Compilation and execution path
مُقدَّم| English | المعنى |
|---|---|
| preprocessor/prɪˈprəʊsesə/ | preprocessor |
| compiler/kəmˈpaɪlə/ | Software that translates a program from a source language into another form, commonly machine or intermediate code. |
| assembler/əˈsemblə/ | Software that translates assembly-language instructions into machine code. |
| object file | object file |
| linker/ˈlɪŋkə/ | linker |
| loader | loader |
| separate compilation | separate compilation |
| bytecode/ˈbaɪtkəʊd/ | Intermediate instructions intended for execution by a virtual machine or interpreter rather than directly by a physical processor. |
It compiled, so why did it fail?
- A student builds a program and gets
undefined reference to 'area'. The code had no errors when it compiled, so which tool is complaining? - Turning source code into a running program takes four stages, and each stage has its own kind of error. Knowing the stages tells you where to look.
Stage 1: the preprocessor
- The preprocessor 预处理器 expands
#includeand#define. It works on plain text and producesmain.i. - Macros have no types: the preprocessor only swaps text, so a bad macro shows up as a strange error later.
The preprocessor checks that a macro is used with the right types.
الصعوبة 1/5 · 100 عملة + مكافأة صغيرة عند الإنجاز الأول اليوم
The preprocessor only swaps text. Types are checked later, by the compiler.
Stage 2: the compiler
- The compiler 编译器 checks types and turns C into assembly language,
main.s. - Type errors and syntax errors come from this stage.
Turning main.i into main.s is the job of —
الصعوبة 1/5 · 100 عملة + مكافأة صغيرة عند الإنجاز الأول اليوم
The preprocessor makes .i, the compiler makes .s, the assembler makes .o and the linker makes the program.
Stage 3: the assembler
- The assembler 汇编器 turns assembly into machine code in an object file 目标文件,
main.o. - A call to a function in another file stays unresolved: the object file just records the name it needs.
Stage 4: the linker, then the loader
- The linker 链接器 joins object files and libraries and matches each needed name to its definition. A name with no definition gives "undefined reference": a linker error, not a compiler error.
- The result is an executable. The loader 加载器 maps it into memory and starts it as a process.
Put the build stages in order.
الصعوبة 3/5 · 200 عملة + مكافأة صغيرة عند الإنجاز الأول اليوم
The file extensions name the stages.
An undefined-reference error comes from —
الصعوبة 1/5 · 100 عملة + مكافأة صغيرة عند الإنجاز الأول اليوم
The linker matches names across object files, so a missing definition is its kind of error.
Match each error to the stage that reports it.
الصعوبة 2/5 · 140 عملة + مكافأة صغيرة عند الإنجاز الأول اليوم
Each stage has its own error class, so the message tells you where to look.
Why keep the stages apart
- Separate compilation 分别编译: change one file in a million-line program and only that file is compiled again; then the linker joins everything.
- Libraries can ship as ready-made object files, and every error belongs to one stage: macro, type, instruction or symbol.
A project has 40 .c files. You edit one of them (not a header). How many files must be compiled again before linking?
الصعوبة 2/5 · 140 عملة + مكافأة صغيرة عند الإنجاز الأول اليوم
Separate compilation recompiles only the changed file; the linker then joins all 40 object files.
Compiled and interpreted
- C goes through all four stages before it runs. The result is fast but tied to one kind of machine.
- Python reads the source each run: CPython compiles it to bytecode 字节码, and a virtual machine runs that. The difference is when translation happens.
Compiled and interpreted execution differ mainly in —
الصعوبة 1/5 · 100 عملة + مكافأة صغيرة عند الإنجاز الأول اليوم
Compile once and run many times, or read and run through a virtual machine every time.
CPython runs Python source with no translation step at all.
الصعوبة 1/5 · 100 عملة + مكافأة صغيرة عند الإنجاز الأول اليوم
CPython first compiles the source to bytecode, then its virtual machine runs the bytecode.
Worked example: the four files
main.cbecomesmain.i(preprocess),main.s(compile),main.o(assemble), thenprog(link). Each file can be kept and inspected with the right compiler flag.- In the opening story,
areawas defined ingeom.c, but onlymain.owas linked. Linkmain.oandgeom.otogether and the error disappears.
Three common slips
- Calling every failure a compile error. "Undefined reference" comes from the linker.
- Forgetting that headers are text.
#includepastes the file in; macros have no types. - Assuming optimisation never changes float results. Some optimisations reorder float arithmetic, and rounding then changes.
Recap
- Preprocessor (text) to compiler (types) to assembler (machine code) to linker (symbols); then the loader starts it.
- Undefined reference means a linker problem: a missing object file or library.
- Compiled or interpreted is about when translation happens: once before running, or every run.