On Sat, Feb 13, 2016 at 03:13:37PM -0500, Leo Famulari wrote:
> On Sat, Feb 13, 2016 at 06:20:01PM +0100, Andreas Enge wrote:
> > On Sat, Feb 13, 2016 at 06:01:19PM +0100, Pjotr Prins wrote:
> > > cat 0001-gnu-Add-slurm.patch |patch -p1  # same patch, different name
> > > patching file gnu/packages/parallel.scm
> > > patching file gnu/packages/patches/slurm-wlm-configure-remove-nonfree-contribs.patch
> > > Works fine. Maybe because I used an attachment previously?
> > 
> > No, that should not be a problem (after I removed the first line ">From..."
> > of the attachment). I still get:
> > 
> > patching file gnu/packages/parallel.scm
> > Hunk #1 FAILED at 2.
> > 1 out of 3 hunks FAILED -- saving rejects to file gnu/packages/parallel.scm.rej
> > patching file gnu/packages/patches/slurm-wlm-configure-remove-nonfree-contribs.patch
> > 
> > The thing that does not work is adding your copyright line. Very strange;
> > we seem to have the same white-space in the file.
> 
> I had the same issue with some of Jan's patches yesterday.
> 
> I used diffoscope to compare the non-working patch and a "should-be"
> identical patch that I generated locally. The only differences were in
> the patch headers and the encoding of the file (UTF-8 vs ISO-8859-1).
> 
> I couldn't figure out why these differences would matter and moved on,
> but it's disturbing that others are hitting the same issue!

I investigated some more. I used the patch that is upthread of what I am
replying to.

I applied the patch with `patch -p1 < ...`. I could have used `git am`
but I'm not doing it again ;)

Everything applied except for Pjotr's attribution in parallel.scm, so I
manually copied that into the file.

Then, I added all the changes to the index, committed, and used `git
format-patch` to create a patch on master. I'll show you:

$ git add -u
$ git add gnu/packages/patches/slurm-wlm-configure-remove-nonfree-contribs.patch
$ git commit -m "patch-test"
$ git format-patch -n --thread=shallow master
$ git reset --hard HEAD^ # roll back one commit
$ git am 0001-patch-test.patch
$ echo $?
0
$ diffoscope --html report.html 0001-patch-test.patch 0001-gnu-Add-slurm.patch

I've attached report.html and 0001-patch-test.patch for your inspection.

I wonder if the version of git is a factor? 2.1.4 vs 2.6.3. Jan was also
using 2.1.4, and his patches also failed *only* on the attribution line.