empty.html' \
+    --not-found 'errpage/missing.html' \
+    httrack 'BASEURL/errpage/index.html' '-o0'
+
+# Control -o1 (default): the 404 error page is written.
+bash "$top_srcdir/tests/local-crawl.sh" --errors 1 \
+    --found 'errpage/missing.html' \
+    httrack 'BASEURL/errpage/index.html' '-o1'

tests/24_local-resume-overlap.test

@@ -0,0 +1,109 @@
+#!/bin/bash
+# Issue #198: on a resumed download the server may answer the Range with a 206
+# that starts *before* the offset we asked for (block-aligned ranges). httrack
+# must honor the returned Content-Range, not blindly append, or the overlap
+# bytes get duplicated and the file grows (corrupt PDFs). Pass 1 interrupts
+# flaky.bin mid-body (partial + temp-ref); pass 2 resumes against a 206 that
+# backs up 8 bytes. The result must equal the same bytes fetched whole (full.bin).
+set -eu
+
+: "${top_srcdir:=..}"
+testdir=$(cd "$(dirname "$0")" && pwd)
+server="${testdir}/local-server.py"
+
+command -v python3 >/dev/null || ! echo "python3 not found; skipping" || exit 77
+
+tmpdir=$(mktemp -d "${TMPDIR:-/tmp}/httrack_198.XXXXXX") || exit 1
+serverpid=
+crawlpid=
+cleanup() {
+    if test -n "$crawlpid"; then kill -9 "$crawlpid" 2>/dev/null || true; fi
+    if test -n "$serverpid"; then
+        kill "$serverpid" 2>/dev/null || true
+        wait "$serverpid" 2>/dev/null || true
+    fi
+    rm -rf "$tmpdir"
+}
+trap cleanup EXIT HUP INT QUIT PIPE TERM
+
+# OVERLAP_COUNTER gets a byte per flaky.bin request so pass 1 knows when to interrupt.
+serverlog="${tmpdir}/server.log"
+counter="${tmpdir}/hits"
+resumed="${tmpdir}/resumed" # gets a byte when the server serves a resume 206
+OVERLAP_COUNTER="$counter" OVERLAP_RESUMED="$resumed" \
+    python3 "$server" --root "${testdir}/server-root" \
+    >"$serverlog" 2>&1 &
+serverpid=$!
+port=
+for _ in $(seq 1 50); do
+    line=$(head -n1 "$serverlog" 2>/dev/null)
+    if test "${line%% *}" == "PORT"; then
+        port="${line#PORT }"
+        break
+    fi
+    kill -0 "$serverpid" 2>/dev/null || {
+        echo "server exited early: $(cat "$serverlog")"
+        exit 1
+    }
+    sleep 0.1
+done
+test -n "$port" || {
+    echo "could not discover server port"
+    exit 1
+}
+base="http://127.0.0.1:${port}"
+
+which httrack >/dev/null || {
+    echo "could not find httrack"
+    exit 1
+}
+out="${tmpdir}/crawl"
+common=(-O "$out" --quiet --disable-security-limits --robots=0 --timeout=30 --retries=0 -c1)
+refdir="${out}/hts-cache/ref"
+
+# pass 1: interrupt once flaky.bin's prefix is streaming (partial + temp-ref).
+printf '[pass 1: interrupt flaky.bin] ..\t'
+httrack "${common[@]}" "${base}/overlap/index.html" >"${tmpdir}/log1" 2>&1 &
+crawlpid=$!
+for _ in $(seq 1 300); do
+    test -s "$counter" && break
+    kill -0 "$crawlpid" 2>/dev/null || break
+    sleep 0.1
+done
+sleep 0.5
+kill -TERM "$crawlpid" 2>/dev/null || true
+wait "$crawlpid" 2>/dev/null || true
+crawlpid=
+test -n "$(find "$refdir" -name '*.ref' 2>/dev/null)" || {
+    echo "FAIL: no temp-ref survived pass 1; cannot drive the resume"
+    exit 1
+}
+echo "OK (temp-ref present)"
+
+# pass 2: --continue -> resume Range -> 206 that starts 8 bytes early.
+printf '[pass 2: resume flaky.bin] ..\t'
+httrack "${common[@]}" --continue "${base}/overlap/index.html" >"${tmpdir}/log2" 2>&1 || true
+echo "OK"
+
+# Guard against a silent full re-download: the byte-compare below only tests the
+# fix if pass 2 actually went through the resume Range -> 206 path.
+printf '[resume path was exercised] ..\t'
+if ! test -s "$resumed"; then
+    echo "FAIL: pass 2 never triggered a resume 206; the overlap fix was not exercised"
+    exit 1
+fi
+echo "OK"
+
+printf '[resumed file is not corrupted] ..\t'
+dir=$(find "$out" -maxdepth 1 -type d -name '127.0.0.1*' | head -1)
+flaky="${dir}/overlap/flaky.bin"
+full="${dir}/overlap/full.bin"
+if ! test -f "$flaky" || ! test -f "$full"; then
+    echo "FAIL: flaky.bin or full.bin missing after pass 2"
+    exit 1
+fi
+if ! cmp -s "$flaky" "$full"; then
+    echo "FAIL: resumed flaky.bin ($(wc -c <"$flaky")) != full.bin ($(wc -c <"$full")); overlap duplicated"
+    exit 1
+fi
+echo "OK ($(wc -c <"$flaky") bytes, byte-identical)"

