pmxbot IRC Log Viewer

[00:14:10] <aclark> wiredfool: getting windows errors too… testing with 3.3. at the moment

[00:14:42] <wiredfool> I just got it going on python27/mac, and it's working as expected: https://gist.github.com/wiredfool/5157719

[00:15:03] <wiredfool> my 3.3 install had warnings on the python compile

[00:15:32] <wiredfool> the one test that's got the expected error is failing, but other than that, it's looking good

[00:16:59] <aclark> https://gist.github.com/aclark4life/5157733

[00:20:29] <wiredfool> I've got to run for now. That gist doesn't look like it's failing with the tiff bits

[00:33:55] <aclark> k ttyl

[19:42:12] <aclark> wiredfool: any idea about: https://github.com/python-imaging/Pillow/issues/118 ?

[19:43:34] <wiredfool> I saw that, and it seems like it happens when the python interpreter is trying to read from the Pillow/PIL directory when it should be reading from the installed version

[19:43:52] <wiredfool> and the compiled versions aren't in there or aren't correct for the python version

[19:44:46] <wiredfool> either that or a PyString_asfoo snuck by your defenses and borked the compile

[19:46:09] <wiredfool> I'm making progress on the weird lena issue. There's a spurious seek, now I just have to find it.

[20:13:38] <aclark> wiredfool: Hmmm, PyString is bad now?

[20:13:49] <wiredfool> in py3

[20:13:57] <wiredfool> it's pyunicode or pybytes

[20:14:21] <aclark> Ah, right

[20:14:23] <wiredfool> I think I have the lena issue solved.

[20:14:46] <wiredfool> I can't find the seek, but I can reset the file position just before I open the file in libtiff

[20:17:56] <aclark> There are pystrings in _imagingft.c …

[20:18:13] <wiredfool> you know, I think you handle that in py3.h

[20:18:53] <wiredfool> Is it actually compiling and putting the .so in a reasonable place?

[20:21:20] <aclark> Well, now I'm not getting the errors :-)

[20:21:30] <aclark> Blew away my Pillow dir and started over…

[20:21:35] <aclark> Maybe a recent pull fixed it

[20:22:02] <aclark> I'm running tox, which tests everything in a venv for each pyver

[20:22:18] <wiredfool> ah

[20:23:30] <aclark> Now I'm going to see if it works on Windows

[20:32:47] <aclark> Windows == Pain

[20:33:33] <wiredfool> yeah

[20:39:09] <wiredfool> aclark: pull req for the lena tiff issue is up

[21:27:48] <aclark> wiredfool: thx, any chance you could merge with the latest? otherwise I can do it

[21:33:04] <wiredfool> right

[21:44:40] <wiredfool> ok, It's merged

[21:44:49] <wiredfool> tests are still failing on histogram and jpeg

[22:38:26] <aclark> wiredfool: thanks, now maybe i can get a release out

[22:38:48] <wiredfool> are you going to do anything about the histograms?

[22:39:31] <aclark> wiredfool: which one was that?

[22:40:01] <wiredfool> the test that's blowing up travis with a buffer overflow

[22:40:07] <wiredfool> #117

[22:40:44] <aclark> Oh, no because I suspect travis is not stable…

[22:40:51] <aclark> But I'm not sure

[22:40:54] <wiredfool> It happens omm

[22:42:04] <aclark> wiredfool: It happens… what?

[22:43:24] <wiredfool> I get the buffer overflow when testing

[22:44:01] <wiredfool> 32bit and 64 bit

[22:45:13] <wiredfool> https://gist.github.com/wiredfool/5165896

[22:45:20] <aclark> Ah

[22:46:23] <wiredfool> also, I think the test_file_jpeg.test_truncated_jpeg can be dropped without issue

[22:46:25] <aclark> Yeah, I mean I suspect this will have to go out as-is and we'll wait to see if it shows up in the wild and/or for someone to figure out a fix. Not sure what else to do.

[22:46:55] <aclark> Delete the test you mean?

[22:47:04] <wiredfool> yep

[22:48:06] <aclark> Oh, because it's bogus, yeah ok.

[22:52:37] <wiredfool> the histogram thing looks odd

[22:52:58] <wiredfool> its dying in libImaging/Histo.c, line 44

[22:53:13] <wiredfool> where it's copying a newly allocated item over the image mode

[22:53:21] <wiredfool> which seems… wrong

[22:54:03] <wiredfool> line 44: https://github.com/wiredfool/Pillow/blob/master/libImaging/Histo.c

[22:54:52] <wiredfool> stacktrace: https://gist.github.com/wiredfool/5165952

[22:59:41] <aclark> Well we could easily test without the strcpy

[23:00:05] <wiredfool> I'm in gdb right now seeing what's really going on

[23:01:16] <aclark> ah cool

[23:08:25] <wiredfool> ok, I think it's a real buffer overflow

[23:08:49] <wiredfool> I was wrong, it's copying the right way, just that the dest mode is too small

[23:09:42] <aclark> ah

[23:10:05] <wiredfool> https://github.com/wiredfool/Pillow/commit/8c7c3736767e2e86bec46e239438dea0a93ea4a3

[23:10:09] <wiredfool> There's the source

[23:10:50] <aclark> ooo

[23:10:52] <aclark> nice :-)

[23:10:56] <wiredfool> or at least, that's where things got out of sync.

[23:10:59] <wiredfool> I'll fix that one

[23:12:00] <aclark> k

[23:13:47] <aclark> Have you been doing any Windows testing? I can't get anything to link

[23:13:54] <wiredfool> nope

[23:14:41] <aclark> I'm wondering if I should be using setuptools instead of distribute on windows

[23:14:52] <aclark> But I can't on Python 3

[23:15:08] <wiredfool> I've got 0 experience with py3 on windows

[23:15:23] <wiredfool> and I've only really done 2.4 and 2.6

[23:15:43] <wiredfool> It's real touchy about which compiler you're using for which python

[23:16:42] <aclark> Yeah me too :-) I can compile but then from PIL import Image fails with https://gist.github.com/aclark4life/5166087

[23:16:49] <aclark> hah

[23:16:55] <aclark> yeah I've got all that sorted, i think

[23:17:02] <aclark> maybe not

[23:20:07] <aclark> Well cgohlke seems to be able to run the tests at least

[23:21:19] <wiredfool> bingo. fixed the buffer overflow

[23:21:28] <wiredfool> now there are just two more places

[23:21:39] <wiredfool> to check, cause they look suspicious

[23:21:55] <aclark> \o/

[23:22:26] <wiredfool> I'm sorely tempted to do a strncpy there

[23:25:52] <wiredfool> ok, pretty sure that the char[4] in ImDib.h is ok, since the pnly place where it's assigned, it's forced to be either 1, L or RGB

[23:32:11] <aclark> k

[23:32:27] <wiredfool> The other one is the palette stuff

[23:32:38] <wiredfool> It filters out rgba,

[23:33:00] <wiredfool> but I'm going to bet that you can push in any image, and then it does the copy from the passed in mode

[23:33:03] <wiredfool> and then boom

[23:44:25] <wiredfool> i'm not sure about the palette one. I don't think that it's dangerous to fix it though

[23:48:11] <aclark> Go for it I say

[23:57:38] <wiredfool> Ok, it's up, travis fails with that jpeg issue and the numpy one

[23:57:45] <wiredfool> ;<

[23:58:21] <wiredfool> I'd be nice to have a travis test actually pass

Log file Viewer

Help | Karma | Search:

#pil logs for Thursday the 14th of March, 2013