Return -1 in error condition instead of removing diff result.

Return -1 in error condition instead of removing diff result.

BUG=skia:2160
R=epoger@google.com

Committed: https://code.google.com/p/skia/source/detail?r=13407

[epoger, 2014-02-11 16:05:07]

When I try to view the diffs at
https://codereview.chromium.org/145123013/diff/1/tools/skpdiff/SkDiffContext.cpp
, I see:

error: old chunk mismatch

Derek, does it work for you?

[epoger, 2014-02-11 16:06:52]

LGTM

Looking at https://codereview.chromium.org/download/issue145123013_1.diff , I
think the change is fine.  But I would like to figure out why we get these
"chunk mismatches" so often...

[rmistry, 2014-02-11 16:08:15]

On 2014/02/11 16:06:52, epoger wrote:
> LGTM
> 
> Looking at https://codereview.chromium.org/download/issue145123013_1.diff , I
> think the change is fine.  But I would like to figure out why we get these
> "chunk mismatches" so often...

Something to do with git cl upload not being able to completely upload the
original and changed file contents. Typically git cl upload outputs an error
message when this happens.

[epoger, 2014-02-11 16:10:16]

On 2014/02/11 16:08:15, rmistry wrote:
> Something to do with git cl upload not being able to completely upload the
> original and changed file contents. Typically git cl upload outputs an error
> message when this happens.

I've seen this happen quite often when I upload CLs over the past ~3 weeks, and
*never* have I gotten an error message from "git cl upload".

My current suspicion is that this happens when I upload my CL from a working
directory other than trunk.

[rmistry, 2014-02-11 16:11:25]

On 2014/02/11 16:10:16, epoger wrote:
> On 2014/02/11 16:08:15, rmistry wrote:
> > Something to do with git cl upload not being able to completely upload the
> > original and changed file contents. Typically git cl upload outputs an error
> > message when this happens.
> 
> I've seen this happen quite often when I upload CLs over the past ~3 weeks,
and
> *never* have I gotten an error message from "git cl upload".
> 
> My current suspicion is that this happens when I upload my CL from a working
> directory other than trunk.

Interesting, grab me (by IM/email/in-person) when this happens again. I would
like to fix it in depot_tools if possible.

[rmistry, 2014-02-11 16:14:56]

On 2014/02/11 16:11:25, rmistry wrote:
> On 2014/02/11 16:10:16, epoger wrote:
> > On 2014/02/11 16:08:15, rmistry wrote:
> > > Something to do with git cl upload not being able to completely upload the
> > > original and changed file contents. Typically git cl upload outputs an
error
> > > message when this happens.
> > 
> > I've seen this happen quite often when I upload CLs over the past ~3 weeks,
> and
> > *never* have I gotten an error message from "git cl upload".
> > 
> > My current suspicion is that this happens when I upload my CL from a working
> > directory other than trunk.
> 
> Interesting, grab me (by IM/email/in-person) when this happens again. I would
> like to fix it in depot_tools if possible.

I ran into this right now, looks like its a problem with Rietveld timing out
when it tries to create the Patches in a Patchset.

[epoger, 2014-02-11 16:21:34]

On 2014/02/11 16:14:56, rmistry wrote:
> I ran into this right now, looks like its a problem with Rietveld timing out
> when it tries to create the Patches in a Patchset.

I guess it's all related to http://crbug.com/245890 ('Can't "git cl upload" -
error says "Base files missing"')

And no, Derek, you can't have your codereview back.

[djsollen, 2014-02-11 16:23:05]

Committed patchset #1 manually as r13407 (presubmit successful).