tests/Makefile.am

@@ -26,6 +26,7 @@ TESTS = \
 	00_runnable.test \
 	01_engine-cache.test \
 	01_engine-cache-golden.test \
+	01_engine-cache-writefail.test \
 	01_engine-charset.test \
 	01_engine-cmdline.test \
 	01_engine-cookies.test \
@@ -41,6 +42,7 @@ TESTS = \
 	01_engine-rcfile.test \
 	01_engine-relative.test \
 	01_engine-savename.test \
+	01_engine-selftest-dispatch.test \
 	01_engine-simplify.test \
 	01_engine-strsafe.test \
 	02_manpage-regen.test \
@@ -61,6 +63,9 @@ TESTS = \
 	18_local-update.test \
 	19_local-connect-fallback.test \
 	20_local-resume-loop.test \
-	21_local-intl-update.test
+	21_local-intl-update.test \
+	22_local-broken-size.test \
+	23_local-errpage.test \
+	24_local-resume-overlap.test
 
 CLEANFILES = check-network_sh.cache

tests/local-crawl.sh

@@ -14,7 +14,9 @@
 # Usage:
 #   bash local-crawl.sh [--tls] [--root DIR] \
 #       --errors N --files N --found PATH ... --directory PATH ... \
+#       --log-found REGEX ... --log-not-found REGEX ... \
 #       httrack BASEURL/some/path [httrack-args...]
+# --log-found/--log-not-found grep (ERE) the crawl's hts-log.txt.
 
 set -u
 
@@ -107,7 +109,7 @@ while test "$pos" -lt "$nargs"; do
         audit+=("${args[$pos]}" "${args[$((pos + 1))]}")
         pos=$((pos + 1))
         ;;
-    --found | --not-found | --directory)
+    --found | --not-found | --directory | --log-found | --log-not-found)
         audit+=("${args[$pos]}" "${args[$((pos + 1))]}")
         pos=$((pos + 1))
         ;;
@@ -257,6 +259,22 @@ while test "$i" -lt "${#audit[@]}"; do
             exit 1
         fi
         ;;
+    --log-found)
+        i=$((i + 1))
+        info "checking log matches ${audit[$i]}"
+        if grep -aqE "${audit[$i]}" "${out}/hts-log.txt"; then result "OK"; else
+            result "not in log"
+            exit 1
+        fi
+        ;;
+    --log-not-found)
+        i=$((i + 1))
+        info "checking log lacks ${audit[$i]}"
+        if grep -aqE "${audit[$i]}" "${out}/hts-log.txt"; then
+            result "present in log"
+            exit 1
+        else result "OK"; fi
+        ;;
     esac
     i=$((i + 1))
 done

tests/local-server.py

@@ -225,6 +225,104 @@ class Handler(SimpleHTTPRequestHandler):
         self.send_header("Content-Length", "0")
         self.end_headers()
 
