Re: [PATCH v7] git-prompt: make colourization consistent

["SZEDER Gábor" <szeder.dev@gmail.com>]

On Thu, Jun 09, 2022 at 11:29:25AM -0700, Junio C Hamano wrote:
> Joakim Petersen <joak-pet@online.no> writes:
> 
> > On 09/06/2022 11:03, SZEDER Gábor wrote:
> >> This patch seems to break colorization when __git_ps1() is invoked
> >> from $PROMPT_COMMAND:
> >>    ~/src/git (master)$ echo $PROMPT_COMMAND
> >> __git_ps1 "\[\e]0;\w - Terminal\a\e[01;32m\]\h\[\e[01;34m\] \w" "\[\e[01;34m\]\$\[\e[00m\] " " \[\e[01;34m\](%s\[\e[01;34m\])"
> >>    ~/src/git (master)$ git checkout 9470605a1b
> >>    HEAD is now at 9470605a1b git-prompt: make colourization consistent
> >>    ~/src/git ((9470605a1b...))$ source contrib/completion/git-prompt.sh
> >>    ~/src/git (\[\e[31m\](9470605a1b...)\[\e[0m\])$ # uh-oh
> >>    ~/src/git (\[\e[31m\](9470605a1b...)\[\e[0m\])$ git checkout 9470605a1b^
> >>    Previous HEAD position was 9470605a1b git-prompt: make colourization consistent
> >>    HEAD is now at 2668e3608e Sixth batch
> >>    ~/src/git (\[\e[31m\](2668e3608e...)\[\e[0m\])$ source contrib/completion/git-prompt.sh
> >>    ~/src/git ((2668e3608e...))$ # Looks good.
> >> 
> >
> > While I did test this on my own prompt for v6 (which is identical to v7
> > in terms of code) and not see any breakage, I have the same issue with
> > v7. Maybe I forgot to re-source the changed git-prompt.sh. Either way,
> > The issue stems from $b being wrapped in $__git_ps1_branch_name and then
> > back into itself after colouring. Moving this wrapping to before colour
> > is applied fixes this. I will submit a v8 shortly.
> 
> As the topic is already in 'next' (and presumably that is how SZEDER
> noticed the breakage),

Indeed.  I usually use a custom git built from 'next' with a couple of
my forever-WIP topics merged on top, and I just happened to build and
deploy a version with this patch already merged the other day, with
the additional stroke of luck that I opened a new terminal window
(what I normally rarely do) whose shell sourced the buggy prompt
script.

I did notice this patch being discussed on the ML, and found the
amount of changes to the expected output in the tests somewhat
suspicious, but, alas, haven't managed to take a closer look before
the patch went into 'next'.  Still hasn't, actually, but FWIW Joakim's
fix (as 0e5d9ef395 in 'seen') does work for me.


Thanks,
Gábor