Actually, I just realized that what I was doing with line-init is not
actually solving my problem. So, here's a new version of the patch
that simply lets chpwd call precmd. See attachment.

On Thu, Jun 10, 2021 at 4:47 PM Marlon Richert <marlon.richert@gmail.com> wrote:
>
> On Thu, Jun 10, 2021 at 12:30 PM Roman Perepelitsa
> <roman.perepelitsa@gmail.com> wrote:
> >
> > There is //'%%'/% in a few places. What's the logic behind it?
>
> Some prompt escape codes (such as %b for bold) need %% when used in
> vcs_info -- except when they are used in nvcsformats.
>
>
> > I think invoking prompt_vcs_setup twice in the same shell will produce
> > an error. I haven't tried it though.
>
> I just tried it and I didn't get any errors. Why do you think it would
> produce an error?
>
>
> > prompt_vcs_chpwd always prints at least one line to the TTY regardless
> > of configuration options. Is this intended?
>
> Yes, that was my intention, but I can see how it could be useful to be
> able to customize that. I'll change it.
>
>
> > Why is prompt_vcs_precmd checking whether prompt_vcs_fd is a readable
> > fd? How can it not be?
>
> I don't know how or why, but if I don't check for that, then I get
> this error repeatedly:
>
> prompt_vcs_precmd:6: failed to close file descriptor 11: bad file descriptor
>
>
> > There is a comment in prompt_vcs_precmd that says it's going to kill a
> > process while in fact it closes the file descriptor. That may
> > *eventually* kill the process when it decides to print something
> > (assuming it hasn't blocked SIGPIPE or is checking for write errors).
>
> True. I'll change it to actually kill the process, too.
>
>
> > What is the purpose of the "Info is not for current dir" check? Is
> > this to account for the possibility of `cd` from a zle widget? If so,
> > perhaps handle this by hooking chpwd?
>
> The purpose is to account for the possibility that we receive the info
> _after_ we've already changed to a new dir. However, if I correct the
> piece of code you mentioned previously, then that shouldn't be
> necessary.
>
>
> > prompt_vcs_fd-widget has a check for [[ -n $1 ]]. When is it false?
>
> When prompt_vcs_fd-widget is being called incorrectly. :) Perhaps it
> should not return 1 in that case, though.
>
>
> Thanks again. New version of the patch attached.