cloud storage

Dirk Hohndel dirk at hohndel.org
Sat Jun 13 21:33:19 PDT 2015


OK, we're making progress...

On Fri, Jun 12, 2015 at 02:09:08PM -0700, Dirk Hohndel wrote:
> I just pushed a bunch of changes to the cloud storage.
> 
> This should be ready for testing now.

Yep, both still true. We're at .726 right now but check if there's
anything newer :-)

> The email / password / pin logic is unchanged to what I explained a few
> days ago. Except that now not saving the password locally actually works
> as expected :-)
> 
> There is a new checkbox for automated sync in the background. This may
> need a different name - all it means that if you save, it tries to then
> push the changes to the remote.
> 
> I also implemented the teapot protocoll. libgit takes forever to realize
> that it can't connect to the remote. We now figure that out in a couple of
> seconds BEFORE we try to connect with libgit2.
> 
> And if you are using a bunch of patches to libgit2 that haven't landed in
> libgit2 master, yet, then all this even works across an http proxy. And in
> my limited testing having the code there doesn't break anything with the
> libgit version that is used by the build script. It just annoyingly fails
> behind a proxy (and here the teapot protocol fails to help because that
> does use the proxy).

Also still all true.

What's new is that things work MUCH more seamlessly now.

- configure email / password
- enter the PIN
- check background sync (since we don't have manual sync, yet)

- open your regular dive file
- save to cloud storage

This should take care of creating both local and remote repositories and
branches and should set everything up just right.

I tried this with yet another email address completely from scratch
without any interaction besides what I listed here. You should never be
aware that this is a git based remote storage protocol.

Now if you quit Subsurface, run it again and load from the Cloud storage
you should get your data back. You can make changes, save them, it all
should work

As always, please please please test.

> Linus, if you have a moment I'd love for you to review the changes I made
> to save-git.c and git-access.c - I think they all make sense but that
> doesn't mean that I'm right.

Umm, yeah, I know you're in Hawaii, testing Subsurface with real data, but
if you have a moment to look through all the patches... I'm reasonably
sure that some of them might kill kittens or cause new religeous cults to
suddenly appear...

But hey, it seems to work and that's a step in the right direction :-)

For the couple of people who tried with yesterday's version... It's
possible that the new code will recover and fix things (I tried to make it
figure out if the remote branch is missing), but especially if you
manually pushed things into the remote from the command line it's possible
that Subsurface can't recover. If you see strange issues, try with a
different email address (and let me know so I can delete the entries /
repositories for the hosed accounts so you can restart from scratch for
those.

Binaries for Windows and Mac are up on the server.

/D

PS: thanks to glance the dependency on QtLocation is gone again. And I
added a crude hack to work around the system_default_directory issue on
Mac with Qt-5.4.


More information about the subsurface mailing list