docs/CODING_STYLE.md

   1 ---
   2 title: Coding Style
   3 ---
   4
   5 # Coding Style
   6
   7 ## Formatting
   8
   9 - 8ch indent, no tabs, except for files in `man/` which are 2ch indent, and
  10   still no tabs, and shell scripts, which are 4ch indent, and no tabs either.
  11
  12 - We prefer `/* comments */` over `// comments` in code you commit,
  13   please. This way `// comments` are left for developers to use for local,
  14   temporary commenting of code for debug purposes (i.e. uncommittable stuff),
  15   making such comments easily discernible from explanatory, documenting code
  16   comments (i.e. committable stuff).
  17
  18 - Don't break code lines too eagerly. We do **not** force line breaks at 80ch,
  19   all of today's screens should be much larger than that. But then again, don't
  20   overdo it, ~109ch should be enough really. The `.editorconfig`, `.vimrc` and
  21   `.dir-locals.el` files contained in the repository will set this limit up for
  22   you automatically, if you let them (as well as a few other things). Please
  23   note that emacs loads `.dir-locals.el` automatically, but vim needs to be
  24   configured to load `.vimrc`, see that file for instructions.
  25
  26 - Try to write this:
  27
  28   ```c
  29   void foo() {
  30   }
  31   ```
  32
  33   instead of this:
  34
  35   ```c
  36   void foo()
  37   {
  38   }
  39   ```
  40
  41 - Single-line `if` blocks should not be enclosed in `{}`. Write this:
  42
  43   ```c
  44   if (foobar)
  45           waldo();
  46   ```
  47
  48   instead of this:
  49
  50   ```c
  51   if (foobar) {
  52           waldo();
  53   }
  54   ```
  55
  56 - Do not write `foo ()`, write `foo()`.
  57
  58 - Preferably allocate local variables on the top of the block:
  59
  60   ```c
  61   {
  62           int a, b;
  63
  64           a = 5;
  65           b = a;
  66   }
  67   ```
  68
  69 ## Other
  70
  71 - structs in `PascalCase` (with exceptions, such as public API structs),
  72   variables and functions in `snake_case`.
  73
  74 - To minimize strict aliasing violations, we prefer unions over casting.
  75
  76 - Be exceptionally careful when formatting and parsing floating point
  77   numbers. Their syntax is locale dependent (i.e. `5.000` in en_US is
  78   generally understood as 5, while in de_DE as 5000.).
  79
  80 - Do not mix function invocations with variable definitions in one
  81   line. Wrong:
  82
  83   ```c
  84   {
  85           int a = foobar();
  86           uint64_t x = 7;
  87   }
  88   ```
  89
  90   Right:
  91
  92   ```c
  93   {
  94           int a;
  95           uint64_t x = 7;
  96
  97           a = foobar();
  98   }
  99   ```
 100
 101 - Use `goto` for cleaning up, and only use it for that. i.e. you may
 102   only jump to the end of a function, and little else. Never jump
 103   backwards!
 104
 105 - Avoid static variables, except for caches and very few other
 106   cases. Think about thread-safety! While most of our code is never
 107   used in threaded environments, at least the library code should make
 108   sure it works correctly in them. Instead of doing a lot of locking
 109   for that, we tend to prefer using TLS to do per-thread caching (which
 110   only works for small, fixed-size cache objects), or we disable
 111   caching for any thread that is not the main thread. Use
 112   `is_main_thread()` to detect whether the calling thread is the main
 113   thread.
 114
 115 - Do not write functions that clobber call-by-reference variables on
 116   failure. Use temporary variables for these cases and change the
 117   passed in variables only on success.
 118
 119
 120 - Instead of using `memzero()`/`memset()` to initialize structs allocated
 121   on the stack, please try to use c99 structure initializers. It's
 122   short, prettier and actually even faster at execution. Hence:
 123
 124   ```c
 125   struct foobar t = {
 126           .foo = 7,
 127           .bar = "bazz",
 128   };
 129   ```
 130
 131   instead of:
 132
 133   ```c
 134   struct foobar t;
 135   zero(t);
 136   t.foo = 7;
 137   t.bar = "bazz";
 138   ```
 139
 140 - The order in which header files are included doesn't matter too
 141   much. systemd-internal headers must not rely on an include order, so
 142   it is safe to include them in any order possible.
