VTK doesn't build (freetype headers problem)

[remram44]

OS X 10.9 build stuck on this error, building VTK:

  /Users/remirampin/Documents/programming/uvcdat/build-e03fa52b/build/VTK/ThirdParty/freetype/vtkfreetype/include/freetype/internal/t1types.h:81:17:
  error: redefinition of 'T1_EncodingType_'

    typedef enum  T1_EncodingType_
                  ^


  /Users/remirampin/Documents/programming/uvcdat/build-e03fa52b/install/Externals/include/freetype/t1tables.h:506:17:
  note: previous definition is here

    typedef enum  T1_EncodingType_

Full log

[aashish24]

@remram44 @doutriaux1 Can you try this branch in devel: vtk_use_cdat_packages

[aashish24]

Please merge if works.

[doutriaux1]

@aashish24 I never had the issue, so let's wait to hear from @remram44 first ok?

[aashish24]

Sure.. but you may want to make sure that it does not break build on your system. Checking on that would be helpful as well.

[remram44]

I'm building right now

[aashish24]

Thanks. Make sure that run you ctest afterwards just to make sure it didn't break your code.

cd build_dir
ctest -jN

[williams13]

If you can, please run CTEST before and after to see if there's any difference.

From: Aashish Chaudhary <notifications@github.commailto:notifications@github.com>
Reply-To: UV-CDAT/uvcdat <reply@reply.github.commailto:reply@reply.github.com>
Date: Monday, September 8, 2014 9:09 AM
To: UV-CDAT/uvcdat <uvcdat@noreply.github.commailto:uvcdat@noreply.github.com>
Subject: Re: [uvcdat] VTK doesn't build (freetype headers problem) (#594)

Thanks. Make sure that run you ctest afterwards just to make sure it didn't break your code.

cd build_dir
ctest -jN

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/594#issuecomment-54844369.

[aashish24]

that is correct. You want to run ctest before as well to be precise. But I assumed that you already have done it and for most of us 8 tests are failing.

• Aashish

[aashish24]

@remram44 Any update?

[remram44]

Unfortunately I won't be able to build on OS X until tonight, however running CTest on Debian showed a lot more failing tests than expected.

[williams13]

This is not good! Can you send us the CTEST "Failed" list?

From: Remi Rampin <notifications@github.commailto:notifications@github.com>
Reply-To: UV-CDAT/uvcdat <reply@reply.github.commailto:reply@reply.github.com>
Date: Monday, September 8, 2014 2:16 PM
To: UV-CDAT/uvcdat <uvcdat@noreply.github.commailto:uvcdat@noreply.github.com>
Cc: Dean Williams <williams13@llnl.govmailto:williams13@llnl.gov>
Subject: Re: [uvcdat] VTK doesn't build (freetype headers problem) (#594)

Unfortunately I won't be able to build on OS X until tonight, however running CTest on Debian showed a lot more failing testshttps://gist.github.com/remram44/6c8302cdaf2d3350e47d than expected.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/594#issuecomment-54888652.

[remram44]

Built on OS X, results (29% tests passed, 200 tests failed out of 282).

[williams13]

OS X 10.8 or 10.9?

From: Remi Rampin <notifications@github.commailto:notifications@github.com>
Reply-To: UV-CDAT/uvcdat <reply@reply.github.commailto:reply@reply.github.com>
Date: Monday, September 8, 2014 7:21 PM
To: UV-CDAT/uvcdat <uvcdat@noreply.github.commailto:uvcdat@noreply.github.com>
Cc: Dean Williams <williams13@llnl.govmailto:williams13@llnl.gov>
Subject: Re: [uvcdat] VTK doesn't build (freetype headers problem) (#594)

Built on OS X, resultshttps://gist.github.com/remram44/81357eabd3da05089df0 (29% tests passed, 200 tests failed out of 282).

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/594#issuecomment-54915634.

[remram44]

This is 10.9.

[aashish24]

Looks like initial attribute issue:

/Users/remirampin/.uvcdat/initial.attributes

[williams13]

Yes, we need to get the new initial.attributes file out so others can avoid this problem.

