Re: Remove lily-git?

[Jean Abou Samra]

Hi,
Le 04/06/2020 à 08:31, James Lowe a écrit :
> On 03/06/2020 21:25, Karlin High wrote:
> > On 6/3/2020 3:14 PM, Jean Abou Samra wrote:
> > > There is a discussion at 
> > > https://gitlab.com/lilypond/lilypond/-/issues/1012
> > > about the future of lily-git.Basically, I think that it no longer 
> > > makes sense to keep it now that we switched to GitLab.
> >
> > I remember seeing this thing bring in over 500MB of dependencies on a 
> > Debian Linux system. And I was thinking, "If that's the only piece of 
> > TCL in the whole LilyPond ecosystem, there has GOT to be a way to 
> > avoid having this."
>
> I am not sure that is correct, lily-git is just a set of python 
> commands with a Front End GUI (for the likes of me) that made sure 
> that you had set your git repo correctly and could easily download 
> $LILYPOND_GIT. It also forced you to set your git user and email.
>
> Lily-git in and of itself was tiny and needed hardly anything to run 
> (wish lily-git.tcl).
>
> The 500MB of dependencies was, I expect, for dblatex et al. That we 
> needed for doc building at the time but lily-git only cloned the repo 
> (and allowed a button to hard reset - again for idiots like me).
>
> I am a bit older and wiser now, but even so git is still a terrible 
> 'ecosystem' not made much better by the gitlabs and githubs of the 
> world (I have the joy of having to interface with both as a 
> non-developer). That said, yes we don't 'need' lily-git, however I'd 
> like to give a hat-tip to the few devs that kept it going so I could 
> do my work. If it weren't for lily-git (and at the start 'lily-dev' - 
> still less faff than containers and jails BTW for non-devs) I'd have 
> not been able to easily contribute to this project and may have simply 
> given up having to learn the terrible interface that is git cli with 
> all the breakages of master we had at the start of when I joined.
>
> James
I do agree with you that Git can be a bit of a trick to learn (at least,
there is a long path before you fully master it). What if right now we
just added a link to some Git graphical client like 
https://www.syntevo.com/smartgit/?
It doesn't remove the complexity of Git (obviously that's quite more
involved than lily-git, the target not being the same), but at the very
least, you don't need to bother with a command-line interface.

lily-git is not going to be usable in an immediate future.
Anyway, if I had to write something, I would do it in Python as per
the abovementioned issue (now that recent versions of Python ship
with tkinter). And as we're still trying to figure out how exactly
we are going to work with GitLab, starting a new tool right now
doesn't look like a good idea. So, what I would propose at this point is
to drop lily-git.tcl, as it doesn't provide a value neither for a user
these days, nor for the people that could develop a new tool in Python
in the future; then, depending on how things go on, and if we feel the
need for it, write lily-git.py or whatever. Maybe we could open an issue
to track that. Makes sense?

Best,
Jean Abou Samra