Bug#733650: readseq: Buffer overflow in ureadseq.c

[Michael Bienia]

Package: readseq
Version: 1-10
Severity: normal
Tags: patch

Hello,

looking at why readseq failed to build in Ubuntu trusty, gcc found a
buffer overflow in ureadseq.c:

In function 'strcpy',
    inlined from 'writeSeq' at ureadseq.c:1925:13:
/usr/include/x86_64-linux-gnu/bits/string3.h:104:3: warning: call to 
__builtin___memcpy_chk will always overflow destination buffer [enabled by 
default]

which then got triggered by the test suite:

Test of NCBI ASN.1 conversions:
./readseq -p -f=asn test.m-ig  -otest.m-asn
*** buffer overflow detected ***: ./readseq terminated

Looking at ureadseq.c:writeSeq() it's pretty obvious that
  char  idword[31], endstr[10];  (line 1771)
and
  strcpy(endstr,"\"\n      } } ,"); (line 1925)
won't match (the string copied is 14 bytes long).

Regards,
Michael

Index: readseq-1/ureadseq.c
===================================================================
--- readseq-1.orig/ureadseq.c	2013-12-30 18:06:59.000000000 +0100
+++ readseq-1/ureadseq.c	2013-12-30 18:13:37.000000000 +0100
@@ -1768,7 +1768,7 @@
 
   short linesout = 0, seqtype = kNucleic;
   long  i, j, l, l1, ibase;
-  char  idword[31], endstr[10];
+  char  idword[31], endstr[14];
   char  seqnamestore[128], *seqname = seqnamestore;
   char  s[kMaxseqwidth], *cp;
   char  nameform[10], numform[10], nocountsymbols[10];