lines modified on deletion

[Neil Conway]

If a file is deleted that contains keyword substitutions, CVS will
include a non-zero "lines modified" count for the deleted revision. Is
there a good reason for this behavior? (From my perspective, it seems
needlessly inconsistent.)

Specifically:

% cvs --version
Concurrent Versions System (CVS) 1.12.9 (client/server)
[...]
% touch a.c
% cvs add a.c
cvs add: use `cvs commit' to add this file permanently
% cvs ci -m "File added." a.c 
/var/lib/cvs/cvs_test/a.c,v  <--  a.c
initial revision: 1.1
% rm a.c 
% cvs delete a.c 
cvs remove: use `cvs commit' to remove this file permanently
% cvs ci -m "File removed." a.c
/var/lib/cvs/cvs_test/a.c,v  <--  a.c
new revision: delete; previous revision: 1.1
% cvs log -r1.2 a.c
[...]
revision 1.2
date: 2005-02-15 01:17:56 +0000;  author: neilc;  state: dead;  lines:
+0 -0
File removed.

However, if the file contains a keyword substitution, the "lines:"
figure for the deleted revision is non-zero:

% cat > b.c
$Id$

...
% cvs add b.c
cvs add: use `cvs commit' to add this file permanently
[neilc:/home/neilc/cvs_test]% cvs ci -m "File added." b.c 
/var/lib/cvs/cvs_test/b.c,v  <--  b.c
initial revision: 1.1
% rm b.c
% cvs delete b.c
cvs remove: use `cvs commit' to remove this file permanently
% cvs ci -m "File removed." b.c
/var/lib/cvs/cvs_test/b.c,v  <--  b.c
new revision: delete; previous revision: 1.1
[neilc:/home/neilc/cvs_test]% cvs log -r1.2 b.c
[...]
revision 1.2
date: 2005-02-15 01:20:02 +0000;  author: neilc;  state: dead;  lines:
+1 -1
File removed.

The context here is that I'm examining CVS metadata to get a sense of
the development activity of a project. Varying the "lines:" figure of a
removed file based on whether any keyword substitution took place seems
arbitrary, and introduces some error into my analysis. Of course, it's
not a huge deal (I can work around this by ignoring "lines modified"
figures in dead revisions), but IMHO it would make more sense to change
this behavior. You could:

- include the number of lines in the file's former content as being
"removed" in the deleted revision

- or, make the lines modified "0" for a deleted revision, whether
keyword substitution took place or not

I'd have no quarrel with either behavior; I guess I'd prefer the former,
actually, but I'm not militant about it.

-Neil