Build on Windows #163
Comments
|
Do you need a native build of Artery on Windows 10, or would a virtualised setup using Docker or Vagrant be suitable as well? I have no resources to actively maintain and test Artery on Windows; however, I am happy to include the necessary patches to make our build system compatible. For example, the run_artery.sh related errors are probably easy to fix, e.g. by omitting this script on Windows and/or when particular CMake generators are used. |
Dear Raphael, |
|
I am willing to support any efforts to get Artery running on Windows 10; however, I cannot take the lead on this topic because I rarely develop on Windows systems. |
|
Update: I managed to get rid of the errors what I had, and managed to generate with cmake. I had to omit run_artery.sh (I don't know if it will cause other problems for building the solution or not.) Also in the ROOT folder's CmakeLists.txt I added the line unset(CMAKE_IMPORT_LIBRARY_SUFFIX) which dealt with these types of errors: If I try to cmake --build . I get the following: I guess it can't run some exe files and some of the build will be missing, causing a lot of errors later. Update 2: If I use cmake-gui, I can set the omnet message compiler manually, so the build process will initialize it correctly. Unfortunatley I still get tons of warnings and errors. I don't know if it generates a full log of this. But the console is running out, so I cant paste it here and I don't even want to because it is super long. (Most of the content are warnings btw) |
|
Have you built OMNeT++ with VC++ or with MinGW64 shipped with OMNeT++? |
With the shipped MinGW64. |
|
Then you have to build Artery and its dependencies with MinGW64 as well, instead of MSVC. As far as I know, you cannot mix C++ libraries built with MinGW64 and MSVC. |
Dear Raphael, I would like to use the asn1 CAM message structure from your vanetza project in omnet with veins_inet. (I wanted to use artery because of this.) I put every asn1 related files in my omnet project. I include the CAM.h in my VeinsInetSampleApplication.cc, so I can access the CAM structure and I can fill CAM fields with values. Then I would like to encode it with the der_encode_to_buffer() function. The problem is, that I get undefined reference error for the function. I tried to include a lot of things. I think it should be a compiler/linker problem, because omnet++ can see the function. I noticed that it won't make any obj files from .c files, maybe thats the problem? Something should be wrong with this C++ mixed with C thing, but as every header contains extern "C" {}, I think that should work. I also tried to include the complete vanetza pack to try something similar like what I found in the built artery (I managed to build it on a virtual linux mint.) I saw that you have a wrapper around this asn1 CAM structure (which is something like a black magic for me :) ) So I would be really happy if I could just simply generate the encoded CAM message by including the asn1 CAM. Sorry for bothering you with this. And if it should be at the vanetza project, then I can create a topic there. Here is a link to a picture of my error: https://imgur.com/a/uuNZwS4 |
|
You can copy the directories vanetza/asn1/support and vanetza/asn1/its to your project. These folders contain only code generated by asn1c based on the ASN.1 grammar, i.e. it is not specific to Vanetza. I ship these files with Vanetza because it can be quite cumbersome to generate the code correctly. |
I did that exactly. My problem is, that I don't know where can I set omnetpp's compiler (clang right now) to compile and link .c source files too. I added the its and support folders just under my project's src folder and put those folders in my projects makemake options in the include path -I under compile and under link tabs. In my out folder it doesn't create any of the folders with the .o files. My generated make file looks like this: I think omnet++ just skips .c files. Workspace settings contain .c files as source files, so I don't really understand why it is not working. |
|
Hm, |
I tried that earlier, but it produces a tons of new errors like: Fixing these errors in every .cc file would be a little painful :D |
|
Okay, so the asn1 generated code cannot be compiled with a C++ compiler. You could compile the asn1 stuff with a separate Makefile and link them to a "standalone" library. Then, you can instruct Edit: You may now understand why I am using CMake instead of |
I will try it. Yeah I definitely understand it :) I'm just really a newbie in the world of compilers. The amount of options and possibilities are really high for me right now :D Opening these makefiles make me feel like I'm reading a foreign language. |
|
Okay, so I created a makefile: Which created the .o objects for me and the libasn1c.a library. I tried to add it to omnet++ with additional libraries to link (-l). I also tried to link with the created object files. I noticed, that the reqired .h files were appearing in the mingw environments include folder, so my code can recognize |
|
I think linking to a static library like libasn1c.a is the less cumbersome way to go: You don't have to worry about runtime paths to look up the library then. Your Makefile looks reasonable to me. |
I added the library path and the library itself in makemake options under compile tab Include directories and in Link tab Additional libraries to link with. and the modified one:
Edit: the makefile generating cmd: My makefile was using clang to create the library. And omnet++ uses clang++ is that a problem? |
|
Maybe it works with |
Okay I think it found it. Now I get some new errors (just in case if I didn't have enough of them :) ): I found this: https://sourceforge.net/p/asn1c/discussion/357921/thread/de1dfec6/ Shall I just exclude the GeneralizedTime.c from the library build? I feel like it's a never ending error train what Im sitting on. |
|
Just try omitting GeneralizedTime.c then. Beware, there be dragons… |
Dragons like it was built without errors in Omnet++ now, but the simulation fails without any error message, just saying Simulation terminated with exit code: -1073741819 ? :D I will try to go in debug mode and find out what the hell is happening. If I just use the CAM_t I can assign values to anything it runs without a problem. When I call the encode function it crashes. EDIT: |
|
Hello @OrmandiTamas, how did you solve the -1073741819 error? AFAIK it is 0xC0000005 in hex which is an error Regards, |
Hello, unfortunatley I dont think you have the same problem. If I remember well, my issue was mainly related to OMNeT++ and not knowing how to use the CAM structure properly. Thats why I got the memory access violation errors. |
|
Thanks for quick response. I think the problem is not the same but the solution might be. I have found your paper about generating CAM messages directly and I think I will solve it as you did - fill in the CAM_t structure directly without using the Vanetza's wrapper classes. I might have been using the structures somehow incorrectly too. |
|
Dear @OrmandiTamas , |
Dear @adel0049 , you can use (ASN1C) to generate the CAM structure's C++ compatible C code from the ASN1 standard. However the source is also available in many places, so you can just download it directly (like this one). What I did, is that I built a static library ( .a format on windows) from the source files, that can be linked to the OMNeT++ project. Once it is done, you can just include the CAM.h in your application code. You can fill, encode and decode the created CAM structure. |
Dear all,
I'm trying to build artery in Windows 10 (unfortunatley I need Windows). I built and installed every required libraries and software. When I'm running cmake .. in the build directory, I get the following errors:
Did anybody install artery on Windows successfully? Can somebody help with this?
The text was updated successfully, but these errors were encountered: