How to build notepad++?

How does one run a standalone executable built on another Linux platform, so that all libraries and dependencies are taken from the build system and not the host?

• If I copy the executable from the build node, I need at least to include all library dependencies when I run it on the host node.  And then I modify LD_LIBRARY_PATH to ensure that the dependent libraries are seen.  However, library versions (between the build and host nodes) may be different in the case of duplicates, so I may need to do an LD_PRELOAD to ensure that the host sees the library copied from the build node instead.  Furthermore, glibc and ld-linux may be different, so this requires running the build node's version of ld-linux (where not only LD_PRELOAD includes glibc and system libraries, but also ld-linux is used to invoke the executable). I understand that this approach works (there may be tweaks or ideas I didn't mention, in which case comments are invited), and that proot or a type of chroot can be used to run the executable in an isolated way, but the challenge comes up when the build node's executable is to be run on the host, but at the same time descendant processes are to use the host's native environment.  If I use the LD_LIBRARY_PATH approach, then descendant processes inherit that variable, which can interfere with the normal functionality of the host's binaries.  Furthermore, using LD_PRELOAD makes this an even more likely situation, since LD_PRELOAD overrides the standard path.  Also, LD_PRELOAD may pose a security risk.  Finally, when using ld-linux to invoke the build node's binary and when also overriding the glibc for the desired process to be that of the build node, the problem becomes the most clear, because descendant processes are run using the build environment and with the ld-linux override, the native host binaries don't run. I don't know if there's a replacement environment variable I can pass to let new processes use a particular ld-linux.  In that case, the original executable that was intended to be built on the build node and run on the host, if its source is available, can be modified to (re)set LD_PRELOAD and LD_LIBRARY_PATH to some ORIGINAL_LD_LIBRARY_PATH value for any subsequent spawned processes.  It wouldn't be enough to cause descendant processes to be run transparently with the host system's ld-linux but it would be a start.  A static binary is not always possible to create. Furthermore, solving this well can help in the case that a third party product is distributed for a particular Linux platform (its build host) that differs, even slightly, to that of the native host system.  If software is released and is intended to run on a particular RedHat or Debian derivative, then the libraries and prerequisite environment components can be separately downloaded (updated to the latest stable/security patch) and then used to run the supplied executable.  But the question is broader than this use case.

• Answer:

  You could use static compiling. Or using interpreted languages (with the needed libraries installed locally for your app).  Go could be also a valid solution. Or you could use Docker containers to put with you app the entire filesystem it depends on, as long as Docker is available for the target systems.