8

First I want to state for the record that this question is related to school/homework.

Let’s say computers CP1 and CP2 both share the same operating system and machine language. If a C program is compiled on CP1, in order to move it to CP2, is it necessary to transfer the source code and recompile on CP2, or simply transfer the object files.

My gut answer is that the object files should suffice. The C code is translated into assembly by the compiler and assembled into machine code by the assembler. Because the architecture shares the same machine code and operating system, I don't see a problem.

But the more I think about it, the more confused I’m starting to get.

My questions are:

a) Since its referring to object files and not executables, I’m assuming there has been no linking. Would there be any problems that surface when linking on CP2?

b) Would it matter if the code used C11 standard on CP1 but the only compiler on CP2 was C99? I'm assuming this is irrelevant once the code has been compiled/assembled.

c) The question doesn't specify shared/dynamic linked libraries. So this would only really work if the program had no dependencies on .dll/.so/ .dylib files, or else these would be required on CP2 as well.

I feel like there are so many gotchas, and considering how vague the question is I now feel that it would be safer to simply recompile.

Halp!

Improve this question
1
  • distcc, which copies object files between computers, recommends using the same version of compiler on all nodes, even for the C language. Commented Jan 13, 2013 at 6:10

4 Answers 4

Reset to default
3

The answer is, it depends. When you compile a C program and move the object files to link on a different computer, it should work. But because of factors such as endianness or name mangling, your program might not work as intended, and even might crash when you try to run it.

C11 is not supported by a C99 compiler, but it does not matter if the source has been compiled and assembled.

As long as the source is compiled with the libraries on one machine, you don't need the libraries to link or run the file(s) on the other computer (static libraries only, dynamic libraries will have to be on the computer you run the application on). This said, you should make the program independent so you don't run into the same problems as before where the program doesn't work as intended or crashes.

You could get a compiler that supports EABI so you don't run into these problems. Compilers that support the EABI create object code that is compatible with code generated by other such compilers, thus allowing developers to link libraries generated with one compiler with object code generated with a different compiler.

I have tried to do this before, but not a whole lot, and not recently. Therefore, my information may not be 100% accurate.

Improve this answer
Sign up to request clarification or add additional context in comments.

2 Comments

"You don't need the libraries" - that's only true for static linking. Dynamic libs are only referenced by your program, not compiled into it. The program won't start if they are missing.
Your rev 2 is substantially right, you should mention name mangling and runtime compatibility, there are systems where you can actually ask for different kinds of object file outputs, for when a new ABI is so much better that it's worth rewriting the toolchain to get it.
1

a) I've already heard the term "object files" being used to refer to linked binaries - even though it's kinda inaccurate. So maybe they mean "binaries". I'd say linking on a different machine could be problematic if it has a different compiler - unless object file formats are standardized, which I'm not sure about.

b) Using different standards or even compilers doesn't matter for binary code - if it's linked statically. If it relies on functions from a dynamic lib, there could be problems. Which answers c) as well: Yes, this will be a problem. The program won't start if it doesn't have all required dynamic libs in the correct version. Depends on linking mode (static vs. dynamic), again.

Improve this answer

Comments

1

Q: Let’s say computers CP1 and CP2 both share the same operating system and machine language.

A: Then you can run the same .exe's on both computers

Q: If a C program is compiled on CP1, in order to move it to CP2, is it necessary to transfer the source code

A: No. You only need the source code if you want to recompile. You only need to recompile if it's a different, incompatible CPU and/or OS.

"Object files" are generally not needed at all for program execution:

http://en.wikipedia.org/wiki/Object_files

An object file is a file containing relocatable format machine code that is usually not directly executable. Object files are produced by an assembler, compiler, or other language translator, and used as input to the linker.

An "executable program" might need one or more "shared libraries" (aka .dll's). In which case the same restrictions apply: the shared libraries, if not already resident, must be copied along with the .exe, and must also be compatible with the CPU and OS.

Finally, "scripts" do not need to be recompiled. You may copy the script freely from computer to computer. But each computer must have an "interpreter" to run the script: a Perl script needs a Perl interpreter, a Python script a python interpreter, and so on.

Improve this answer

1 Comment

The problem is that CPU's are different. We are starting to see Red Hat options to compile packages from source by default. Not only that, libraries may be different, appimage and flatpack being necessary. Minecraft and Arduino bring their own Java now. Even statically linking crypto is not safe of late. I need the originals to exist on the target despite static linking. You need to compile to a safe baseline level to distribute binaries. Chrome no longer runs on a Pentium M. The binaries on a Microsoft Surface are lean, they are not the same as as the binaries on your HP laptop.
1

What matters for object files are the ABIs supported by the target machine (C2) and required by the code. As long as C2 has all of the ABIs needed by the code and supports the same (or at least compatible) versions of those ABIs as C1, things should work. The problem comes if some ABI has a different, incompatible version, a problem that is sometimes referred to as "dll hell", even on machines that don't have dlls (which is a Windows term).

Improve this answer

Comments

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.