CODING_STYLE

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