[Quilt-dev] Re: can't apply orign.patch from latest mm snashot

[Andrew Morton]

On Tue, 25 Mar 2008 22:49:27 +0100
trem <address@hidden> wrote:

> Andrew Morton a __crit :
> > On Tue, 25 Mar 2008 22:15:20 +0100
> > trem <address@hidden> wrote:
> >
> >   
> >> Hi
> >>
> >> I've just tried to test the mm snapshot, and the first one don't apply 
> >> (origin.patch).
> >> When I try to apply it (without quilt), I've got :
> >>
> >> |    Exposing the binary blob which is the md 'super-block' via sysfs 
> >> doesn't
> >> |    really fit with the whole sysfs model, and ever since commit
> >> --------------------------
> >> File to patch:
> >>
> >>     
> >
> > That's patch(1) being stupid interpreting the text as a diff.  Use `patch
> > -u', or quilt.
> >
> >
> >   
> Hi
> 
> Thanks for your quick answer.
> 
> patch -u works fine, but quilt (I've used quilt push -a) don't work.
> I've removed all text in origin.patch and now everything works fine.

Let's cc quilt-dev.

Guys, quilt cannot apply http://userweb.kernel.org/~akpm/origin.patch
because patch(1) misinterprets some of the changelog text as being the
start of a context diff.

quilt should pass `-u' to patch(1) to prevent this.




Even then, patch(1) still gets confused when people go and quote unified
diffs within the changelog, even though those unified diffs are indented
four spaces to the right.

So probably an ideal solution would be to write a C program which reliably
detects start-of-patch from a text file and instead of doing

        patch < file.patch

do

        cat file.patch | filter-leading-gunk | patch