New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

VTK doesn't build (freetype headers problem) #594

Closed
remram44 opened this Issue Sep 3, 2014 · 48 comments

Comments

Projects
None yet
4 participants
@remram44
Contributor

remram44 commented Sep 3, 2014

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

@doutriaux1 doutriaux1 added this to the 2.0.0 milestone Sep 3, 2014

@doutriaux1 doutriaux1 added bug labels Sep 3, 2014

@aashish24

This comment has been minimized.

Contributor

aashish24 commented Sep 7, 2014

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

@aashish24 aashish24 added the gatekeeper label Sep 7, 2014

@aashish24

This comment has been minimized.

Contributor

aashish24 commented Sep 8, 2014

Please merge if works.

@doutriaux1

This comment has been minimized.

Member

doutriaux1 commented Sep 8, 2014

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

@aashish24

This comment has been minimized.

Contributor

aashish24 commented Sep 8, 2014

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

This comment has been minimized.

Contributor

remram44 commented Sep 8, 2014

I'm building right now

@aashish24

This comment has been minimized.

Contributor

aashish24 commented Sep 8, 2014

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

cd build_dir
ctest -jN

@williams13

This comment has been minimized.

Contributor

williams13 commented Sep 8, 2014

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

This comment has been minimized.

Contributor

aashish24 commented Sep 8, 2014

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

This comment has been minimized.

Contributor

aashish24 commented Sep 8, 2014

@remram44 Any update?

@remram44

This comment has been minimized.

Contributor

remram44 commented Sep 8, 2014

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 comment has been minimized.

Contributor

williams13 commented Sep 8, 2014

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

This comment has been minimized.

Contributor

remram44 commented Sep 9, 2014

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

@williams13

This comment has been minimized.

Contributor

williams13 commented Sep 9, 2014

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 comment has been minimized.

Contributor

remram44 commented Sep 9, 2014

This is 10.9.

@aashish24

This comment has been minimized.

Contributor

aashish24 commented Sep 9, 2014

Looks like initial attribute issue:

/Users/remirampin/.uvcdat/initial.attributes

@williams13

This comment has been minimized.

Contributor

williams13 commented Sep 9, 2014

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

This comment has been minimized.

Member

doutriaux1 commented Sep 9, 2014

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

@doutriaux1

This comment has been minimized.

Member

doutriaux1 commented Sep 9, 2014

@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

This comment has been minimized.

Contributor

williams13 commented Sep 9, 2014

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

This comment has been minimized.

Contributor

aashish24 commented Sep 10, 2014

@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

This comment has been minimized.

Contributor

remram44 commented Sep 10, 2014

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

This comment has been minimized.

Contributor

remram44 commented Sep 10, 2014

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

This comment has been minimized.

Contributor

williams13 commented Sep 10, 2014

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

This comment has been minimized.

Contributor

remram44 commented Sep 10, 2014

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

This comment has been minimized.

Contributor

williams13 commented Sep 10, 2014

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

This comment has been minimized.

Contributor

remram44 commented Sep 10, 2014

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

This comment has been minimized.

Contributor

aashish24 commented Sep 10, 2014

@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

This comment has been minimized.

Member

doutriaux1 commented Sep 10, 2014

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

This comment has been minimized.

Member

doutriaux1 commented Sep 10, 2014

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

This comment has been minimized.

Contributor

remram44 commented Sep 10, 2014

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

image:
test_vcs_gen_meshfill.png
diff:
test_vcs_gen_meshfill_diff.png

Could be an antialiasing or font issue?

@aashish24

This comment has been minimized.

Contributor

aashish24 commented Sep 10, 2014

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

@doutriaux1

This comment has been minimized.

Member

doutriaux1 commented Sep 10, 2014

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

      self.renWin.SetMultiSamples(0)
@remram44

This comment has been minimized.

Contributor

remram44 commented Sep 10, 2014

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

This is the only setting I could find:
System Preferences/General
didn't change anything. How do I disable antialiasing?

@doutriaux1

This comment has been minimized.

Member

doutriaux1 commented Sep 10, 2014

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

@doutriaux1

This comment has been minimized.

Member

doutriaux1 commented Sep 10, 2014

@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

This comment has been minimized.

Member

doutriaux1 commented Sep 10, 2014

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

@aashish24

This comment has been minimized.

Contributor

aashish24 commented Sep 11, 2014

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

This comment has been minimized.

Member

doutriaux1 commented Sep 11, 2014

@aashish24 @remram44 already posted the diff image

@aashish24

This comment has been minimized.

Contributor

aashish24 commented Sep 11, 2014

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

@doutriaux1

This comment has been minimized.

Member

doutriaux1 commented Sep 11, 2014

the second one.

@aashish24

This comment has been minimized.

Contributor

aashish24 commented Sep 11, 2014

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

@doutriaux1

This comment has been minimized.

Member

doutriaux1 commented Sep 11, 2014

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?

1 similar comment
@doutriaux1

This comment has been minimized.

Member

doutriaux1 commented Sep 11, 2014

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

This comment has been minimized.

Contributor

aashish24 commented Sep 11, 2014

Let me have a look

@aashish24

This comment has been minimized.

Contributor

aashish24 commented Sep 11, 2014

@doutriaux1 what font is used by the vcs?

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

Where do I find that information?

@doutriaux1

This comment has been minimized.

Member

doutriaux1 commented Sep 11, 2014

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

This comment has been minimized.

Member

doutriaux1 commented Sep 11, 2014

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

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

@aashish24 aashish24 removed the gatekeeper label Sep 12, 2014

@remram44

This comment has been minimized.

Contributor

remram44 commented Sep 17, 2014

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

@aashish24 aashish24 closed this Sep 25, 2014

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment