Daniel,

Thank you for reviewing multi-wc-format.

I think it would be good for us to be able to refer to the issue
​tracker. I'll file a 'multi-wc-format' issue and we can add sub-issues
​for changes we need or want.

https://subversion.apache.org/issue/4883 "multi-wc-format"


Daniel Shahaf wrote:
> 1. How can a user ask a working copy what range of minor versions it
> supports?  Cf. "Compatible With Version:" in `svnadmin info` output.
> [...]

I agree with all of that, and it seems like something we should do
​before releasing the feature. Filed, quoting your comments, as:

https://subversion.apache.org/issue/4884 "multi-wc-format: user
​visibility of WC version"


> 2. The comment in trunk just above STMT_UPGRADE_TO_32 is just an
> ellipsis.  It should be fleshed out.  Would it be correct to change it
> to say "Format 32 is identical to format 31"? [...]

Yes. Done in r1898364.


> 3. «svn upgrade» of an f31 working copy does nothing and prints nothing.
> I see this is deliberate behaviour (by the docstring of 
> SVN_WC__DEFAULT_VERSION).
> 
> 3.1. Should «svn upgrade» upgrade to the newest supported format?
> That's what «svnadmin upgrade» does and what «svn upgrade» has done to
> date, so users may expect this.

Given how Subversion has matured, and how we are aware of the
​difficulties that WC version upgrades cause for users of multiple
​clients, we are now deliberately choosing compatibility as the default,
​now that we have the ability to do so.

> 3.2. Failing that, should «svn upgrade» print an info message to the
> effect of "I upgraded your wc to f31 but I can also upgrade it to f32 [...]?

Good idea. Filed as:

https://subversion.apache.org/issue/4885 "multi-wc-format: WC upgraded
​and not-upgraded notifications"

with a short reply.


> 4. Since wc-queries-test.c special-cases STMT_UPGRADE_TO_32,
> wc-metadata.sql should grow comments pointing to the places outside
> wc-metadata.sql itself that need to be changed when a creating a new
> format.

Done in r1898365.


> 5. We should probably keep a list somewhere of what set of «make check
> WC_FORMAT_VERSION=%s» values is needed in order to cover all
> currently-supported codepaths.

A good idea in principle. Ideally, we should probably keep a
​machine-readable list of all test parameters, along with which subset we
​recommend for different purposes such as pre-commit checks, backport
​proposal checks, release candidate checks.

I do not see a good place to put this right now. The closest thing we
​have already might be 'subversion/tests/README' or
​'tools/dev/unix-build' but neither of these enumerates the test parameters.


> Sorry for not getting to this before the merge.

That's no problem.

- Julian

Reply via email to