 143   However, to not clutter global includes, and to make sure internal
 144   definitions will not affect global headers, please always include the
 145   headers of external components first (these are all headers enclosed
 146   in <>), followed by our own exported headers (usually everything
 147   that's prefixed by `sd-`), and then followed by internal headers.
 148   Furthermore, in all three groups, order all includes alphabetically
 149   so duplicate includes can easily be detected.
 150
 151 - To implement an endless loop, use `for (;;)` rather than `while (1)`.
 152   The latter is a bit ugly anyway, since you probably really
 153   meant `while (true)`. To avoid the discussion what the right
 154   always-true expression for an infinite while loop is, our
 155   recommendation is to simply write it without any such expression by
 156   using `for (;;)`.
 157
 158 - Avoid leaving long-running child processes around, i.e. `fork()`s that
 159   are not followed quickly by an `execv()` in the child. Resource
 160   management is unclear in this case, and memory CoW will result in
 161   unexpected penalties in the parent much, much later on.
 162
 163 - Don't block execution for arbitrary amounts of time using `usleep()`
 164   or a similar call, unless you really know what you do. Just "giving
 165   something some time", or so is a lazy excuse. Always wait for the
 166   proper event, instead of doing time-based poll loops.
 167
 168 - To determine the length of a constant string `"foo"`, don't bother with
 169   `sizeof("foo")-1`, please use `strlen()` instead (both gcc and clang optimize
 170   the call away for fixed strings). The only exception is when declaring an
 171   array. In that case use STRLEN, which evaluates to a static constant and
 172   doesn't force the compiler to create a VLA.
 173
 174 - Please avoid using global variables as much as you can. And if you
 175   do use them make sure they are static at least, instead of
 176   exported. Especially in library-like code it is important to avoid
 177   global variables. Why are global variables bad? They usually hinder
 178   generic reusability of code (since they break in threaded programs,
 179   and usually would require locking there), and as the code using them
 180   has side-effects make programs non-transparent. That said, there are
 181   many cases where they explicitly make a lot of sense, and are OK to
 182   use. For example, the log level and target in `log.c` is stored in a
 183   global variable, and that's OK and probably expected by most. Also
 184   in many cases we cache data in global variables. If you add more
 185   caches like this, please be careful however, and think about
 186   threading. Only use static variables if you are sure that
 187   thread-safety doesn't matter in your case. Alternatively, consider
 188   using TLS, which is pretty easy to use with gcc's `thread_local`
 189   concept. It's also OK to store data that is inherently global in
 190   global variables, for example data parsed from command lines, see
 191   below.
 192
 193 - Make sure to enforce limits on every user controllable resource. If the user
 194   can allocate resources in your code, your code must enforce some form of
 195   limits after which it will refuse operation. It's fine if it is hard-coded (at
 196   least initially), but it needs to be there. This is particularly important
 197   for objects that unprivileged users may allocate, but also matters for
 198   everything else any user may allocated.
 199
 200 - You might wonder what kind of common code belongs in `src/shared/` and what
 201   belongs in `src/basic/`. The split is like this: anything that is used to
 202   implement the public shared object we provide (sd-bus, sd-login, sd-id128,
 203   nss-systemd, nss-mymachines, nss-resolve, nss-myhostname, pam_systemd), must
 204   be located in `src/basic` (those objects are not allowed to link to
 205   libsystemd-shared.so). Conversely, anything which is shared between multiple
 206   components and does not need to be in `src/basic/`, should be in
 207   `src/shared/`.