From: Aashish Chaudhary <notifications@github.commailto:notifications@github.com>
Reply-To: UV-CDAT/uvcdat <reply@reply.github.commailto:reply@reply.github.com>
Date: Monday, September 8, 2014 10:17 PM
To: UV-CDAT/uvcdat <uvcdat@noreply.github.commailto:uvcdat@noreply.github.com>
Cc: Dean Williams <williams13@llnl.govmailto:williams13@llnl.gov>
Subject: Re: [uvcdat] VTK doesn't build (freetype headers problem) (#594)

Looks like initial attribute issue:

/Users/remirampin/.uvcdat/initial.attributes

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/594#issuecomment-54925264.

[doutriaux1]

@aashish24 I have a branch that fixes the "no initial.attributes" bug.

[doutriaux1]

@aashish24 and another one with a touched up initial.atrributes 614 and 611 are the branches.

@williams13 you sent me some templates, etc.. yesterday, is that all you need for diagnostics? Or do you have more in the work?

[williams13]

For diagnostics, this is all I have for now. Either user UVWG (lower case uvwg), UVWG_1of3, UVWG_2or3, and UVWG_3of3. Or some other descriptive name. I am okay with it as long as I can call it.

From: Charles <notifications@github.commailto:notifications@github.com>
Reply-To: UV-CDAT/uvcdat <reply@reply.github.commailto:reply@reply.github.com>
Date: Tuesday, September 9, 2014 8:14 AM
To: UV-CDAT/uvcdat <uvcdat@noreply.github.commailto:uvcdat@noreply.github.com>
Cc: Dean Williams <williams13@llnl.govmailto:williams13@llnl.gov>
Subject: Re: [uvcdat] VTK doesn't build (freetype headers problem) (#594)

@aashish24https://github.com/aashish24 and another one with a touched up initial.atrributes 614 and 611 are the branches.

@williams13https://github.com/williams13 you sent me some templates, etc.. yesterday, is that all you need for diagnostics? Or do you have more in the work?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/594#issuecomment-54984504.

[aashish24]

@remram44 Can you try this:

git fetch --all -p
git chckout master
git reset --hard origin/master
git merge --no-ff devel/vtk_use_cdat_packages

then build and test as ususal.

[remram44]

Alright, merging uvcdat/master (4d6ee33), and devel/vtk_use_cdat_packages (582a775). I didn't think our Git workflow could get worse...

I set CDAT_BUILD_MATPLOTLIB to on, building now on Debian and OS X 10.9.

[remram44]

Results on Debian: https://gist.github.com/remram44/84045ef50a2af1adf7c1 (76% tests passed, 68 tests failed out of 286)
Results on OS X 10.9: https://gist.github.com/remram44/f75dd00d9b1896085da5 (86% tests passed, 40 tests failed out of 286)

[williams13]

You need to do the following:

cd HOME

mv .uvcdat .uvcdat.old

source /Users/williams13/work/uvcdate/build_uvcdat/install/bin/setup_runtime.sh (point to wherever you built UV-CDAT)

uvcdat

This will generate a new .uvcdat directory with the appropriate initial.attribute file

cd /Users/williams13/work/uvcdate/build_uvcdat/ (go to the directory where you built UV-CDAT)

ctest –j4

Let us know what you get.
From: Remi Rampin <notifications@github.commailto:notifications@github.com>
Reply-To: UV-CDAT/uvcdat <reply@reply.github.commailto:reply@reply.github.com>
Date: Wednesday, September 10, 2014 12:02 PM
To: UV-CDAT/uvcdat <uvcdat@noreply.github.commailto:uvcdat@noreply.github.com>
Cc: Dean Williams <williams13@llnl.govmailto:williams13@llnl.gov>
Subject: Re: [uvcdat] VTK doesn't build (freetype headers problem) (#594)

Results on Debian: https://gist.github.com/remram44/84045ef50a2af1adf7c1 (76% tests passed, 68 tests failed out of 286)
Results on OS X 10.9: https://gist.github.com/remram44/f75dd00d9b1896085da5 (86% tests passed, 40 tests failed out of 286)

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/594#issuecomment-55165814.

