You make quite inconsistent thing in your C subroutine, at the very first step. You either replace SV *a with a new SV or do not, based on its constant-ness. Its a hole for scalar leaks.

I make a copy of the SV*a if it is readonly because otherwise I will get Modification of a read-only value attempted at c:\test\test.pl line 24. when I attempt to append to it. Otherwise, there is no need to copy rw scalars.

What will happen if you do not do "REFCNT_inc" exactly? Will the subroutine work for non-constant scalars?

If I do not increment the refcount, I get <c>

c:\test>test
billfred
fred
billfred
Attempt to free unreferenced scalar: SV 0x182445c, Perl interpreter: 0
+x2240d4 at c:\test\test.pl line 32.
1fred
Attempt to free unreferenced scalar: SV 0x182445c, Perl interpreter: 0
+x2240d4 at c:\test\test.pl line 35.
1
Attempt to free unreferenced scalar: SV 0x182445c, Perl interpreter: 0
+x2240d4.
Attempt to free unreferenced scalar: SV 0x182445c, Perl interpreter: 0
+x2240d4.
[download]

Why not write following way: ...

Using your version, which is close to where I started, I get the following results:

c:\test>test
billfred
Use of uninitialized value in subroutine entry at c:\test\test.pl line
+ 35.
fred
billfred
1fred
11
[download]

Note:

• The warning if SV*a was undefined.
• How the modification does not change the source.

  Ie. With my version, '1' becomes '1fred' becomes '1fred1'.

  With yours, '1' becomes '1fred', then becomes '11'.