Comment #1 originally posted by sandwichmaker on 2012-06-12T14:48:03.000Z:

Additional information:

This problem occurs on ceres-solver 1.2.0 and it doesn't occur on 1.1.1

from ceres-solver.

Comment #2 originally posted by sandwichmaker on 2012-06-12T17:30:50.000Z:

I do not have a reproduction environment for this.
I am going to try and get g++4.7 on my desktop to see if I can reproduce it.

from ceres-solver.

Comment #3 originally posted by sandwichmaker on 2012-06-21T12:38:46.000Z:

ceres-solver 1.2.2 failed to test on Ubuntu 12.04 (i386)
I tried building it on launchpad PPA and it failed autodiff_test

2/43 Test # 2: autodiff_test .............................***Exception: SegFault 0.00 sec
[==========] Running 4 tests from 1 test case.
[----------] Global test environment set-up.
[----------] 4 tests from AutoDiff
[ RUN ] AutoDiff.ProjectiveCameraModel

  Start  3: autodiff_cost_function_test

build log:
https://launchpadlibrarian.net/108043372/buildlog_ubuntu-precise-i386.ceres-solver_1.2.2-1~precise1_FAILEDTOBUILD.txt.gz

from ceres-solver.

Comment #4 originally posted by sandwichmaker on 2012-06-21T12:48:39.000Z:

Hi Koichi,
Is this a 64bit or a 32bit system?
could you run autodiff_test manually with -logtostderr and post the results please.

from ceres-solver.

Comment #5 originally posted by sandwichmaker on 2012-06-21T12:49:01.000Z:

<empty>

from ceres-solver.

Comment #6 originally posted by sandwichmaker on 2012-06-21T12:54:05.000Z:

actually, what would help most is to build the test in debug mode and run it manually on your system. I suspect this is an eigen alignment problem, but building it in debug mode will confirm that.

from ceres-solver.

Comment #7 originally posted by sandwichmaker on 2012-06-21T13:19:33.000Z:

Koichi,

I tried this on my ubuntu precise (with some modifications that I do not have any control over) with g++ version 4.6.3-1ubuntu5 and everything worked fine.

I will have to look at the compiler flags you are using to build this.
Sameer

from ceres-solver.

Comment #8 originally posted by sandwichmaker on 2012-06-21T15:02:43.000Z:

Koichi,

I tried building with the
"-g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -Werror=format-security"

flags added to my c++ build and everything seems to be working. Could you tell me what if any changes are to the build environment when building and executing this? I am unable to reproduce this bug on my end.

My gcc is different by a minor revision, but other than that things seem to be sane.

Sameer

from ceres-solver.

Comment #9 originally posted by sandwichmaker on 2012-06-21T15:14:50.000Z:

Hi Sameer,

The following message is shown when it is built with -DCMAKE_BUILD_TYPE=Debug

2/43 Test # 2: autodiff_test .............................**Exception: Other 0.01 sec
[==========] Running 4 tests from 1 test case.
[----------] Global test environment set-up.
[----------] 4 tests from AutoDiff
[ RUN ] AutoDiff.ProjectiveCameraModel
autodiff_test: /usr/include/eigen3/Eigen/src/Core/DenseStorage.h:69: Eigen::internal::plain_array<T, Size, MatrixOrArrayOptions, 16>::plain_array() [with T = double, int Size = 16, int MatrixOrArrayOptions = 0]: Assertion `(reinterpret_cast<size_t>(array) & 0xf) == 0 && "this assertion is explained here: " "http://eigen.tuxfamily.org/dox-devel/TopicUnalignedArrayAssert.html" " *** READ THIS WEB PAGE !!! ****"' failed.

from ceres-solver.

Comment #10 originally posted by sandwichmaker on 2012-06-21T15:18:04.000Z:

This bug occurs on launchpad build server. So I think it has no changes with official packages.

from ceres-solver.

Comment #11 originally posted by sandwichmaker on 2012-06-21T15:27:50.000Z:

and this error doesn't occurs on 64bit

from ceres-solver.

Comment #12 originally posted by sandwichmaker on 2012-06-21T15:31:10.000Z:

okay now we are making some progress.

Its quite possible that our alignment stuff is still broken on 32-bit. Part of the problem is that I do not have a 32 bit dev environment/tool-chain. So this is very useful.

This may take a few days to fix, as keir or I will have to get our hands a 32-bit dev environment.

from ceres-solver.

Comment #13 originally posted by sandwichmaker on 2012-06-21T15:31:45.000Z:

Could you please file a separate bug to track this. Since this is quite different from the floating point tests that are failing in 12.10

from ceres-solver.

Comment #14 originally posted by sandwichmaker on 2012-06-22T08:22:01.000Z:

Hi,
I report 12.04 bug to http://code.google.com/p/ceres-solver/issues/detail?id=52

from ceres-solver.

Comment #15 originally posted by sandwichmaker on 2012-09-25T21:42:06.000Z:

Can you still reproduce this? I'm closing for now; reopen if it still happens.

from ceres-solver.