[remram44]

Alright, new logs (same results, obviously):

Results on Debian: https://gist.github.com/remram44/ea8f68e9bb81a01f5466 (76% tests passed, 68 tests failed out of 286)
OS X 10.9: https://gist.github.com/remram44/3af7f200ff3af849c77d (86% tests passed, 40 tests failed out of 286)

[williams13]

I cannot speak to Debian. Didn't know we had someone building on this flavor of Linux. But the Mac 10.9 should have the results below.

@aashish and @doutriaux how can we get Remi on the same page as the rest of us. WE have got to get his changes into master!

All the

282/286 Test #20: testUnstructGrid ................................... Passed 5.55 sec
Start 269: CDMS_Test_09
283/286 Test #269: CDMS_Test_09 ....................................... Passed 0.20 sec
284/286 Test #28: testTasRegrid ...................................... Passed 5.25 sec
285/286 Test #26: testGridTypes ...................................... Passed 5.35 sec
286/286 Test #42: testReverseOrderInterp ............................. Passed 5.32 sec

98% tests passed, 6 tests failed out of 286

Total Test time (real) = 605.19 sec

The following tests FAILED:
3 - cdms_verify (Failed)
4 - cdms_load_and_plot_axis_variable (Failed)
5 - run_uvcdat_gui_tests (Failed)
54 - vcs_test_dump_json (Failed)
239 - vcs_test_aspect_ratio (Failed)
286 - diags_test_01 (Failed)

From: Remi Rampin <notifications@github.commailto:notifications@github.com>
Reply-To: UV-CDAT/uvcdat <reply@reply.github.commailto:reply@reply.github.com>
Date: Wednesday, September 10, 2014 12:54 PM
To: UV-CDAT/uvcdat <uvcdat@noreply.github.commailto:uvcdat@noreply.github.com>
Cc: Dean Williams <williams13@llnl.govmailto:williams13@llnl.gov>
Subject: Re: [uvcdat] VTK doesn't build (freetype headers problem) (#594)

Alright, new logs (same results, obviously):

Results on Debian: https://gist.github.com/remram44/ea8f68e9bb81a01f5466 (76% tests passed, 68 tests failed out of 286)
OS X 10.9: https://gist.github.com/remram44/3af7f200ff3af849c77d (86% tests passed, 40 tests failed out of 286)

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/594#issuecomment-55173083.

[remram44]

If you look closely at LastTest.log, you'll see things like that: image fails comparison threshold 0.0736404483779>0.05

Are the thresholds too strict?

[aashish24]

@williams13 Remi's changes are in. This is my change (well my fix to the bug).

@remram44 the threshold is bit strict but on Mac we should have gotten close to what we all get. Can you send me the diff.png images in your build directory (just few of them?)

@Looks like the build issue is fixed but we need to make sure that it didn't cause these new test failures. @doutriaux1 can you try this branch?

[doutriaux1]

The threshold is definitely not too strict because adding the logo to the pictures did not break ANY test last time we merged it back accident, that was actually scary!

C.

From: Remi Rampin <notifications@github.commailto:notifications@github.com>
Reply-To: UV-CDAT/uvcdat <reply@reply.github.commailto:reply@reply.github.com>
Date: Wednesday, September 10, 2014 at 1:10 PM
To: UV-CDAT/uvcdat <uvcdat@noreply.github.commailto:uvcdat@noreply.github.com>
Cc: Charles Doutriaux <doutriaux1@llnl.govmailto:doutriaux1@llnl.gov>
Subject: Re: [uvcdat] VTK doesn't build (freetype headers problem) (#594)

If you look closely at LastTest.loghttps://gist.githubusercontent.com/remram44/3af7f200ff3af849c77d/raw/osx10.9-LastTest.log, you'll see things like that: image fails comparison threshold 0.0736404483779>0.05

Are the thresholds too strict?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/594#issuecomment-55175311.

[doutriaux1]

Remi,

In your build directory under testing/vcs there should be the pictures produced by failed test and also the same file with a _diff.png that shows (poor man implementation) the diff with original.

Please take a look to see if they are very different or not. Send us some example of ones that you think are identical and where failure is due to bad threshold.

Thanks,

C.

