Inconsistency WRT ghosts

John Arbash Meinel john at arbash-meinel.com
Wed Aug 6 23:45:14 BST 2008


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I'm trying to test my lazy revno numbering system against merge_sort,
and I get a weird result when it comes to Ghosts.

I believe the problem is that 'merge_sort' is defined in the absence of
NULL_REVISION, and doesn't handle when a node is referred to, but not in
the graph.

So if you want to include ghosts, you have to set their parents to (),
and then you also set other revisions from (NULL_REVISION,) => ().

One specific node which is a problem is:
$ bzr log --short -r \
~  revid:john at arbash-meinel.com-20050726150912-9d6366fa2dbb2a14
0.8.1 John Arbash Meinel        2005-07-26
~      Applied Aaron's 1-line patch.

The specific issue is that

$ bzr log --short -r 0.8.1

Returns the whole ancestry (which is a different bug, that when a revno
doesn't match anything it returns None, which causes us to assume
None..None, aka the whole list)

I'm trying to debug some other things at the same time, but I at least
wanted to make public *this* case.

(Note also, that if 0.8.1 is being pruned, then likely 0.9.1 will get
mis-numbered.)

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkiaKXoACgkQJdeBCYSNAAMLcACfbRL6X9PCNkYyKbq6bqLahwAO
d1AAniZdCu2MyVr1PZxvMJR9eIQaPWDl
=2Ooj
-----END PGP SIGNATURE-----



More information about the bazaar mailing list