tomoaki0705's profile - activity

It would be very nice if you could paste your entire log on somewhere more easy to read, such as https://pastebin.com/ o

Additional question, Is really "/usr/bin/aarch64-linux-gnu-c++" linked to "/usr/bin/aarch64-linux-gnu-g++" ? What happ

Oh, really sorry I missed your answer. I forgot to reply. Really, thank you for your reply. The fix is now merged in t

Let's separate the OpenCV problem and your configuration problem. compiler There's a very simple short code in OpenC

summary

If you have VS2013, can you tell me the result of FileStorage_DMatch from opencv_test_core ?

Description

I'm using the master branch and realized that following three tests were failing

• FileStorage_DMatch
• FileStorage_DMatch_vector
• FileStorage_DMatch_vector_vector

All of them comes from opencv_test_core

I'm using VS2012 on 64bit Win7. I digged in a while, and I came to a conclusion that this test failure is fixed behavior for VS2012. For example, in test FileStorage_DMatch, the expected result is hard coded as

"%YAML:1.0\n---\nd: [ 1, 2, 3, -1.5000000000000000e+00 ]\n"

when the actual result is

"%YAML:1.0\n---\nd: [ 1, 2, 3, -1.5000000000000000e+000 ]\n"

Please note that the number of the last 0s are the only difference. I traced a bit and I this comes from the result of sprintfof VS2012, and there is no other option to modify the result.

Question

I don't have other compiler than VS2012, but having a look to pullrequest.opencv.org, I believe this behavior is normal for VS2015 and modern GCC. I'm planning to create PR to fix this result, but before that, I'm not sure about the behavior of VS2013.

I guess the behavior of VS2010 and older will be same as VS2012. Also, the behavior of VS2017 and later will be same as VS2015. I have no idea what will happen for VS2013.

If anyone can try this situation on VS2013 (i.e. whether these tests from master will fail or not) I'm happy to hear the result.

Support of CMake 2.8.7 is now restore on PR #6551 https://github.com/Itseez/opencv/pull...

Thanks ! that explains the situation.

I faced on build error on Ubuntu 12.04. on cmake stage.

The error says CMake Error at CMakeLists.txt:608 (find_package): find_package called with invalid argument "CONFIG"

I'm on master and my cmake is version 2.8.7

I searched on internet and it seems that there are similar situation happening.

• https://bugs.webkit.org/show_bug.cgi?...
• https://github.com/code-iai/iai_kinec...
• https://bugs.freedesktop.org/show_bug...

The reason seems clear. This CONFIG keywords are supported from 2.8.8, which is later than the version I use. The package CMake is looking is for OpenCV_HAL, which I don't use.

Am I supposed to upgrade my cmake or is this a bug ?

I had a look in modules/core/src/convert.cpp There is getConvertFunc which is non-static and declared in modules/core/src/precomp.hpp

Just above getConvertFunc, there's getCvtScaleAbsFunc which is static, and this doesn't appear in modules/core/src/precomp.hpp I checked in 3.1.0 and yesterday's master/head. The situation was same in both revision.

These two functions are very similar, but the way they are declared are different. I feel that there was some mistake during in the past commits, but I couldn't figure out which way is the correct way to go. There's also getCopyMaskFunc which appears in precomp.hpp, and non-static, so I think getCvtScaleAbsFunc should become non-static and declared in precomp.hpp. If anyone knows if there are any particular reason why only this getCvtScaleAbsFunc has been defined as static, I'd like to know.

Thanks in advance.