 208
 209   To summarize:
 210
 211   `src/basic/`
 212   - may be used by all code in the tree
 213   - may not use any code outside of `src/basic/`
 214
 215   `src/libsystemd/`
 216   - may be used by all code in the tree, except for code in `src/basic/`
 217   - may not use any code outside of `src/basic/`, `src/libsystemd/`
 218
 219   `src/shared/`
 220   - may be used by all code in the tree, except for code in `src/basic/`,
 221     `src/libsystemd/`, `src/nss-*`, `src/login/pam_systemd.*`, and files under
 222     `src/journal/` that end up in `libjournal-client.a` convenience library.
 223   - may not use any code outside of `src/basic/`, `src/libsystemd/`, `src/shared/`
 224
 225 - Our focus is on the GNU libc (glibc), not any other libcs. If other libcs are
 226   incompatible with glibc it's on them. However, if there are equivalent POSIX
 227   and Linux/GNU-specific APIs, we generally prefer the POSIX APIs. If there
 228   aren't, we are happy to use GNU or Linux APIs, and expect non-GNU
 229   implementations of libc to catch up with glibc.
 230
 231 - Whenever installing a signal handler, make sure to set `SA_RESTART` for it, so
 232   that interrupted system calls are automatically restarted, and we minimize
 233   hassles with handling `EINTR` (in particular as `EINTR` handling is pretty broken
 234   on Linux).
 235
 236 - When applying C-style unescaping as well as specifier expansion on the same
 237   string, always apply the C-style unescaping fist, followed by the specifier
 238   expansion. When doing the reverse, make sure to escape `%` in specifier-style
 239   first (i.e. `%` → `%%`), and then do C-style escaping where necessary.
 240
 241 ## Destructors
 242
 243 - The destructors always deregister the object from the next bigger object, not
 244   the other way around.
 245
 246 - For robustness reasons, destructors should be able to destruct
 247   half-initialized objects, too.
 248
 249 - When you define a destructor or `unref()` call for an object, please accept a
 250   `NULL` object and simply treat this as NOP. This is similar to how libc
 251   `free()` works, which accepts `NULL` pointers and becomes a NOP for them. By
 252   following this scheme a lot of `if` checks can be removed before invoking
 253   your destructor, which makes the code substantially more readable and robust.
 254
 255 - Related to this: when you define a destructor or `unref()` call for an
 256   object, please make it return the same type it takes and always return `NULL`
 257   from it. This allows writing code like this:
 258
 259   ```c
 260   p = foobar_unref(p);
 261   ```
 262
 263   which will always work regardless if `p` is initialized or not,x and
 264   guarantees that `p` is `NULL` afterwards, all in just one line.
 265
 266 ## Error Handling
 267
 268 - Error codes are returned as negative `Exxx`. e.g. `return -EINVAL`. There are
 269   some exceptions: for constructors, it is OK to return `NULL` on OOM. For
 270   lookup functions, `NULL` is fine too for "not found".
 271
 272   Be strict with this. When you write a function that can fail due to more than
 273   one cause, it *really* should have an `int` as the return value for the error
 274   code.
 275
 276 - Do not bother with error checking whether writing to stdout/stderr worked.
 277
 278 - Do not log errors from "library" code, only do so from "main program"
 279   code. (With one exception: it is OK to log with DEBUG level from any code,
 280   with the exception of maybe inner loops).
 281
 282 - In public API calls, you **must** validate all your input arguments for
 283   programming error with `assert_return()` and return a sensible return
 284   code. In all other calls, it is recommended to check for programming errors
 285   with a more brutal `assert()`. We are more forgiving to public users than for
 286   ourselves! Note that `assert()` and `assert_return()` really only should be
 287   used for detecting programming errors, not for runtime errors. `assert()` and
 288   `assert_return()` by usage of `_likely_()` inform the compiler that he should
 289   not expect these checks to fail, and they inform fellow programmers about the
 290   expected validity and range of parameters.
