New website and using gerrit

[Rob Spanton]

Hey Guys,

I've just switched studentrobotics.org to point at the new PR website.  It now
points at a google container engine machine that is using the config found in
the website repo on github [1].  Yay, new website! :D  This proxies a whole load
of things through to the existing Linode box.

This doesn't proxy gerrit's ssh through to the linode machine -- I've made it so
that gerrit.studentrobotics.org (and gerrit.srobo.org) point at the linode
machine.  If you wish to continue using gerrit, then you can update your ssh
config to point at this new destination.  For example, my ssh config changes to
the following:

Host sr-gerrit
HostName gerrit.studentrobotics.org
User rspanton
Port 29418

Cheers,

Rob

[1] https://github.com/srobo/website/tree/master/_env

[Peter Law]

Hi,

From a quick look at the config, it seems to proxy the IDE back to
itself (line 37 [1] doesn't have a subdomain in the proxy url). Is
that intentional?

Thanks,
Peter

[1] https://github.com/srobo/website/blob/master/_env/nginx.conf#L37

> --
> You received this message because you are subscribed to the Google Groups "Student Robotics Development" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to srobo-devel...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Peter Law]

Hi,

I've also just noticed that there's no mention of subversion in the
config. Are we abandoning our SVN repo at this point?

Thanks,
Peter

[Peter Law]

I wrote:
> From a quick look at the config, it seems to proxy the IDE back to
> itself (line 37 [1] doesn't have a subdomain in the proxy url). Is
> that intentional?

Oddly, it still seems to work though.

Thanks,
Peter

[Peter Law]

Scratch that, I assume that it works because it's just been updated.

P

[Andy Busse]

While I enjoy the new style, is this still a WIP? e.g. Docs are
difficult to navigate (some images also don't load, and interactive
troubleshooter is down), and there appear to be plenty of dead links on
the front page news article.

Nice work on the new format, though.

Andy

[Rob Spanton]

On Mon, 2016-09-05 at 23:00 +0100, Andy Busse wrote:
> While I enjoy the new style, is this still a WIP?

Indeed, as was the old site!

The docs are of course a priority for existing in a usable form by
kickstart.

Cheers,

Rob

[Jimmy Thompson]

Cheers for the comments, and the bug find!

It's definitely still a work in progress, and likely will be for a long time. :-)

The docs look a bit shit right now; they are the first in a line of things we want to try and make look more consistent, the aim is likely an identical masthead and a consistent set of colours and fonts.

As for the public website, there's plenty to do. I'm trying to put them into Github issues[1], if you find some time, PRs are always appreciated.

[1] https://github.com/srobo/website/issues 

--
You received this message because you are subscribed to the Google Groups "Student Robotics Development" group.

To unsubscribe from this group and stop receiving emails from it, send an email to srobo-devel+unsubscribe@googlegroups.com.

[Peter Law]

Hi,

Having just tried to update some of my local git repos I'm finding
that this no longer works. I'm guessing that this is because the
git:// protocol also uses a port which isn't currently forwarded.
Please could this be resolved ASAP since I don't have a central place
to update my git repos, and don't fancy going through several hundred
making a fix manually.

A (temporary [1]) fix is to change the url of the relevant remote to
explicitly point at saffron (e.g:
git://saffron.studentrobotics.org/inventory.git).

Thanks,
Peter

[1] And undesirable since it will need to be repeated whenever our git
host moves.

On 5 September 2016 at 22:16, Rob Spanton <r...@robspanton.com> wrote:

> --
> You received this message because you are subscribed to the Google Groups "Student Robotics Development" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to srobo-devel...@googlegroups.com.

[Rob Spanton]

On Fri, 2016-09-09 at 00:05 +0100, Peter Law wrote:
> Having just tried to update some of my local git repos I'm finding
> that this no longer works. I'm guessing that this is because the
> git:// protocol also uses a port which isn't currently forwarded.
> Please could this be resolved ASAP since I don't have a central place
> to update my git repos, and don't fancy going through several hundred
> making a fix manually.

You don't have to!  Just stick something along the lines of this in your global
git config:

[url "git://git.studentrobotics.org"]
    insteadOf = "git://studentrobotics.org"

[url "git://git.srobo.org"]
    insteadOf = "git://srobo.org"

Cheers,

Rob

[Peter Law]

Hi,

> On Fri, 2016-09-09 at 00:05 +0100, Peter Law wrote:
>> Having just tried to update some of my local git repos I'm finding
>> that this no longer works. I'm guessing that this is because the
>> git:// protocol also uses a port which isn't currently forwarded.
>> Please could this be resolved ASAP since I don't have a central place
>> to update my git repos, and don't fancy going through several hundred
>> making a fix manually.

