When `$PAGER` is set in the environment, jj uses that instead of the
default (`:builtin` on Windows, `less -FRX` everywhere else). Commonly,
users will have `PAGER=less` in their environment for various reasons,
and this is respected by jj. This means that every jj command, even one
that only outputs one or two lines, will still invoke a full screen
pager. It also means that every jj command which uses escape sequences
for color, which is most of them, will be output through a pager that
doesn't handle that well, so users see output that looks like this,
which isn't very readable:
```
ESC[1mESC[38;5;2m@ESC[0m ESC[1mESC[38;5;13mspmESC[38;5;8mwzlkq...
```
To fix this issue that new users stumble upon, ignore `$PAGER` from the
environment, and always use our per-platform default. Users can set
`ui.pager` to select whichever pager they prefer.
This seems like the least bad option for resolving #3502. The cases I
considered were:
1. User doesn't have `PAGER` set. No change.
2. User has `PAGER=less` in their environment. We'll still run `less`,
just with `-FRX`, so this seems fine. This case is surprisingly common.
3. User has `PAGER` set because they prefer another pager. We'll ignore
that preference and run `less -FRX`.
4. User has `PAGER` set because `less` isn't available on their
platform. This is uncommon except for Windows, where we'll run
`:builtin` instead of `less -FRX` by default anyway.
This may cause some users who have intentionally set and configured
`PAGER` to be frustrated that we aren't respecting that value, but it's
generally not possible to respect that value in all cases _and_ have a
consistent and usable experience out of the box for the majority of
users.
#### Alternatives considered
1. Disable color and OSC8 hyperlinks if `PAGER` is set, since we can't
be sure the pager supports the color codes.
2. Don't paginate by default if `PAGER` is set. This seems
counterintuitive, but would at least resolve the problem. Users would
assume that the `jj` CLI doesn't support paginating, and either wrap it
in a pager themselves (this is a bad outcome) or find `ui.pager` and
change the setting.
3. Set `LESS` (iff it's not set already), then invoke `PAGER`. This
means that users setting things like `LESS=i` breaks our output as well,
and cases where `PAGER` isn't 'less' aren't fixed.
Fixes#3502
Colocation is about sharing the working copy between jj and git. It's
less important where the repo is stored. I therefore think we should
not call it "colocated repo". I considered renaming it to "colocated
working copy" but that sounded awkward in many places because we often
talk about the whole workspace (repo + working copy), so "In colocated
workspaces with a very large number of branches or other refs" sounds
better than "In colocated working copies with a very large number of
branches or other refs".
Once we support colocate workspaces in non-main Git worktrees, I think
this rename will be even more relevant because then all those
workspaces share the same repo but only some of them may be colocated.
The original description suggests jj somehow converts line endings, when in fact all it
does is commit files as is (as described in the previous paragraph), whether the line
endings are CRLF or LF.
According to the discussion on Discord, streampager is still maintained.
It brings more dependencies, but seems more reliable in our use case. For
example, the streampager doesn't consume inputs indefinitely whereas minus
does. We can also use OS-level pipe to redirect child stderr to the pager.
