snappy: fix multi-chunk framed data silently dropped

After the frame parsing loop accumulates the total decompressed
length in aggregated_data_length, the variable was incorrectly
reset to zero before being used for the output buffer allocation.
This caused the multi-chunk code path to skip allocation entirely,
producing a NULL output buffer and zero length with a success
return code — silently dropping all decompressed data.

Remove the erroneous reset so the accumulated length is preserved
and used for the aggregation buffer allocation. Also gate the
allocation on result == 0 and free any allocated buffer on error,
preventing a memory leak when a multi-chunk stream fails mid-way
(e.g. checksum mismatch after one valid chunk).

Signed-off-by: Gabriele Facundo <gabriele.facundo@cern.ch>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: thewillyhuman

src/flb_snappy.c

@@ -302,7 +302,6 @@ int flb_snappy_uncompress_framed_data(char *in_data, size_t in_len,
     }
 
     aggregated_data_buffer = NULL;
-    aggregated_data_length = 0;
 
     if (compressed_chunk_count == 1 &&
         uncompressed_chunk_count == 0 &&
@@ -322,7 +321,7 @@ int flb_snappy_uncompress_framed_data(char *in_data, size_t in_len,
         flb_free(chunk);
     }
     else {
-        if (aggregated_data_length > 0) {
+        if (aggregated_data_length > 0 && result == 0) {
             aggregated_data_buffer = flb_calloc(aggregated_data_length,
                                                 sizeof(char));
 
@@ -357,6 +356,12 @@ int flb_snappy_uncompress_framed_data(char *in_data, size_t in_len,
         }
     }
 
+    if (result != 0) {
+        flb_free(aggregated_data_buffer);
+        aggregated_data_buffer = NULL;
+        aggregated_data_offset = 0;
+    }
+
     *out_data = (char *) aggregated_data_buffer;
     *out_len = aggregated_data_offset;