[Bug-tar] [PATCH] Deterministic archive type detection

[Pavel Raiskup]

Heuristic for guessing the archive/compression format works with
archive magic and tar_checksum() which depends on the first block
(BLOCKSIZE) of archive data in pre-fetched buffer.

Very small files though (compressed size is smaller than
BLOCKSIZE) don't initialize a whole data block in buffer, so the
heuristic calculated with uninitialized data previously.  In rare
situations compressed archives were marked as non-compressed,
which in turn caused extraction failure.

* src/buffer.c (check_compressed_archive): Don't assume archives
smaller than BLOCKSIZE could be non-compressed, as tar header has
always has at least one block.
---
 src/buffer.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/src/buffer.c b/src/buffer.c
index fb34834..f064d2a 100644
--- a/src/buffer.c
+++ b/src/buffer.c
@@ -391,10 +391,12 @@ check_compressed_archive (bool *pshort)
   /* Restore global values */
   read_full_records = sfr;
 
-  if ((strcmp (record_start->header.magic, TMAGIC) == 0 ||
-       strcmp (record_start->buffer + offsetof (struct posix_header, magic),
-              OLDGNU_MAGIC) == 0) &&
-      tar_checksum (record_start, true) == HEADER_SUCCESS)
+  if (record_start != record_end /* no files smaller than BLOCKSIZE */
+      && (strcmp (record_start->header.magic, TMAGIC) == 0
+          || strcmp (record_start->buffer + offsetof (struct posix_header,
+                                                      magic),
+                     OLDGNU_MAGIC) == 0)
+      && tar_checksum (record_start, true) == HEADER_SUCCESS)
     /* Probably a valid header */
     return ct_tar;
 
-- 
2.9.3