I doubt ERT can be changed, it probably sets `debug-on-error' for a reason,
i.e. without it it simply couldn't work. As far as I can tell, the only
good option is to never trigger an error just because a directory looks
like it might be Git, but Git is not installed.

Also consider this from a user point of view. Let's say I have nothing to
do with programming at all and don't have Git installed. Someone emails me
a cool Emacs package as a tarball and accidentally archives `.git' along,
because he is a developer. I unpack it, open some file and now my Emacs
warns me that Git is not installed. But why? I didn't even ask it to do
anything. It *itself* decided to do something (because it saw a `.git'
directory), failed (Git is not installed), and now warns *me*. From my
point of view, if there is a failure for action that wasn't direct result
of user order, Emacs should stay silent.

In other words, I think there are two sane options here:

1) Simply check if Git is installed before doing anything Git-related from
`find-file-hook' callback. If it is not installed, just silently don't do
anything. Reserve errors and warnings until the user actually ask you to do
something Git-related, not simply to open a file.

2) Make sure that this error is always demoted to a warning, even if
`debug-on-error' is t.

I like option 1 better.

Paul

On Wed, 6 Sept 2023 at 14:52, Dmitry Gutov <dmitry@gutov.dev> wrote:

> On 06/09/2023 15:49, Paul Pogonyshev wrote:
> > It feels conceptually wrong to require all tests that open files to
> > rebind `vc-handled-backends'. This is not what they are testing. It also
> > depends on knowing particular Emacs quirks (which I, for example, didn't
> > know one day earlier). If those were to change in some way, would all
> > tests everywhere need to accomodate?
>
> But does that mean that VC needs to be changed, or ERT?
>
> E.g. ERT has an interactive mode where it drops into a debugger for the
> user ('d'), but does it need to enable the debugger for its simple runs
> as well?
>