Re: [gpsd-dev] Upcoming migration to GitLab

[Fred Wright]

On Mon, 27 May 2019, Eric S. Raymond wrote:

> Fred Wright <address@hidden>:
> > Did you do something to make Fork and Clone appear?
>
> Enrolling you as a dev probly did it.

Perhaps.  I thought I'd still observed the trouble after that, but maybe I 
remembered wrong, or maybe there was some "percolation time" involved.

In any case, that's clearly not an acceptable state of affairs for the 
final result, since Fork and Clone shouldn't be limited to project 
members.  Obviously Fork is limited to GitLab users, since it needs some 
place to put the fork, but it could be any GitLab user.  And of course 
Clone (which is just copying a URL to the clipboard, anyway) should be 
available to anyone.

> >         git fetch source_repo source_branch:local_branch
> >         git push dest_repo local_branch:dest_branch
>
> Right, that's a variation on the push method I had in mind.  I should have 
> realized
> one can stage through a third repo.

And it's not even very expensive when all three repos are in similar 
states, since both fetch and push are incremental.

On Mon, 27 May 2019, Eric S. Raymond wrote:

> Hal Murray <address@hidden>:
> > > I just enrolleed you as a developer. See if you can clone now?
> >
> > Looks good.
> >
> > > I'm a bit puzzled.  The project visibility is definirely "public".
> >
> > You might want to leave me as a non-developer so I can help debug things like
> > this.
>
> Let's run that expriment after I turn off the mirroring.  That seems to
> make a big difference inn the access rules.

Maybe.  So far it's only conjecture that the problem relates to mirroring.

> > I think we (both gpsd and ntpsec) need some simple HOWTOs to cover 2 cases.
> >
> > The first is just a simple download, build, and test and/or install.  Download
> > probably includes both git to get the latest and tarballs for those that don't
> > like git and/or want to use official releases.
> >
> > The second is how to contribute.  I don't understand this area.  I can
> > clone/pull/push, but don't know what I'm missing.
>
> If you know how to make MRs from a clone, you're set.  Do you?

The first step is to clone from a personal fork of the main repo, not the 
main repo itself.  Having done that, the general idea is to make changes 
in a local branch, push the branch to the fork, and then initiate the MR 
from the fork's webpage.

Personally, with any project hosted on a site with forking like GitHub or 
GitLab, I always make a fork, clone from the fork, and then add the main 
repo as 'upstream'.  Even if the only immediate goal is to examine and/or 
build the code, having the right setup for a later PR or MR is a plus.

On Mon, 27 May 2019, Eric S. Raymond wrote:
> Chris Kuethe <address@hidden>:
> > It's easy to check the project visibility status - just pop open an
> > incognito window.
> >
> > I think the project is set to public, so its viewable, but the source repo
> > is not. But I don't have access to the project setting so I can't check
> > that myself. https://docs.gitlab.com/ee/user/project/settings/
>
> I can't find a separate visibility setting for the repository.

I looked at the GitLab documentation somewhat, and didn't see anything 
promising.

On Mon, 27 May 2019, Eric S. Raymond wrote:
> Greg Troxel <address@hidden>:
>
> > I am also seeing strangeness.  Going to
> >
> >   https://gitlab.com/gpsd/gpsd
> >
> > when not logged in, I see no repository.   I logged in, and then there
> > is the repository.  Sign out, and the same URL is 404.

Odd.  I don't get a 404 when not logged in; I just get the version 
without the Fork and Clone buttons.

> > I have no useful advice about what's going on or what to do; just a data
> > point.
>
> This is consistent with the repo not actually being public even though
> Public is set.  I have to wonder if we might not have stumbled over
> a bug in the GitLab software,,,

I just noticed something else which might be a clue:  When not logged in, 
https://gitlab.com/gpsd/gpsd doesn't show the usual "repo view" at all, 
but instead shows a page listing the project members (though with the same 
initial boilerplate as the repo view).  It of course makes perfect sense 
that the member list isn't forkable or clonable. :-)

I wonder if it confuses GitLab to have a project name, a repo name, and a 
group name that are all identical.  I notice that with NTPsec, there's a 
redirection:

https://gitlab.com/NTPsec/NTPsec -> https://gitlab.com/NTPsec/ntpsec

This happens with or without logging in, with the Clone button being 
present in both cases, but the Fork button only being present when logged 
in.  And the redirected URL doesn't have the same name at both levels, if 
one assumes case sensitivity.

Fred Wright