I have an issue with custom libraries that reference headers in other custom libraries.
This is under Eclipse (2020-03) with the 9.11 CDT - Arduino Plugins. (So the latest Version at this time)
Until now I have been using Sloeber, which included all include paths from all custom libraries so everythig was building fine. But with the latest Eclipse/CDT version sloeber does not work anymore. Since it should not really be necessary anymore anyway, I am trying to get my projects to build with just the CDT-Arduino Plugin(s).
Simple projects are Building fine, but more complex have issues…
The build process (makefile generation) seems to scan though the libraries directory, because I can see all the entries in the makefile to build the things it findes in there. But I one of these libraries referneces a header of another one it failes to find it.
Of course I could just edit the include and make it relative, but since these are imported libraries I prefer not to make changes in the code.
With the Arduino plugin there is also no configuration for include paths etc in the project properties any more :-( Normally I would have added the include path there, but the arduino Plugin seems to want to do it all by itself...
Any idea, how I can add the required include paths? Is there a way to do this in the makefile.ini?
I found a different and better way to add custom libraries, which also solves the include path problem. Sadly there is no documentaion about this.
Anyway, it is very simple, just copy the library to the …/.arduinocdt/libraries/ folder. The custom library must contain a library.jason and library.properties file. When opening the Arduino download manager, you can see the new lib and ýou can add it.
Eclipse will read the library version from the json/properties and copy the data in the correct form in a sub directory with the version number.
For libraries added this way, all includes/headers will be found without Problems including libraries referencing other libraries.
Related
I have an application where I am compiling against the libraries Qt (5.11) and Paraview (5.6). I am developing in C++ and I am also coding in Eclipse.
Yes, I am aware that Paraview is more of an application then a library. However, I will be using it as a library to intergrate Paraview in my application in order to visualize some data. I could use QT creator and that would take care of the QT library. However, I am not very familiar with QT creator as an IDE and my entire workflow has been based around eclipse. For this, I would like to utilize components that I am familiar with.
Now, as many of you know, the QT library is large, very large. So is Paraview. I could create a search path for each folder containing the header files I need in the C++ include settings. If I go in this direction, then my library search path will be long, very long. Also, there might be a situation where I need to change computers (for whatever reason), I would then need to edit the search paths for the new computer and that will take time.I had thought that I could only include the paths to the include files that I needed. And I have started to do that. However, when I compile my application, I noticed that the libraries would use include files that are located in a directory that Eclipse is not even searching in. At the end of the day, I would need to include all of the directories to which the header files are stored in.
I was wondering if there might be a better method to reference all of the files that my project needs? Both come with make files (I believe that Qt includes a QMake file?). Is there a simpler method to include all of the files instead of editing the C++ include path for eclipse? I did see an option for QMake in Eclipse, is there anything that I can do with that feature to compile to QT library with my application?
Thank you and any feedback is appreciated. Also, feel free to suggest edits or clarification so that my description is clear.
Note some good practices when working with eclipse and qt or other large library.
Include the directories with a headers
If you use option "Generate Makefile automatically", add group include path with GUI Eclipse:
copy the directories with a headers (via system explorer);
go to project properties/C/C++Build/Settings/Tool Settings/GCC C++
Compiler/Includes.
paste directories (Ctrl+V) into "Include paths" list.
If you are using a qt-eclipse-integration plugin, you can add folders via the Discovered scanner configuration. You can find it if open: project properties/C/C++ Include Paths and Symbols/Add Contributed...
Create new project
if you already have a customized project, you can use it as a template:
make a copy of the customized project and name "mytempate";
delete all files from "mytempate".
When you want to create a new project, copy the template and specify (in the copy dialog) the path where the source codes are located.
Now you have a finished new project with settings and code.
Change computers
Use one of the ways:
1. Export C/C++ project settings for use in another workspace via project properties/C/C++ General/Paths and symbols/Includes/Export Settings...
2. Pack in the archive eclipse and workspace and copy your another computer.
Using netbeans 7.2 and the most recent version of MinGW (using installer) I can't use the libnoise library. I am properly including the header files, the auto completion confirms this, however the library is simply not working. There is a .lib file and a .dll. I have tried every possible combination of adding them under project > properties > Build > Linker as well as putting the .dll in the base project directory. Despite all this I am still getting undefined reference errors whenever I try and run the code. Does anyone know what to do?
I know that it is possible to link import library files (*.lib) with MinGW, but I still suggest to recompile libnoise.
