Revert "fix(network): advance send checkpoint forward-only and keep first send failure"

This reverts commit 88ddeb1.

The forward-only guard was wrong. The cloudsync server computes
lastOptimisticVersion as the last *continuous* version over applied ranges plus
active (pending/running) apply jobs (internal/syncstatus/status.go,
ranges.go:LastContinuousVersion). When an apply job fails it leaves the active
set and is not in the applied ranges, so its range becomes a gap and
lastOptimisticVersion regresses to just before the failed range. That regression
is the server's re-send signal: the client must follow lastOptimisticVersion
backward and re-stream from there. The forward-only guard suppressed it, turning
a recoverable apply failure into a permanently skipped change.

Restoring the original behavior (follow lastOptimisticVersion wherever it points)
also makes review finding #5 ("checkpoint can regress") a non-issue -- the
regression is intended -- and the keep-first failure change moot, since the
server returns a single current failure per response rather than distinct
per-chunk failures.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: andinux

src/network/network.c

@@ -1616,20 +1616,16 @@ static void network_sync_state_update_from_response(NETWORK_RESULT *res,
     int parsed_gaps_size = json_extract_array_size(res->buffer, res->blen, "gaps");
     if (parsed_gaps_size >= 0) *gaps_size = parsed_gaps_size;
 
-    // Keep the FIRST failure observed across the chunk stream, not the last: an
-    // earlier chunk's failure is the more likely root cause, and a later chunk's
-    // response must not silently mask it (each chunk's response can carry a
-    // failures.{apply,check} object).
     char *apply_failure = json_extract_failure_stage(res->buffer, res->blen, "apply");
     if (apply_failure) {
-        if (*apply_failure_json) cloudsync_memory_free(apply_failure);
-        else *apply_failure_json = apply_failure;
+        if (*apply_failure_json) cloudsync_memory_free(*apply_failure_json);
+        *apply_failure_json = apply_failure;
     }
 
     char *check_failure = json_extract_failure_stage(res->buffer, res->blen, "check");
     if (check_failure) {
-        if (*check_failure_json) cloudsync_memory_free(check_failure);
-        else *check_failure_json = check_failure;
+        if (*check_failure_json) cloudsync_memory_free(*check_failure_json);
+        *check_failure_json = check_failure;
     }
 }
 
@@ -1790,19 +1786,14 @@ int cloudsync_network_send_changes_internal (sqlite3_context *context, int argc,
     }
     if (gaps_size < 0) gaps_size = 0;
 
-    // Advance the send checkpoint FORWARD ONLY. The server's lastOptimisticVersion
-    // can momentarily report a value below our current checkpoint (a stale or
-    // out-of-order status response, or a server that is behind this site); writing
-    // it back verbatim (the old "!= db_version" guard) would rewind the cursor and
-    // re-stream already-sent data, or worse, interact with a later advance to skip
-    // a range. Only move the cursor when the target is strictly ahead. When the
-    // server returns no optimistic version, fall back to the max watermark we
-    // streamed (new_db_version, always >= db_version).
+    // update db_version in settings
     char buf[256];
-    if (last_optimistic_version > db_version) {
-        snprintf(buf, sizeof(buf), "%" PRId64, last_optimistic_version);
-        dbutils_settings_set_key_value(data, CLOUDSYNC_KEY_SEND_DBVERSION, buf);
-    } else if (last_optimistic_version < 0 && new_db_version > db_version) {
+    if (last_optimistic_version >= 0) {
+        if (last_optimistic_version != db_version) {
+            snprintf(buf, sizeof(buf), "%" PRId64, last_optimistic_version);
+            dbutils_settings_set_key_value(data, CLOUDSYNC_KEY_SEND_DBVERSION, buf);
+        }
+    } else if (new_db_version != db_version) {
         snprintf(buf, sizeof(buf), "%" PRId64, new_db_version);
         dbutils_settings_set_key_value(data, CLOUDSYNC_KEY_SEND_DBVERSION, buf);
     }