vgx11: fix translate offset

[sbinet]

Address issue #179 by translating the image to (0,-h) so (0,0) is at the
bottom-left corner for the X11 backend.

go test ./... will skip the X11 test if $DISPLAY is not defined.

[kortschak]

No space after comma.

[sbinet]

done

[kortschak]

I feel unsatisfied by the fact that the test is there but not run. It seems to me that a better approach would be to more heavily rely on vgimg (possibly extending it) to create the vgimg.Canvas and thereby trust the test for its creation in vg.

[sbinet]

alternatively, I could try to have the travis-ci actually run the X11 test.

it seems to me like the approach you're advising would be more "testing vgimg" or "testing how xgbutil is using image" rather than actually testing how an X11 window is being displayed.

[kortschak]

I'm happy if you can get travis to test the X11 behaviour.

[sbinet]

PTAL.

now the test is run if $DISPLAY is defined (which should work as a poor man's way to detect whether X11 is available. more refined ways could be devised if/when deemed necessary)

[kortschak]

Do we need the windows build now? It seems not.

[sbinet]

you mean the !windows tag in the canvas_test.go file?
probably not.

[kortschak]

The entire canvas_nox11.go file.

[kortschak]

The !windows constraint does need to be there - not for us, but for people on Windows who do go test themselves.

[sbinet]

the canvas_nox11.go file still needs to be there I think.
at least, when I tested on my linux+darwin machines with:
$ go test -v -tags=windows ./...
it would fail (w/o that canvas_nox11 file) complaining there were no source files for package vgx11 (paraphrasing.)

you are right though that even with the hasX11 function, the !windows build tag is still necessary (otherwise, compilation of the test binary would fail. d'oh!)

[kortschak]

If there is only source for a non windows build environment the directory will be skipped when GOOS=windows.

[sbinet]

ok. I wasn't indeed completely sure this wasn't an artifact of my unorthodox way of simulating a Windows environment.

[kortschak]

Nope, you're right. That does not seem reasonable. But it is what it is.

[sbinet]

ok. removed

[kortschak]

I going to have a look at build and see why this is how it behaves.

[kortschak]

Sorry, I'd really like this to be "Package vgx11 implements X11 vg support." or words to that effect. The basis for the implementation is secondary and should not be in the first sentence.

[sbinet]

done. (I was just following the other vg-foo backends descriptions)

[kortschak]

I think it is a result of doing -tags=windows. When I change the build constraints to windows (without the bang), I get that error if the test is invoked in vgx11, but not if I invoke go test ./... in plot. I think we are good.

LGTM apart from minor comment. Please wait for @eaburns though.

[eaburns]

return os.Getenv("DISPLAY") != ""

[sbinet]

done.

[eaburns]

LGTM.
I assume that you two worked out all of the build issues.

[kortschak]

I'm satisifed that it's correct.

[sbinet]

I'll rebase, squash and make a nice commit message. (tomorrow)