Read embedded thumbnail simon987/sist2#74

CMakeLists.txt

@@ -53,7 +53,6 @@ find_library(GOMP_LIB NAMES libgomp.a gomp PATHS /usr/lib/gcc/x86_64-linux-gnu/5
 target_compile_options(
         scan
         PRIVATE
-        -Werror
         -g
 )
 

libscan/ooxml/ooxml.c

@@ -145,6 +145,23 @@ static int read_doc_props(scan_ooxml_ctx_t *ctx, struct archive *a, text_buffer_
     return 0;
 }
 
+#define MAX_TN_SIZE 1024 * 1024 * 15
+
+void read_thumbnail(scan_ooxml_ctx_t *ctx, document_t *doc, struct archive *a, struct archive_entry *entry) {
+    size_t entry_size = archive_entry_size(entry);
+
+    if (entry_size <= 0 || entry_size > MAX_TN_SIZE) {
+        return;
+    }
+
+    char* buf = malloc(entry_size);
+    archive_read_data(a, buf, entry_size);
+
+    APPEND_TN_META(doc, 1, 1) // Size unknown
+    ctx->store((char *) doc->uuid, sizeof(doc->uuid), buf, entry_size);
+    free(buf);
+}
+
 void parse_ooxml(scan_ooxml_ctx_t *ctx, vfile_t *f, document_t *doc) {
 
     size_t buf_len;
@@ -180,6 +197,8 @@ void parse_ooxml(scan_ooxml_ctx_t *ctx, vfile_t *f, document_t *doc) {
                 if (read_doc_props(ctx, a, &tex, doc) != 0) {
                     break;
                 }
+            } else if (strcmp(path, "docProps/thumbnail.jpeg") == 0) {
+                read_thumbnail(ctx, doc, a, entry);
             }
         }
     }

libscan/ooxml/ooxml.h

@@ -8,6 +8,7 @@ typedef struct {
     long content_size;
     log_callback_t log;
     logf_callback_t logf;
+    store_callback_t store;
 } scan_ooxml_ctx_t;
 
 void parse_ooxml(scan_ooxml_ctx_t *ctx, vfile_t *f, document_t *doc);

test/main.cpp

@@ -388,6 +388,21 @@ TEST(Ooxml, Docx1) {
     cleanup(&doc, &f);
 }
 
+TEST(Ooxml, Docx2Thumbnail) {
+    vfile_t f;
+    document_t doc;
+    load_doc_file("libscan-test-files/test_files/ooxml/embed_tn.docx", &f, &doc);
+
+    size_t size_before = store_size;
+
+    parse_ooxml(&ooxml_500_ctx, &f, &doc);
+
+    ASSERT_NEAR(strlen(get_meta(&doc, MetaContent)->str_val), 500, 4);
+    ASSERT_NE(size_before, store_size);
+
+    cleanup(&doc, &f);
+}
+
 TEST(Ooxml, Xlsx1) {
     vfile_t f;
     document_t doc;
@@ -550,13 +565,13 @@ int main(int argc, char **argv) {
 
     arc_recurse_media_ctx.log = noop_log;
     arc_recurse_media_ctx.logf = noop_logf;
-    arc_recurse_media_ctx.store = noop_store;
+    arc_recurse_media_ctx.store = counter_store;
     arc_recurse_media_ctx.mode = ARC_MODE_RECURSE;
     arc_recurse_media_ctx.parse = _parse_media;
 
     arc_list_ctx.log = noop_log;
     arc_list_ctx.logf = noop_logf;
-    arc_list_ctx.store = noop_store;
+    arc_list_ctx.store = counter_store;
     arc_list_ctx.mode = ARC_MODE_LIST;
 
     text_500_ctx.content_size = 500;
@@ -564,7 +579,7 @@ int main(int argc, char **argv) {
     text_500_ctx.logf = noop_logf;
 
     ebook_ctx.content_size = 999999999999;
-    ebook_ctx.store = noop_store;
+    ebook_ctx.store = counter_store;
     ebook_ctx.tesseract_lang = "eng";
     ebook_ctx.tesseract_path = "./tessdata";
     ebook_ctx.tn_size = 500;
@@ -576,7 +591,7 @@ int main(int argc, char **argv) {
 
     media_ctx.log = noop_log;
     media_ctx.logf = noop_logf;
-    media_ctx.store = noop_store;
+    media_ctx.store = counter_store;
     media_ctx.tn_size = 500;
     media_ctx.tn_qscale = 1.0;
     media_ctx.max_media_buffer = (long)2000 * 1024 * 1024;
@@ -584,6 +599,7 @@ int main(int argc, char **argv) {
     ooxml_500_ctx.content_size = 500;
     ooxml_500_ctx.log = noop_log;
     ooxml_500_ctx.logf = noop_logf;
+    ooxml_500_ctx.store = counter_store;
 
     mobi_500_ctx.content_size = 500;
     mobi_500_ctx.log = noop_log;
@@ -591,7 +607,7 @@ int main(int argc, char **argv) {
 
     raw_ctx.log = noop_log;
     raw_ctx.logf = noop_logf;
-    raw_ctx.store = noop_store;
+    raw_ctx.store = counter_store;
     raw_ctx.tn_size = 500;
     raw_ctx.tn_qscale = 5.0;
 

test/test_util.h

@@ -17,8 +17,10 @@ static void noop_log(const char *filepath, int level, char *str) {
     // noop
 }
 
-static void noop_store(char* key, size_t key_len, char *value, size_t value_len) {
-    // noop
+static size_t store_size = 0;
+
+static void counter_store(char* key, size_t key_len, char *value, size_t value_len) {
+    store_size += value_len;
 }
 
 meta_line_t *get_meta(document_t *doc, metakey key);