Inspired by a recent SoPW, how short can you write a diff of two simple csv files that displays the missing keys from the two files prefixed by the fileno they were missing from?

Some test data:

P:\test>type t1.csv
1,the,rest
2,the,rest
3,the,rest
4,the,rest
5,the,rest

P:\test>type t2.csv
2,the,rest
4,the,rest
6,the,rest
8,the,rest
[download]

And my best that at 114 strokes can surely be well beaten?

Rules: Command, including all switches count; filenames don't.

# 23456789 123456789 123456789 123456789 123456789 123456789 123456789
+ 123456789 123456789 123456789 123456789 1234
perl -anF, -le"$h{$F[0]}=1;eof&&last}{$h{(split',',$_)[0]}--while<>;$h
+{$_}&&print $h{$_}<0?'1:':'2:',$_ for keys%h" t1.csv t2.csv
1:8
1:6
2:1
2:3
2:5
[download]