STATUS
------
Patch approved pending resolution of requests.
REQUESTS
--------
1. Correct spelling errors.
2. When I applied the first patch to my tree, the patch had already been
applied. Is it true that the first patch was pushed to the
mysql-6.0-backup tree? If not, then why doesn't the patch apply
correctly?
COMMENTARY
----------
Good explanations in tests. Well done.
SUGGESTIONS
-----------
None.
DETAILS
-------
> Bug#32465 - Backup: Commit blocker does not block tables in default driver
>
> Patch (2/2): The patch contains new validity point tests written for the
> temporary commit blocker solution of setting Global Read Lock before backup prepare phase
> (fixed in bug#39602). The backup_vp* tests replace the old backup_commit_blocker tests.
...
> === added file 'mysql-test/suite/backup/t/backup_vp_nontx.test'
> --- a/mysql-test/suite/backup/t/backup_vp_nontx.test 1970-01-01 00:00:00 +0000
> +++ b/mysql-test/suite/backup/t/backup_vp_nontx.test 2008-11-17 09:17:15 +0000
> @@ -0,0 +1,253 @@
> +#
> +# This test is one of two validity point tests. See:
> +# backup_vp_tx.test
> +#
> +# The goal of the test should be to ensure the following assumptions
> +# for the behaviour of validy point (VP) hold. Validity point is
> +# implemented using commit blocker (CB).
[1] Spelling errors. Behaviour is ok (King's English as it were), but
validy needs correcting.
...
> +# Backup will be blocked from setting CB by the ongoing insert in
> +# con_ntx1. Backup will signal con_ntx1 that it has been blocked. The
> +# insert will then finish, making backup able to set CB. When CB has
> +# been set, ntx2 will be singaled to try to insert. ntx2 insert will
> +# be blocked by the CB and send a signal that it has been blocked. CB
> +# is then released, but backup is stopped immediately after releasing
> +# CB. ntx2 will now be able to complete the commit, and ntx1 will
> +# issue and complete another insert. Finally, backup is allowed to
> +# complete.
[1] Spelling errors. 'singaled' needs correcting.
...
> +# The goal of the test should be to ensure the following assumptions
> +# for the behaviour of validy point (VP) hold. Validity point is
> +# implemented using commit blocker (CB).
[1] Spelling errors. Behaviour is ok (King's English as it were), but
validy needs correcting.
...
> +# Backup will be blocked from setting CB by the ongoing commit in
> +# con_tx1. Backup will signal con_tx1 that it has been blocked. The
> +# commit will then finish, making backup able to set CB. When CB has
> +# been set, tx2 will be singaled to try to commit. tx2 commit will be
> +# blocked by the CB and send a signal that it has been blocked. CB is
> +# then released, but backup is stopped immediately after releasing CB.
> +# tx2 will now be able to complete the commit, and tx3 will issue and
> +# complete a commit. Finally, backup is allowed to complete.
[1] Spelling errors. 'singaled' needs correcting.
...
Chuck
