DBD::Oracle 1.26 new_tablename / synonyms issue

[Jan Pazdziora]

Hello,

The DBD::Oracle 1.26 code

        http://cpansearch.perl.org/src/TIMB/DBD-Oracle-1.26/oci8.c

declares

        char new_tablename[100];

in the if branch in init_lob_refetch, and then assigns

        tablename=new_tablename;

and then uses the tablename outside of that if block. So the
tablename pointer points to a random address on the stack.

It causes problems described at

        http://rt.cpan.org/Public/Bug/Display.html?id=61511
        https://bugzilla.redhat.com/show_bug.cgi?id=548489

Note that it's kinda related to

        http://rt.cpan.org/Public/Bug/Display.html?id=36069#txn-464638

where it's said

        Used some of the code form the patch there was no need to
        create a char [100] outside the if statement though.

I'd claim that it is actually necessary to declare it outside of
that block because even if the name goes out of scope at the end
of the block, the tablename pointer still points to it.

Please apply

--- oci8.c.orig 2010-11-16 16:26:42.459208675 +0100
+++ oci8.c      2010-11-16 16:27:27.194207404 +0100
@@ -4030,6 +4030,7 @@
        ub2 numcols = 0;
        imp_fbh_t *fbh;
        int unmatched_params;
+       char new_tablename[100];
        I32 i,j;
        char *p;
        lob_refetch_t *lr = NULL;
@@ -4067,7 +4068,6 @@
        if (status == OCI_SUCCESS) { /* There is a synonym, get the schema */
                char *syn_schema=NULL;
                char syn_name[100];
-               char new_tablename[100];
                ub4  tn_len = 0, syn_schema_len = 0;
 
                strncpy(syn_name,tablename,strlen(tablename));

to DBD::Oracle.

Thank you,

-- 
Jan Pazdziora