Fix issue created in PR#1848

We shouldn't be attempting to create embedded reference sequences for
CRAM containers with reads mapped to chr -1 (ie unmapped).  We don't
permit embed_ref in multi-ref mode and it's nonsensical for entirely
unmapped data.

The only real fix needed here is "c->ref_id < 0" just before
calling cram_generate_reference(), but checking elsewhere can sidestep
other potential issues and we have safety in checking in more than one
place.

Credit to OSS_Fuzz
Fixes oss-fuzz issue 372547397

Author: jkbonfield

cram/cram_encode.c

@@ -1842,7 +1842,7 @@ int cram_encode_container(cram_fd *fd, cram_container *c) {
     // Don't try embed ref if we repeatedly fail
     pthread_mutex_lock(&fd->ref_lock);
     int failed_embed = (fd->no_ref_counter >= 5); // maximum 5 tries
-    if (!failed_embed && c->embed_ref == -2) {
+    if (!failed_embed && c->embed_ref == -2 && c->ref_id >= 0) {
         hts_log_warning("Retrying embed_ref=2 mode for #%d/5", fd->no_ref_counter);
         fd->no_ref = c->no_ref = 0;
         fd->embed_ref = c->embed_ref = 2;
@@ -1921,6 +1921,12 @@ int cram_encode_container(cram_fd *fd, cram_container *c) {
                 // Do not confuse with fd->ref_free which is a pointer to a
                 // reference string to free.
                 c->ref_free = 1;
+            } else {
+                // Double check for broken input.  We shouldn't have
+                // embedded references enabled for unmapped data, but our
+                // data could be broken.
+                embed_ref = 0;
+                no_ref = c->no_ref = 1;
             }
         }
         c->ref_seq_id = c->ref_id;
@@ -1967,7 +1973,7 @@ int cram_encode_container(cram_fd *fd, cram_container *c) {
 
         // Embed consensus / MD-generated ref
         if (embed_ref == 2) {
-            if (cram_generate_reference(c, s, r1) < 0) {
+            if (c->ref_id < 0 || cram_generate_reference(c, s, r1) < 0) {
                 // Should this be a permanent thing via fd->no_ref?
                 // Doing so means we cannot easily switch back again should
                 // things fix themselves later on.  This is likely not a

cram/cram_io.c

@@ -4954,6 +4954,8 @@ int cram_write_SAM_hdr(cram_fd *fd, sam_hdr_t *hdr) {
                         hts_log_warning("NOTE: the CRAM file will be bigger "
                                         "than using an external reference");
                         pthread_mutex_lock(&fd->ref_lock);
+                        // Best guess.  It may be unmapped data with broken
+                        // headers, in which case this will get ignored.
                         fd->embed_ref = 2;
                         pthread_mutex_unlock(&fd->ref_lock);
                         break;