Closed dgkf closed 5 months ago
Since GitLab can have arbitrarily nested projects and some alternative syntax would be needed to differentiate user/repo/sub/dir
from /group/subgroup/subgroup2/subgroup3
, would an alternative syntax more akin to Authors@R
be up for consideration?
Remotes:
list(name = "pkg", type = "gitlab", url = "https://acme.co/a/b/c.git", subdir = "x/y/z", ref = "dev")
I appreciate that the current string style is quite brief, but maybe something more explicit would be a nice fallback?
Thanks! I changed this a bit to add auth to all git HTTP queries. There is two things to do still:
would an alternative syntax more akin to Authors@R be up for consideration?
That is not great, because other tools will interpret the Remotes
field differently. It also does not play well with other, non-R tools that try to use the DESCRIPTION
file.
This throws a wrench in the subdir parsing. To allow a subdir, I added a /-/ separator:
group/subgroup/project/-/sub/dir@ref
Yes, this could work.
Actually, some the git and gitlab tests are already "live", so we can update those later, and I am going to merge this now, after updating the docs.
Thanks you again!
This PR is intended to fix #328 with a few other related fixes
/
between the host and "username". Unlike GitHub, GitLab allows arbitrary nesting of groups:username/repo
group/subgroup/subsubgroup/project
This throws a wrench in the subdir parsing. To allow a subdir, I added a
/-/
separator:group/subgroup/project/-/sub/dir@ref
/-/
is a reserved name used for other server paths, so there's no risk of conflict with project paths./info/refs?service=git-upload-pack
require authentication, which is currently ignored. I added this authentication toasync_git_list_refs_v2
andasync_git_send_message_v2
, necessary for solving package dependencies.Let me know if I'm on the right track, or if you think this is better solved some other way.