On 9 September 2016 at 20:38, Rob Spanton <r...@robspanton.com> wrote:
> You don't have to! Just stick something along the lines of this in your global
> git config:
>
> [url "git://git.studentrobotics.org"]
> insteadOf = "git://studentrobotics.org"
>
> [url "git://git.srobo.org"]
> insteadOf = "git://srobo.org"

Aha, thanks.

P

[Peter Law]

Hi,

I'm not sure if it's related to this, but the gerrit web interface
doesn't work properly (for me) when using the www. site. (It does work
when visiting gerrit., but that's annoying to remember and all the
existing links point at www. and gerrit itself redirects there in some
cases).

The failure mode is that on visiting any patch page directly (for
example via emails) you get a "Session Expired" popup which instructs
you to login. After logging in visiting any patch page then fails with
"The page you requested was not found, or you do not have permission
to view this page."

Rather oddly if you start your visit at the gerrit root page then
everything seems to work ok.

I'm guessing that there's something odd going on in the way that
information is forwarded from the new nginx scenario, but I've no idea
how to go about debugging it. Are there some instructions somewhere on
how I can set up a local copy of the website's forwarding stuff for
testing this?

Thanks,
Peter

On 5 September 2016 at 22:16, Rob Spanton <r...@robspanton.com> wrote:

[Peter Law]

Hi,

I've also just noticed that http://ide.srobo.org/ doesn't redirect
properly. It should put you at https://www.studentrobotics.org/ide/,
but instead puts you at https://www.studentrobotics.org/.

Thanks,
Peter

On 5 September 2016 at 22:16, Rob Spanton <r...@robspanton.com> wrote:

[Peter Law]

Hi,

While this works beautifully for local development where each
developer can modify their config, it doesn't work as well for CIs (or
related environments), particularly where they mix with submodules. I
suggest that in the future we default to using git.-prefixed urls for
submodules, though fixing this so that it's actually supported would
be nice.

For now I've fixed the repo I hit this in and will keep an eye out for
others in this situation.

Thanks,
Peter

[Peter Law]

Hi,

> I'm not sure if it's related to this, but the gerrit web interface
> doesn't work properly (for me) when using the www. site.

*snip*

> I'm guessing that there's something odd going on in the way that
> information is forwarded from the new nginx scenario,

This turned out to be the cause; there's a PR at
https://github.com/srobo/website/pull/53 which resolves the issue.

> but I've no idea
> how to go about debugging it. Are there some instructions somewhere on
> how I can set up a local copy of the website's forwarding stuff for
> testing this?

Update: there are now some instructions in the README, though it
remains far from ideal. https://github.com/srobo/website/issues/50
exists to cover their improvement.

Thanks,
Peter

[Peter Law]

Hi,

> git://srobo.org not working any longer, with various workarounds

I recently hit another situation where this causes weird issues. While
submodules are suitably redirected using the url .. insteadOf config
trick, that clearly relies on all developers having that workaround
installed.

Unfortunately as I noted all non-developer scenarios remain broken, so
I've been migrating any submodules over to using git.srobo.org
directly. In theory that should survive any future host changes as
well as avoiding additional network latency from using a port forward
if one ever appears.

While nice in practise this has some additional complexities. In order
for the new url to used, users must run `git submodule sync` in each
repo containing submodules. Where submodules are nested this means
several layers of update commits and `git submodule sync` runs are
needed to fully convert a repo.

This is an inconvenience locally, but turns out to be a serious issue
on saffron (both the live one and any dev clones) where several such
repos are deployed. While in general on saffron we fetch from the
local version of the repo (rather than fetch over a network something
which is essentially coming from localhost), this does not happen for
submodules [1].

These two factors, combined with puppet not appearing to have support
for running `git submodule sync` mean that some of saffron's repos had
managed to get themselves into a rather weird state of partial update.
I've fixed the couple I spotted which were causing puppet to error,
though there may be others hiding simply because they haven't had any
updates pushed yet.

For anyone needing to fix this in the future the process is largely
the same as for a local repo:
- if needed: `chown` the files in the repo which puppet has mangled ownership of
- `su` to the appropriate user for the repo then:
-- `git fetch --all --recurse-submodules`
-- `git submodule sync`
-- `git submodule foreach --recursive git submodule sync`
-- `git submodule update --recursive`

You may need to repeat these steps a few times to get each layer of
submodules suitably updated.

Thanks,
Peter

[1] We could use relative urls, but there are other reasons not to do
that: mainly that other clones would then need to re-host the entire
chain and be kept in sync, greatly complicating doing so.