+    # 206 resume must honor the server's Content-Range, not the offset we asked
+    # for (#198): a server resuming a few bytes *before* the request must not
+    # leave httrack duplicating the overlap onto the partial. flaky.bin
+    # interrupts once then resumes OVERLAP_EARLY bytes early; full.bin serves
+    # the identical bytes in one shot, so the test can compare the two.
+    OVERLAP_BLOB = b"%PDF-1.4\n" + bytes((i * 37 + 11) % 256 for i in range(8000))
+    OVERLAP_EARLY = 8
+    OVERLAP_PREFIX_LEN = 4000  # flushed before the stall
+    _overlap_started = False
+
+    def route_overlap_index(self):
+        self.send_html('\t<a href="flaky.bin">flaky</a>\n\t<a href="full.bin">full</a>')
+
+    def route_overlap_full(self):
+        self.send_raw(self.OVERLAP_BLOB, "application/octet-stream")
+
+    def route_overlap(self):
+        counter = os.environ.get("OVERLAP_COUNTER")
+        if counter:
+            with open(counter, "a") as fp:
+                fp.write("x")
+        blob = self.OVERLAP_BLOB
+        rng = self.headers.get("Range")
+        # First GET: stream a prefix then stall, so the crawl can be interrupted
+        # mid-body (partial + temp-ref on disk).
+        if rng is None and not Handler._overlap_started:
+            Handler._overlap_started = True
+            self.send_response(200)
+            self.send_header("Content-Type", "application/octet-stream")
+            self.send_header("Content-Length", str(len(blob)))
+            self.send_header("Accept-Ranges", "bytes")
+            self.end_headers()
+            if self.command != "HEAD":
+                self.wfile.write(blob[: self.OVERLAP_PREFIX_LEN])
+                self.wfile.flush()
+                try:
+                    while True:
+                        time.sleep(3600)
+                except OSError:
+                    pass
+            return
+        if rng is None:  # no resume request: serve the whole file
+            return self.route_overlap_full()
+        # Resume: honor the Range, but back up OVERLAP_EARLY bytes.
+        start = (
+            int(rng[len("bytes=") :].split("-")[0]) if rng.startswith("bytes=") else 0
+        )
+        start = max(0, start - self.OVERLAP_EARLY)
+        # Signal that the resume Range -> 206 path actually fired, so the test
+        # can prove it was exercised (not a silent full re-download).
+        resumed = os.environ.get("OVERLAP_RESUMED")
+        if resumed:
+            with open(resumed, "a") as fp:
+                fp.write("x")
+        part = blob[start:]
+        self.send_response(206, "Partial Content")
+        self.send_header("Content-Type", "application/octet-stream")
+        self.send_header("Content-Length", str(len(part)))
+        self.send_header(
+            "Content-Range", "bytes %d-%d/%d" % (start, len(blob) - 1, len(blob))
+        )
+        self.end_headers()
+        if self.command != "HEAD":
+            self.wfile.write(part)
+
+    # error pages / 0-byte files (#17): -o0 ("no error pages") must keep 4xx/5xx
+    # bodies off disk; a genuine 0-byte 200 is a valid file and stays.
+    def route_errpage_index(self):
+        self.send_html(
+            '\t<a href="good.html">good</a>\n'
+            '\t<a href="missing.html">missing</a>\n'
+            '\t<a href="empty.html">empty</a>\n'
+        )
+
+    def route_errpage_good(self):
+        self.send_raw(b"<html><body>good page</body></html>\n", "text/html")
+
+    def route_errpage_missing(self):
+        self.send_html("\t404 error body", status=404, extra_status="Not Found")
+
+    def route_errpage_empty(self):
+        self.send_raw(b"", "text/html")
+
+    # broken Content-Length (#32/#41): declared size != bytes sent. httrack
+    # warns "bogus state (broken size)" and skips the cache unless -%B.
+    def route_size_index(self):
+        self.send_html('\t<a href="oversize.bin">over</a>\n')
+
+    def route_size_oversize(self):
+        body = b"A" * 100
+        self.send_response(200)
+        self.send_header("Content-Type", "application/octet-stream")
+        self.send_header("Content-Length", str(len(body) - 2))  # lie: too short
+        self.send_header("Connection", "close")
+        self.end_headers()
+        if self.command != "HEAD":
+            self.wfile.write(body)
+
     ROUTES = {
         "/cookies/entrance.php": route_entrance,
         "/cookies/second.php": route_second,
@@ -248,6 +346,15 @@ class Handler(SimpleHTTPRequestHandler):
         "/intl/" + INTL_NAME: route_intl_page,
         "/resume/index.html": route_resume_index,
         "/resume/blob.txt": route_resume,
+        "/overlap/index.html": route_overlap_index,
+        "/overlap/flaky.bin": route_overlap,
+        "/overlap/full.bin": route_overlap_full,
+        "/size/index.html": route_size_index,
+        "/size/oversize.bin": route_size_oversize,
+        "/errpage/index.html": route_errpage_index,
+        "/errpage/good.html": route_errpage_good,
+        "/errpage/missing.html": route_errpage_missing,
+        "/errpage/empty.html": route_errpage_empty,
     }
 
     # --- dispatch ----------------------------------------------------------