Log file Viewer

[15:30:49] <wiredfool> aclark: sorry about bringing you in on that buildout question

[16:57:41] <aclark> wiredfool: heh np, seems to have calmed down for now

[17:04:43] <aclark> wiredfool: thinking about leaving that meta ticket open "forever" to draw attention to the issue as requested

[17:04:49] <aclark> and hoping iElectric will fix it :-)

[17:08:05] <wiredfool> there is that

[17:10:28] <wiredfool> Homm seems annoyed about the imagecolor pr

[17:16:24] <wiredfool> I'm not inclined to commit a test that enshrines a WTF behavior

[17:20:12] <aclark> wiredfool: which ticket is that?

[17:22:41] <wiredfool> https://github.com/python-imaging/Pillow/pull/534

[17:22:48] <aclark> Ah

[17:23:29] <aclark> So it's a partial fix?

[17:23:43] <aclark> That leaves other stuff broken?

[17:24:19] <wiredfool> I think so.

[17:25:07] <wiredfool> though, he's testing for broken behavior

[17:26:24] <wiredfool> if we had a sort of 'known broken failing tests' that wouldn't pollute the day to day travis runs

[17:27:27] <aclark> Welp, he seems to be a regular-ish contributor so maybe he knows what he is doing… right, we could probably arrange that

[17:28:27] <wiredfool> to be clear, I don't have a problem with the code, only that one extraneous added test. If he hadn't added it, or if it had been a 'value wrong, but doesn't crash' sort of test, then I'd probably have already merged it.

[17:28:57] <wiredfool> I just wanted to know the reasoning behind that one.

[17:29:56] <wiredfool> maybe I'll do a pass through it soonish.

[17:32:50] <wiredfool> tho, it's behind my raspberry pi controlled chicken coop door in the project queue

[17:33:14] <aclark> IC

[17:33:16] <aclark> hah

[17:38:10] <wiredfool> with flask, websockets, and hardware hacking (tm)

[17:38:50] <aclark> nice

[17:39:08] <aclark> Are you seeing this 10.9.2 failure? I'm not AFAICT

[17:44:09] <wiredfool> I don't have a dev environment on my mavericks machine

[17:44:32] <wiredfool> I blew away all my homebrew and ports, and fink and whatever else on the upgrade

[17:46:37] <aclark> Seems like clean install fixes it, or something

[18:01:29] <wiredfool> between that and the flags in _sysconfigdata, we should be on to something

[18:02:01] <wiredfool> my _sysconfigdata.py dates from Aug25 2013, which I think is roughly maverics GM build time

[18:02:45] <wiredfool> if that's the case, then any maverics system python install will have trouble compiling any python c module.

[18:03:58] <wiredfool> a clean install of what package?

[18:40:59] <aclark> Pillow I assume

[18:41:40] <wiredfool> I'm not sure how that should change anything

[18:47:39] <aclark> If you try to pip install --upgrade without blowing away the old Pillow first, you get the error </guess>

[19:16:19] <stack> hello, I'm trying to transform an image into a circle with this code, http://pastebin.com/fVuWBKxD the problem is that the resulting circle has weird pixelated borders and is not smooth, I've tried to use some ImageFilter or Enhance method but I'm unable to get something better, any hint on that? thanks in advance.

[19:17:46] <wiredfool> what does it look like?

[19:26:21] <stack> wiredfool: https://dl.dropboxusercontent.com/u/174243215/200.jpeg https://dl.dropboxusercontent.com/u/174243215/circle.png

[19:59:55] <wiredfool> stack: so, what's happening there is that draw, for better or worse, draws based on scanlines and pixels. There's no antialiasing. So, two options:

[20:00:33] <wiredfool> 1) draw the mask at double size, then resize down using an antialias/ bicubic interpolation

[20:01:20] <wiredfool> 2) draw the ellipse with a 255 fill and a stroke of 128. I don't remember if you can do those in one call, or if yould have to do it separately

[20:01:44] <wiredfool> also, there's a known bug with ellipse that draws things slightly too large, so you might need to make the x size slightly smaller

[20:02:51] <stack> yes I saw the effect of that bug and the ticket :)

[20:03:35] <stack> I don't know what between 1 and 2 could be the better option, I'll try the second one first

[20:04:15] <wiredfool> I'm going to guess that the first is going to be smoother

[20:06:34] <stack> ok I'll try that instead, given I can't find how to do that stroke thing

[20:07:37] <wiredfool> It should be there, but that doesn't mean that it actually is.

[20:12:54] <stack> wiredfool: the first option, with an ANTIALIAS filter produces optimal results, thanks!

[20:14:17] <stack> I mean, resizin specifying an antialias filter

[20:24:23] <wiredfool> cool, glad it worked