[infinispan-dev] Master not compiling

[Sanne Grinovero]

On 30 September 2011 15:50, Emmanuel Bernard <emmanuel at hibernate.org> wrote:
>
> On 30 sept. 2011, at 16:46, Mircea Markus wrote:
>
> Now there's a reason why merge was used there and not rebase. If you add a
> new file in a topic branch and modify it in several commits on the same
> topic branch, then during rebase git sees all these individual commits on
> the same files as conflicts, which makes rebase a real PTA. Optimistic
> locking impl had 52 commits, most of them touching same classes.
>
> There may be something else at play here. I'm almost certain that if you add
> a new file and subsequently change it in a topic branch and afterwards
> rebase to another branch it works without conflict.

There are ways to avoid most conflicts, it was quite amazing to
realize yesterday that after changing almost all file position in
Search *twice*  in a week, pending complex pull requests (from
contributors!) and work in progress we had in other branches could
just be rebased without any conflict: all the changes we made didn't
make it any harder, so I guess it can help to plan the order of
changes, that's also why I'm reworking the commit sequence regularly
if I'm not integrating soon.

In the meantime I found more problems :)
it's not only the last commit breaking the compilation, even before
that somebody has been cleaning up the TestCacheManagerFactory without
checking what was being used by non-core modules. This likely happened
because tests fail all the time so it's hard to run a full build and
check if any change is safe.

Sanne