Fix a number of ASN1_INTEGER vs ASN1_STRING mixups coming from the

mechanical M_ASN1 macro expansion.  The ASN1_INTEGER_cmp function
takes signs into account while ASN1_STRING_cmp doesn't. The mixups
mostly involve serialNumbers, which, in principle, should be positive.
However, it is unclear whether that is checked or enforced anywhere
in the code, so these are probably bugs.

Patch from Holger Mikolon
ok jsing

1 files changed, 2 insertions, 2 deletions

diff --git a/src/lib/libcrypto/x509/x509_cmp.c b/src/lib/libcrypto/x509/x509_cmp.c
index 2141f871e5..6d6e840899 100644
--- a/src/lib/libcrypto/x509/x509_cmp.c
+++ b/src/lib/libcrypto/x509/x509_cmp.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: x509_cmp.c,v 1.34 2018/08/24 19:59:32 tb Exp $ */
+/* $OpenBSD: x509_cmp.c,v 1.35 2019/03/13 20:34:00 tb Exp $ */
 /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
  * All rights reserved.
  *
@@ -76,7 +76,7 @@ X509_issuer_and_serial_cmp(const X509 *a, const X509 *b)
 
         ai = a->cert_info;
         bi = b->cert_info;
-        i = ASN1_STRING_cmp(ai->serialNumber, bi->serialNumber);
+        i = ASN1_INTEGER_cmp(ai->serialNumber, bi->serialNumber);
         if (i)
                 return (i);
         return (X509_NAME_cmp(ai->issuer, bi->issuer));