Hi all,
If it helps, I noticed some times those test failures and I think they are
caused by the JAI Medialib acceleration.
You should try building GeoTools using the JVM parameter
-Dcom.sun.media.jai.disableMedialib=true.
Hope it helps.
Regards,
Nicola Lagomarsini.
2015-02-13 17:47 GMT+01:00 A
On Thu, Feb 12, 2015 at 11:41 PM, Tyler Battle
wrote:
> I've been having the same tests fail. I tried it in interactive mode and
> could spot differences in some of the images. Disabling these checks on OS
> X seems like the best option to me.
>
Could you share an example of the differences?
Ch
Francesco,
this documentation is obsolete. Andrea recently ported Resemble.js to
Java and updated our image tests on master to use it, so they now no
longer user perceptualdiff, work on all platforms, and are always
enabled (unless -Dskip.image.tests is specified).
Kind regards,
Ben.
On 13/02
Ben, thank you.
Building with the profile “interactive.image” also did the trick to ask for
comparison and it doesn’t fail anymore there. Documentation says that image
tests need the command perceptualdiff whom I did not have. Maybe those were
failed for this reason…after brewing that I will tr
Francesco,
the new Maven flag to disable image tests is:
-Dskip.image.tests
As discussed in this thread, disabling image tests may at this time be
necessary on OS X.
Kind regards,
Ben.
On 13/02/15 10:58, Justin Deoliveira wrote:
> On Thu, Feb 12, 2015 at 2:54 PM, Andrea Aime
> wrote:
>
>> Ye
I've been having the same tests fail. I tried it in interactive mode and
could spot differences in some of the images. Disabling these checks on OS
X seems like the best option to me.
On 12 February 2015 at 13:58, Justin Deoliveira wrote:
>
>
> On Thu, Feb 12, 2015 at 2:54 PM, Andrea Aime > wr
On Thu, Feb 12, 2015 at 2:54 PM, Andrea Aime
wrote:
> Yet... I don't see a way out of this, short of having a OSX build server.
> The renderer is still quote under-tested, whixh also makes it diffucult to
> chance/refactor it.
>
Right. So I think the profile to disable the checks on OS X is a go
Yet... I don't see a way out of this, short of having a OSX build server.
The renderer is still quote under-tested, whixh also makes it diffucult to
chance/refactor it.
Cheers
Andrea
Il 12/feb/2015 21:55 "Justin Deoliveira" ha scritto:
> The last time I brought this up I did this and indeed the
The last time I brought this up I did this and indeed the images looked
fine, but the perceptual diff reported many of the pixels were different.
It was mentioned back then that we could go through the failing tests and
up the tolerance value but you also mentioned that there were more such
tests c
Guys, check the two images running an interactive test (on the phone now,
do not remember the sysvar) and if the two images do not have significante
differences, up the tolerance param accordingly.
Cheers
Andrea
Il 12/feb/2015 21:30 "Justin Deoliveira" ha scritto:
> I consistently get these exac
I consistently get these exact same failures on OS X so I think this is a
known issue, although Jody claims to not have this problem, so not sure
what his secret sauce is.
That said I think someone recently added a built profile to ignore these
failures.
On Thu, Feb 12, 2015 at 12:09 PM, Ben Cara
Francesco,
a full build of master with -Dall passes for me on debian/sid amd64.
This could be a Mac-only failure. I think Jody encountered many of these
in the past. (I had to fix many font problems to get gt-render to pass
on OpenJDK.)
Jody, are you still on Mac?
Francesco, are you able to r
Is it just me who isn’t able to build currently geotools? I’m facing with the
failure in the subject that reports the message below:
Results :
Failed tests:
GridCoverageRendererTest.testReprojectGoogleMercator:416 Images are visibly
different, found 3220 different pixels, against a threshold o
13 matches
Mail list logo