Re: [PATCH 57/89] sched/headers: Split <linux/sched/task_stack> out of <linux/sched.h>

[Ingo Molnar]

* Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:

> Hi Ingo,
> 
> On Tue, Feb 7, 2017 at 9:13 AM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
> > * Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> >> On Mon, Feb 6, 2017 at 10:54 PM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
> >> > Wondering why Git allowed me to be so stupid with those leftover merge markers.
> >> > Git usually doesn't even allow me to commit them so I have these tuned out as a
> >> > possibility. This was just a regular git rebase -i flow, to back-merge fixes and
> >> > reorder/squash patches - nothing fancy that I remember - only the occasional
> >> > --onto option. I'm using Git 2.7.4.
> >>
> >> Git complains about the merge conflicts, and refuses to commit the result
> >> as long as you haven't resolved them, but it will happily commit everything
> >> you add using "git add -u", incl. merge markers.
> >
> > Hm, it should really force that via 'git add -f' or such. The merge markers are
> > _very_ infrequent as naturally occuring source code lines even on a per line basis
> > - and especially the combination of them should be exceedingly unique.
> 
> They were very infrequent, until we switched to RST for documentation,
> causing false positives when searching for "^[<=>].*" in vim...

But the exact merge conflict pattern is generated by Git, and it's far more 
specific than the "^[<=>].*" pattern, right?

So it should be possible to disambiguate?

> > I frequently use:
> >
> >         git add $(git ls-files -m)
> 
> That's identical to "git add -u", right?

Indeed, I'm bad at remembering one letter shortcuts: why is what is '-m' in 
git-ls-files called '-u' in git-add? ;-)

BTW., would 'git add -u' have prevented my mistake?

Thanks,

	Ingo