Merge lp://qastaging/~spiv/bzr/tags-commit-propagation-603395-2.2 into lp://qastaging/bzr/2.2

I think having the news entry saying it propagates "new" tags is misleading. It propagates all tags. (Imagine the master deleted a tag, we'll put it back at this point.)

Anyway the change seems fine, though it seems like it would supersede the other. (rather than have it done at merge time, have it done at commit time seems a better tradeoff. Would be even better if we staged tag changes into the working tree, rather than auto-committing to the branch.)

So, I like this patch more than the other one, just needs a tweak on the NEWS entry.

If I merge this branch with trunk, both new tests fail due to script test output now having to be specified:

Traceback (most recent call last):
  ...
  File "C:\bzr\bzr\tags-commit-propagation-603395-2.2\bzrlib\tests\blackbox\test_tags.py", line 140, in test_commit_in_heavyweight_checkout_copies_tags_to_master
    """)
  File "C:\bzr\bzr\tags-commit-propagation-603395-2.2\bzrlib\tests\script.py", line 514, in run_script
    return ScriptRunner().run_script(test_case, script_string)
  File "C:\bzr\bzr\tags-commit-propagation-603395-2.2\bzrlib\tests\script.py", line 209, in run_script
    self.run_command(test_case, cmd, input, output, error)
  File "C:\bzr\bzr\tags-commit-propagation-603395-2.2\bzrlib\tests\script.py", line 228, in run_command
    raise AssertionError(str(e) + " in stdout of command %s" % cmd)
AssertionError: texts not equal:
+ Created a standalone tree (format: 2a)
 in stdout of command ['bzr', 'init', 'master']

Don't really like the duplication of the setup steps either, it's hard enough to read already without this block twice:

+ $ bzr init master
+ $ cd master
+ $ bzr commit -m "Initial commit." --unchanged
+ $ cd ..
+ $ bzr checkout master child
+ $ bzr branch master fork
+ $ cd fork
+ $ bzr commit -m "Commit in fork." --unchanged
+ $ bzr tag new-tag

mgz: Heh, I agree about the duplication... so much so that I'd already done something about it, but I'd forgotten to push! Have pushed now. Please take another look and let me know what you think. Some assertions have been moved out of run_script too, to innoculate against future changes in the command output that are irrelevant to the purpose of the tests.

I'm happy to take care of tweaking the run_script invocation when merging up to trunk — with the much shorter run_script segments I have now that should be a pretty small change.

Yup, I like the look of the assertions spelt like that.

sent to pqm by email