 291
 292 - When you invoke certain calls like `unlink()`, or `mkdir_p()` and you know it
 293   is safe to ignore the error it might return (because a later call would
 294   detect the failure anyway, or because the error is in an error path and you
 295   thus couldn't do anything about it anyway), then make this clear by casting
 296   the invocation explicitly to `(void)`. Code checks like Coverity understand
 297   that, and will not complain about ignored error codes. Hence, please use
 298   this:
 299
 300   ```c
 301   (void) unlink("/foo/bar/baz");
 302   ```
 303
 304   instead of just this:
 305
 306   ```c
 307   unlink("/foo/bar/baz");
 308   ```
 309
 310   Don't cast function calls to `(void)` that return no error
 311   conditions. Specifically, the various `xyz_unref()` calls that return a
 312   `NULL` object shouldn't be cast to `(void)`, since not using the return value
 313   does not hide any errors.
 314
 315 - When returning a return code from `main()`, please preferably use
 316   `EXIT_FAILURE` and `EXIT_SUCCESS` as defined by libc.
 317
 318 ## Logging
 319
 320 - For every function you add, think about whether it is a "logging" function or
 321   a "non-logging" function. "Logging" functions do logging on their own,
 322   "non-logging" function never log on their own and expect their callers to
 323   log. All functions in "library" code, i.e. in `src/shared/` and suchlike must
 324   be "non-logging". Every time a "logging" function calls a "non-logging"
 325   function, it should log about the resulting errors. If a "logging" function
 326   calls another "logging" function, then it should not generate log messages,
 327   so that log messages are not generated twice for the same errors.
 328
 329 - If possible, do a combined log & return operation:
 330
 331   ```c
 332   r = operation(...);
 333   if (r < 0)
 334           return log_(error|warning|notice|...)_errno(r, "Failed to ...: %m");
 335   ```
 336
 337   If the error value is "synthetic", i.e. it was not received from
 338   the called function, use `SYNTHETIC_ERRNO` wrapper to tell the logging
 339   system to not log the errno value, but still return it:
 340
 341   ```c
 342   n = read(..., s, sizeof s);
 343   if (n != sizeof s)
 344           return log_error_errno(SYNTHETIC_ERRNO(EIO), "Failed to read ...");
 345   ```
 346
 347 ## Memory Allocation
 348
 349 - Always check OOM. There is no excuse. In program code, you can use
 350   `log_oom()` for then printing a short message, but not in "library" code.
 351
 352 - Avoid fixed-size string buffers, unless you really know the maximum size and
 353   that maximum size is small. They are a source of errors, since they possibly
 354   result in truncated strings. It is often nicer to use dynamic memory,
 355   `alloca()` or VLAs. If you do allocate fixed-size strings on the stack, then
 356   it is probably only OK if you either use a maximum size such as `LINE_MAX`,
 357   or count in detail the maximum size a string can have. (`DECIMAL_STR_MAX` and
 358   `DECIMAL_STR_WIDTH` macros are your friends for this!)
 359
 360   Or in other words, if you use `char buf[256]` then you are likely doing
 361   something wrong!
 362
 363 - Make use of `_cleanup_free_` and friends. It makes your code much nicer to
 364   read (and shorter)!
 365
 366 - Use `alloca()`, but never forget that it is not OK to invoke `alloca()`
 367   within a loop or within function call parameters. `alloca()` memory is
 368   released at the end of a function, and not at the end of a `{}` block. Thus,
 369   if you invoke it in a loop, you keep increasing the stack pointer without
 370   ever releasing memory again. (VLAs have better behavior in this case, so
 371   consider using them as an alternative.)  Regarding not using `alloca()`
 372   within function parameters, see the BUGS section of the `alloca(3)` man page.
 373
 374 - If you want to concatenate two or more strings, consider using `strjoina()`
 375   or `strjoin()` rather than `asprintf()`, as the latter is a lot slower. This
 376   matters particularly in inner loops (but note that `strjoina()` cannot be
 377   used there).
 378
 379 ## Types
 380
 381 - Think about the types you use. If a value cannot sensibly be negative, do not
 382   use `int`, but use `unsigned`.
 383
 384 - Use `char` only for actual characters. Use `uint8_t` or `int8_t` when you
 385   actually mean a byte-sized signed or unsigned integers. When referring to a
 386   generic byte, we generally prefer the unsigned variant `uint8_t`. Do not use
 387   types based on `short`. They *never* make sense. Use `int`, `long`, `long
 388   long`, all in unsigned and signed fashion, and the fixed-size types
 389   `uint8_t`, `uint16_t`, `uint32_t`, `uint64_t`, `int8_t`, `int16_t`, `int32_t`
 390   and so on, as well as `size_t`, but nothing else. Do not use kernel types
 391   like `u32` and so on, leave that to the kernel.