With the current Makefile, this is not very easy and may break. Thus I've written a CMake script, which doesn't only work on Windows, but should work on all CMake supported platforms. In addition to this, I've cleaned up the directory structure. The code itself hasn't been touched and when you build a library it should essentially be the same as the current one.
libnoise on GitHub
After you've built your shared library, you'll have a libnoise.dll.a and libnoise.dll file. You then add libnoise.dll.a to the linking settings and put the DLL itself next to the binary, or in the working directory.
You have to link the lib file (= adding it to linker libraries) and put the dll to
<project root>/dist/<...>/
(where your exe is put to). There's no need to add the dll to linker too.
Can you please post the error message you get?
I encountered a problem with linking and distributing libraries in Visual Studio 2010. What I want to do:
Link my project with SDL and use its headers
Keep SDL somewhere in my project directory, so when I clone it from source control on another machine, I won't have to install any additional library.
My attempts so far:
Google points to tutorials like this one : Lazy Foo Productions, but it doesn't do what I want because SDL is placed outside project directory.
When I copy SDL to solution directory and add it via Project->Properties->Linker.... It works, but project settings now contain absolute path to library directory, so I guess it won't compile on other machine.
What should I do? SDL is just a example here, I'd like to know how to solve such problem with any library.
You can use Project->Properties->Linker->General->Additional Library Directories to specify new directory where the linker to look up for libraries (it can use relative path)
You can use Project->Properties->C++->General->Additional Include Directories to specify where to look for the headers.
I have been using libcurl in a C++ project.
I have added the libcurl Include and Library directory's to the VC++ Directories and added the .lib file to the Linkers Input Additional Dependencies.
Everything works fine but when I check in my code (TFS) and somebody gets it on another machine they cannot build etc due to not having libcurl installed on their machine or installed in different paths etc.
How do I add all the necessary files to my C++ Solution so that anyone getting the project from source control can build and link without error.
Thanks
The solution I went with was too add the library file to a folder within my solution, then add
$(SolutionDir)libs\curllib.lib
To the Additional Dependencies within Linker->Input
You can not add the library to the solution. you should distribute the library along the source code. Place it in a subdirectory of the project so that it goes along the whole project source. Then configure the path in Additional Library Dependencies. That way the people should be able to link appropriately when building from source.
We have a large project using VS2008 and boost 1_42. I'm trying to upgrade to VS2010 and boost 1_44. I installed VS2010 and boost 1_44 and converted the project. Now I am trying to build, and everything compiles, but fails when linking:
LINK : fatal error LNK1104: cannot open file 'libboost_thread-vc90-mt-1_42.lib'
I have changed the include and lib directories to point to the new boost 1_44 files and I have renamed the old boost 1_42 directory.
Why is the linker still looking for a vc90-1_42 file, when it is only using 1_44 headers? Is there a way that I can determine WHY the linker wants this file? The linker obviously thinks it needs the file, but why?
I have cleaned the project and I am re-building to ensure any old build files are erased.
I've run into exactly this problem a couple of times too. It's usually been some old temporary files but like in your case cleaning didn't always do the trick straight away. Does your project include any static libs that might have been built with 1.42?
Something you can try which may or may not be helpful in tracking down your issue:
Rename the old boost directory back to it's original name
Clean the solution
Under C/C++->Command Line->Additional Options add "/showIncludes"
Under Linker->Command Line->Additional Options add "/verbose:lib"
Rebuild all
Then when you build you'll be able to see at which point 1.42 headers are included, etc. in the output window. Somehow doing this helped me in tracking down where the problem was.
Along with changing the lib directory, you need to change the name of the boost library. That's in the Linker | Input section of the project settings.
Your added comment makes it clear that the dependency on the Boost 1.42 library was being created indirectly by another library that hadn't been rebuilt.
For this you basically have two choices: either add that library as a project to your main solution, and make sure it has enough dependency information that it'll be re-built when you upgrade Boost, or use the /Zl compiler switch when you build your library. This tells the compiler you're building a library so you do not want to embed library dependencies like this.
Boost uses
#pragma comment(lib)
command to inform the linker of libraries it needs to link with. It is not an error. If Boost says you need it, it's likely you do.
On How can I find out why the linker wants this file?
There are programs which will go through your app and dlls/libs and report the content of manifests and what the binaries report they depend on. You could then scan the report for the unexpected libraries being included. We used this mainly to find libs including the previous version of the VC runtime.
Have not used the one we had in about 5 years though, now if only I could remember the name of the app!
DependancyWalker (depends.exe) will allow you to see dependancies of dll/exe but not static libs.
You could open each binary as a 'file' in MSVS and look at the manifest content by hand, but I imaging this would be a bit painful. I've not tried this with a static lib.