Hi Phillip,
On Mon, 30 Oct 2017, Phillip Wood wrote:
> On 30/10/17 06:26, Jacob Keller wrote:
> > On Sun, Oct 29, 2017 at 8:36 PM, Junio C Hamano <gitster@xxxxxxxxx> wrote:
> >> Jacob Keller <jacob.keller@xxxxxxxxx> writes:
> >>
> >>> I am pretty confident we can fix it....
> >>
> >> I am sure we can eventually, but 3 hours is not enough soak time.
> >>
> >> I am inclined to leave the fix for 2.15.1/2.16.0 instead of delaying
> >> the release by 10 more days.
> >
> > That's fair. I'm not even sure it was introduced since the last
> > release (I tried 2.12, but not 2.13 or 2.14 manually). Thus, it likely
> > wasn't noticed for at least a release, meaning it's less important (to
> > me at least) that we provide a fix immediately, since it went
> > unnoticed this long, likely that means few people will be impacted.
>
> It is in 2.14.3, I haven't bisected but I suspect it was introduced by
> 311af5266b sequencer (rebase -i): implement the 'exec' command
>
> Running
> git rebase -x'perl -e '\''$,=$\="\n"; print grep { /^GIT_/ } sort keys
> %ENV'\' @
> Shows that the rebase--helper version also sets GIT_PREFIX as well as
> GIT_DIR, I suspect the difference is coming from differences in the
> setup for builtin commands vs external commands. The shell version and
> the rebase--helper version set GIT_CHERRY_PICK_HELP, GIT_EDITOR,
> GIT_INTERNAL_GETTEXT_SH_SCHEME, GIT_REFLOG_ACTION
Indeed, when you look at git_dir_init in git-sh-setup, you will see that
Unix shell scripts explicitly get their GIT_DIR turned into an absolute
path.
So my suggested patch is wrong, and it should be more along the lines of
struct strbuf buf = STRBUF_INIT;
const char *child_env[] = { NULL, NULL };
strbuf_addf(&buf, "GIT_DIR=%s", absolute_path(get_git_dir()));
child_env[0] = buf.buf;
...
strbuf_release(&buf);
Jake, can I still take you up on taking it from here?
Ciao,
Dscho
