Conversation
- include <Eigen/LU>, otherwise Eigen::Martix<>::inverse() would be undefined - fix forward declares. if a struct is forward declared as a class, function symbols might be erroneously marked external in the object file
|
Hi, thanks a lot, it's great to see a version that can work on Windows! I think merging it with master could be complicated without breaking building on other systems, and unfortunately I'll be very busy the next weeks. But I think it is already very useful to other people as is, with this PR open :) |
|
Hello, I couldn't find a more appropriate discussion thread for this so I thought I'll comment here so that people interested in this might be able to find it. I am using this fork on Windows, compiled on Windows 10 20H2 with Visual Studio 2019. When building dm-vio, I was getting a bunch of "M_PI undeclared identifier" errors, which I was able to fix by adding _USE_MATH_DEFINES to preprocessor definitions. With the associated fork of Pangolin 0.6, I was getting build failures due to errors in include\mpark\variant.hpp. When using the TUMVI datasets, I noticed that the symbolic link in the dso/cam0 folders do not work on windows, so I had to move the images folder to dso/cam0 to solve the problem. Once all this was sorted, I was able to successfully run dmvio_dataset.exe with the following windows command line dmvio_dataset.exe ^ I hope anyone interested may find this useful! |
Hi,
I've made a basic Windows port of DM-VIO. I wanted to check how much I have broken in the Linux build (e.g. I most likely broke sse2neon in the CMake), but I regrettably haven't got around to it in the past months so I've decided to open this pull request so you can take a look at it as is. No hard feelings if this doesn't get merged :)
My changes fall into these 3 categories:
Let me know if you are interested in this port at all. If you have a specific issue with it, I'll see what I can do about it.