[PATCH 0 of 4] largefiles: make system cache a user cache, and the repo cache a store

Matt Mackall mpm at selenic.com
Thu Oct 20 15:22:20 CDT 2011

On Thu, 2011-10-20 at 14:55 -0400, Benjamin Pollack wrote:
> On Oct 20, 2011, at 2:49 PM, Greg Ward wrote:
> > On Thu, Oct 20, 2011 at 1:24 PM, Benjamin Pollack
> > <benjamin at bitquabit.com> wrote:
> >> This patch series does several things:
> >> 
> >>   - In introduces a test authored by Eli Carter that provides a concrete
> >>     failure for the current behavior
> >>   - It corrects the documentation to clarify what the proper behavior ought to
> >>     be
> >>   - It renames a large number of functions and methods, plus a config setting,
> >>     to actually reflect that desired behavior
> >>   - It actually fixes the behavior to match the description, allowing the
> >>     failing test to pass
> >> 
> >> The last two, or even three, patches can be folded; they were kept separate for
> >> my sanity during development, and for ease of tracking what I actually changed.
> >> Likewise, if we're concerned about such a large renaming, the last patch can be
> >> easily rewritten to use the old function names.
> > 
> > Nice! Since patch #1 introduces a failing test, IMHO the commit
> > message should say so. Or, better, it should be the last patch, since
> > we generally don't like failing tests.
> That's fine with me.  Sorry for doing it backwards; our team's
> practice is to introduce the test first, then the code that allows it
> to pass, so I just did that without thinking.  I can resubmit in the
> other order, or it can just be applied differently (it goes cleanly
> either way).

I'll rearrange them.

The general principle at work here is the classic dictum "don't break
the build". To the extent that a test doesn't run, it means the "build"
is (at least conceptually) broken. While this is not likely to be an
issue in practice, we still prefer every commit to leave the project in
a 'good' state. Also, it gives me the flexibility to accept the first N
patches in a series. So we prefer a new test to either be introduced
together with the fix or after.

Mathematics is the supreme nostalgia of our time.

More information about the Mercurial-devel mailing list