 392
 393 - Stay uniform. For example, always use `usec_t` for time values. Do not mix
 394   `usec` and `msec`, and `usec` and whatnot.
 395
 396 - Never use the `off_t` type, and particularly avoid it in public APIs. It's
 397   really weirdly defined, as it usually is 64-bit and we don't support it any
 398   other way, but it could in theory also be 32-bit. Which one it is depends on
 399   a compiler switch chosen by the compiled program, which hence corrupts APIs
 400   using it unless they can also follow the program's choice. Moreover, in
 401   systemd we should parse values the same way on all architectures and cannot
 402   expose `off_t` values over D-Bus. To avoid any confusion regarding conversion
 403   and ABIs, always use simply `uint64_t` directly.
 404
 405 - Unless you allocate an array, `double` is always a better choice than
 406   `float`. Processors speak `double` natively anyway, so there is no speed
 407   benefit, and on calls like `printf()` `float`s get promoted to `double`s
 408   anyway, so there is no point.
 409
 410 - Use the bool type for booleans, not integers. One exception: in public
 411   headers (i.e those in `src/systemd/sd-*.h`) use integers after all, as `bool`
 412   is C99 and in our public APIs we try to stick to C89 (with a few extension).
 413
 414 ## Deadlocks
 415
 416 - Do not issue NSS requests (that includes user name and host name lookups)
 417   from PID 1 as this might trigger deadlocks when those lookups involve
 418   synchronously talking to services that we would need to start up.
 419
 420 - Do not synchronously talk to any other service from PID 1, due to risk of
 421   deadlocks.
 422
 423 ## File Descriptors
 424
 425 - When you allocate a file descriptor, it should be made `O_CLOEXEC` right from
 426   the beginning, as none of our files should leak to forked binaries by
 427   default. Hence, whenever you open a file, `O_CLOEXEC` must be specified,
 428   right from the beginning. This also applies to sockets. Effectively, this
 429   means that all invocations to:
 430
 431   - `open()` must get `O_CLOEXEC` passed,
 432   - `socket()` and `socketpair()` must get `SOCK_CLOEXEC` passed,
 433   - `recvmsg()` must get `MSG_CMSG_CLOEXEC` set,
 434   - `F_DUPFD_CLOEXEC` should be used instead of `F_DUPFD`, and so on,
 435   - invocations of `fopen()` should take `e`.
 436
 437 - It's a good idea to use `O_NONBLOCK` when opening 'foreign' regular files,
 438   i.e.  file system objects that are supposed to be regular files whose paths
 439   where specified by the user and hence might actually refer to other types of
 440   file system objects. This is a good idea so that we don't end up blocking on
 441   'strange' file nodes, for example if the user pointed us to a FIFO or device
 442   node which may block when opening. Moreover even for actual regular files
 443   `O_NONBLOCK` has a benefit: it bypasses any mandatory lock that might be in
 444   effect on the regular file. If in doubt consider turning off `O_NONBLOCK`
 445   again after opening.
 446
 447 ## Command Line
 448
 449 - If you parse a command line, and want to store the parsed parameters in
 450   global variables, please consider prefixing their names with `arg_`. We have
 451   been following this naming rule in most of our tools, and we should continue
 452   to do so, as it makes it easy to identify command line parameter variables,
 453   and makes it clear why it is OK that they are global variables.
 454
 455 - Command line option parsing:
 456   - Do not print full `help()` on error, be specific about the error.
 457   - Do not print messages to stdout on error.
 458   - Do not POSIX_ME_HARDER unless necessary, i.e. avoid `+` in option string.
 459
 460 ## Exporting Symbols
 461
 462 - Variables and functions **must** be static, unless they have a prototype, and
 463   are supposed to be exported.