On 09/10/2014 01:10 PM, Remi Rampin wrote:

If you look closely at LastTest.loghttps://gist.githubusercontent.com/remram44/3af7f200ff3af849c77d/raw/osx10.9-LastTest.log, you'll see things like that: image fails comparison threshold 0.0736404483779>0.05

Are the thresholds too strict?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/594#issuecomment-55175311.

[remram44]

Here's a test (vcs_test_gen_meshfill) failing with 0.0504090062807>0.05:

image:

diff:

Could be an antialiasing or font issue?

[aashish24]

Seems like it. Can you turn off the anti-aliasing on your system driver? Just for testing

[doutriaux1]

But @aashish24 we do turn off antialiasing (per your instructions a while back). Do we not want to set

      self.renWin.SetMultiSamples(0)

[remram44]

Can you turn off the anti-aliasing on your system driver?

This is the only setting I could find:

didn't change anything. How do I disable antialiasing?

[doutriaux1]

@remram44 I don't think it's the issue, we are turning off antialiasing.

[doutriaux1]

@remram44 @aashish24 ok it is VTK related... I just built this branch on my ubunutu and now I get the same failed tests as @remram44 ! Aashish how can these flags affect the png output so much. Not merging until we figured out what's going on. @aashish24 did you run ctest on your computer when you pushed these changes?

[doutriaux1]

same on mac, tests fail. @aashish24 did it work for you?

[aashish24]

It did work for me on Ubuntu but lets not merge unless we figure it out at this point. One thing would be is that now we are using freetype from cdat. I am wonderin gif that's causing the images to be sligthtly diffirent. Can you post the diff image? Thanks for testing it.

[doutriaux1]

@aashish24 @remram44 already posted the diff image

[aashish24]

That cannot be diff image if the diffirence is that small.?

[doutriaux1]

the second one.

[aashish24]

Ahh... got it -;) yes it is the text rendering. Can you show me where the text rendering code is?

[doutriaux1]

if vcs2vtk.py around line 698.

def genTextActor(renderer,string=None,x=None,y=None,to='default',tt='default',cmap=None)

I'm using TextActor. Maybe we should be using textActor2D?

[doutriaux1]

if vcs2vtk.py around line 698.

def genTextActor(renderer,string=None,x=None,y=None,to='default',tt='default',cmap=None)

I'm using TextActor. Maybe we should be using textActor2D?

[aashish24]

Let me have a look

[aashish24]

@doutriaux1 what font is used by the vcs?

vcs.elements["font"][vcs.elements["fontNumber"][tt.font]]

Where do I find that information?

[doutriaux1]

Packages/vcs/Lib/init.py

for nm,fnt in [
    ("default","AvantGarde-Book_Bold.ttf"),
    ("Clarendon","Clarendon.ttf"),
    ("Courier","Courier.ttf"),
    ("Helvetica","HelvMono.ttf"),
    ("Adelon","Adelon_Regular.ttf"),
    ("Times","Times_CG_ATT.ttf"),
    ("Arabic","Arabic.ttf"),
    ("Chinese","Chinese_Generic1.ttf"),
    ("Greek","Athens_Greek.ttf"),
    ("Hebrew","hebrew.ttf"),
    ("Russian","Russian.ttf"),
    ("Maths1","jsMath-msam10.ttf"),
    ("Maths2","blex.ttf"),
    ("Maths3","jsMath-wasy10.ttf"),
    ("Maths4","blsy.ttf"),
    ("AvantGarde","AvantGarde-Book_Bold.ttf"),
    ]:
  i+=1
  pth  = os.path.join(os.environ["HOME"],d,fnt)
  pthe = os.path.join(os.environ.get(e,""),fnt)
  if os.path.exists(pth):
    vcs.elements["font"][nm]=pth
    vcs.elements["fontNumber"][i]=nm
  elif os.path.exists(pthe):
    vcs.elements["font"][nm]=pthe
    vcs.elements["fontNumber"][i]=nm

[doutriaux1]

In [3]: x.getfontname(1)
Out[3]: 'default'

In [4]: x.getfontname(2)
Out[4]: 'Clarendon'

[remram44]

Looks like discussion on the original issue continues on #664? I don't know.