 464
 465 - Public API calls (i.e. functions exported by our shared libraries)
 466   must be marked `_public_` and need to be prefixed with `sd_`. No
 467   other functions should be prefixed like that.
 468
 469 - When exposing public C APIs, be careful what function parameters you make
 470   `const`. For example, a parameter taking a context object should probably not
 471   be `const`, even if you are writing an otherwise read-only accessor function
 472   for it. The reason is that making it `const` fixates the contract that your
 473   call won't alter the object ever, as part of the API. However, that's often
 474   quite a promise, given that this even prohibits object-internal caching or
 475   lazy initialization of object variables. Moreover, it's usually not too
 476   useful for client applications. Hence, please be careful and avoid `const` on
 477   object parameters, unless you are very sure `const` is appropriate.
 478
 479 ## Referencing Concepts
 480
 481 - When referring to a configuration file option in the documentation and such,
 482   please always suffix it with `=`, to indicate that it is a configuration file
 483   setting.
 484
 485 - When referring to a command line option in the documentation and such, please
 486   always prefix with `--` or `-` (as appropriate), to indicate that it is a
 487   command line option.
 488
 489 - When referring to a file system path that is a directory, please always
 490   suffix it with `/`, to indicate that it is a directory, not a regular file
 491   (or other file system object).
 492
 493 ## Functions to Avoid
 494
 495 - Use `memzero()` or even better `zero()` instead of `memset(..., 0, ...)`
 496
 497 - Please use `streq()` and `strneq()` instead of `strcmp()`, `strncmp()` where
 498   applicable (i.e. wherever you just care about equality/inequality, not about
 499   the sorting order).
 500
 501 - Never use `strtol()`, `atoi()` and similar calls. Use `safe_atoli()`,
 502   `safe_atou32()` and suchlike instead. They are much nicer to use in most
 503   cases and correctly check for parsing errors.
 504
 505 - `htonl()`/`ntohl()` and `htons()`/`ntohs()` are weird. Please use `htobe32()`
 506   and `htobe16()` instead, it's much more descriptive, and actually says what
 507   really is happening, after all `htonl()` and `htons()` don't operate on
 508   `long`s and `short`s as their name would suggest, but on `uint32_t` and
 509   `uint16_t`. Also, "network byte order" is just a weird name for "big endian",
 510   hence we might want to call it "big endian" right-away.
 511
 512 - Please never use `dup()`. Use `fcntl(fd, F_DUPFD_CLOEXEC, 3)` instead. For
 513   two reason: first, you want `O_CLOEXEC` set on the new `fd` (see
 514   above). Second, `dup()` will happily duplicate your `fd` as 0, 1, 2,
 515   i.e. stdin, stdout, stderr, should those `fd`s be closed. Given the special
 516   semantics of those `fd`s, it's probably a good idea to avoid
 517   them. `F_DUPFD_CLOEXEC` with `3` as parameter avoids them.
 518
 519 - Don't use `fgets()`, it's too hard to properly handle errors such as overly
 520   long lines. Use `read_line()` instead, which is our own function that handles
 521   this much nicer.
 522
 523 - Don't invoke `exit()`, ever. It is not replacement for proper error
 524   handling. Please escalate errors up your call chain, and use normal `return`
 525   to exit from the main function of a process. If you `fork()`ed off a child
 526   process, please use `_exit()` instead of `exit()`, so that the exit handlers
 527   are not run.
 528
 529 - We never use the POSIX version of `basename()` (which glibc defines it in
 530   `libgen.h`), only the GNU version (which glibc defines in `string.h`).  The
 531   only reason to include `libgen.h` is because `dirname()` is needed. Every
 532   time you need that please immediately undefine `basename()`, and add a
 533   comment about it, so that no code ever ends up using the POSIX version!
 534
 535 # Committing to git
 536
 537 - Commit message subject lines should be prefixed with an appropriate component
 538   name of some kind. For example "journal: ", "nspawn: " and so on.
 539
 540 - Do not use "Signed-Off-By:" in your commit messages. That's a kernel thing we
 541   don't do in the systemd project.