tor/doc/HACKING/CodingStandards.md

Coding conventions for Tor

tl;dr:

• Run configure with --enable-fatal-warnings
• Document your functions
• Write unit tests
• Run make check before submitting a patch
• Run make distcheck if you have made changes to build system components
• Add a file in changes for your branch.

Patch checklist

If possible, send your patch as one of these (in descending order of preference)

• A git branch we can pull from
• Patches generated by git format-patch
• A unified diff

Did you remember...

• To build your code while configured with --enable-fatal-warnings?
• To run make check-docs to see whether all new options are on the manpage?
• To write unit tests, as possible?
• To run make test-full to test against all unit and integration tests (or make test-full-online if you have a working connection to the internet)?
• To test that the distribution will actually work via make distcheck?
• To base your code on the appropriate branch?
• To include a file in the changes directory as appropriate?

If you are submitting a major patch or new feature, or want to in the future...

• Set up Chutney and Stem, see doc/HACKING/WritingTests.md
• Run make test-full to test against all unit and integration tests.

If you have changed build system components:

• Please run make distcheck
• For example, if you have changed Makefiles, autoconf files, or anything else that affects the build system.

License issues

Tor is distributed under the license terms in the LICENSE -- in brief, the "3-clause BSD license". If you send us code to distribute with Tor, it needs to be code that we can distribute under those terms. Please don't send us patches unless you agree to allow this.

Some compatible licenses include:

• 3-clause BSD
• 2-clause BSD
• CC0 Public Domain Dedication

How we use Git branches

Each main development series (like 0.2.1, 0.2.2, etc) has its main work applied to a single branch. At most one series can be the development series at a time; all other series are maintenance series that get bug-fixes only. The development series is built in a git branch called "main"; the maintenance series are built in branches called "maint-0.2.0", "maint-0.2.1", and so on. We regularly merge the active maint branches forward.

For all series except the development series, we also have a "release" branch (as in "release-0.2.1"). The release series is based on the corresponding maintenance series, except that it deliberately lags the maint series for most of its patches, so that bugfix patches are not typically included in a maintenance release until they've been tested for a while in a development release. Occasionally, we'll merge an urgent bugfix into the release branch before it gets merged into maint, but that's rare.

If you're working on a bugfix for a bug that occurs in a particular version, base your bugfix branch on the "maint" branch for the first supported series that has that bug. (As of June 2013, we're supporting 0.2.3 and later.)

If you're working on a new feature, base it on the main branch. If you're working on a new feature and it will take a while to implement and/or you'd like to avoid the possibility of unrelated bugs in Tor while you're implementing your feature, consider branching off of the latest maint- branch. Never branch off a relase- branch. Don't branch off a tag either: they come from release branches. Doing so will likely produce a nightmare of merge conflicts in the ChangeLog when it comes time to merge your branch into Tor. Best advice: don't try to keep an independent branch forked for more than 6 months and expect it to merge cleanly. Try to merge pieces early and often.

How we log changes

When you do a commit that needs a ChangeLog entry, add a new file to the changes toplevel subdirectory. It should have the format of a one-entry changelog section from the current ChangeLog file, as in

o Major bugfixes (security):

- Fix a potential buffer overflow. Fixes bug 99999; bugfix on
  0.3.1.4-beta.

o Minor features (performance):

- Make tor faster. Closes ticket 88888.

To write a changes file, first categorize the change. Some common categories are: o Minor bugfixes (subheading): o Major bugfixes (subheading): o Minor features (subheading): o Major features (subheading): o Code simplifications and refactoring: o Testing: o Documentation:

The subheading is a particular area within Tor. See the ChangeLog for examples.

Then say what the change does. If it's a bugfix, mention what bug it fixes and when the bug was introduced. To find out which Git tag the change was introduced in, you can use git describe --contains <sha1 of commit>. If you don't know the commit, you can search the git diffs (-S) for the first instance of the feature (--reverse).

For example, for #30224, we wanted to know when the bridge-distribution-request feature was introduced into Tor:

$ git log -S bridge-distribution-request --reverse commit ebab521525
Author: Roger Dingledine <arma@torproject.org>
Date:   Sun Nov 13 02:39:16 2016 -0500

    Add new BridgeDistribution config option

$ git describe --contains ebab521525
tor-0.3.2.3-alpha~15^2~4

If you need to know all the Tor versions that contain a commit, use:

$ git tag --contains 9f2efd02a1 | sort -V
tor-0.2.5.16
tor-0.2.8.17
tor-0.2.9.14
tor-0.2.9.15
...
tor-0.3.0.13
tor-0.3.1.9
tor-0.3.1.10
...

If a bug was introduced before the oldest currently supported release series of Tor, and it's hard to track down where it was introduced, you may say "bugfix on all supported versions of Tor."

If at all possible, try to create the changes file in the same commit where you are making the change. Please give it a distinctive name that no other branch will use for the lifetime of your change. We usually use "ticketNNNNN" or "bugNNNNN", where NNNNN is the ticket number. To verify the format of the changes file, you can use make check-changes. This is run automatically as part of make check -- if it fails, we must fix it as soon as possible, so that our CI passes. These checks are implemented in scripts/maint/lintChanges.py.

Changes file style guide:

• Make everything terse.

• Write from the user's point of view: describe the user-visible changes right away.

• Mention configuration options by name. If they're rare or unusual, remind people what they're for.

• Describe changes in the present tense and in the imperative: not past.

• Every bugfix should have a sentence of the form "Fixes bug 1234; bugfix on 0.1.2.3-alpha", describing what bug was fixed and where it came from.

• "Relays", not "servers", "nodes", or "Tor relays".

When we go to make a release, we will concatenate all the entries in changes to make a draft changelog, and clear the directory. We'll then edit the draft changelog into a nice readable format.

What needs a changes file?

• A not-exhaustive list: Anything that might change user-visible behavior. Anything that changes internals, documentation, or the build system enough that somebody could notice. Big or interesting code rewrites. Anything about which somebody might plausibly wonder "when did that happen, and/or why did we do that" 6 months down the line.

What does not need a changes file?

• Bugfixes for code that hasn't shipped in any released version of Tor
• Any change to a file that is not distributed in the tarball. This includes:
  • Any change to our CI configuration that does not affect the distributed source.
  • Any change to developer-only tools, unless those tools are distributed in the tarball.
• Non-functional code movement.
• Identifier re-namings, comment edits, spelling fixes, and so on.

Why use changes files instead of Git commit messages?

• Git commit messages are written for developers, not users, and they are nigh-impossible to revise after the fact.

Why use changes files instead of entries in the ChangeLog?

• Having every single commit touch the ChangeLog file tended to create zillions of merge conflicts.

Whitespace and C conformance

Tor's C code is written in accordance with the C99 standard. Invoke make check-spaces from time to time, so it can tell you about deviations from our C whitespace style. Generally, we use:

• Unix-style line endings
• K&R-style indentation
• No space before newlines
• Never more than one blank line in a row
• Always spaces, never tabs
• No more than 79-columns per line.
• Two spaces per indent.
• A space between control keywords and their corresponding paren if (x), while (x), and switch (x), never if(x), while(x), or switch(x).
• A space between anything and an open brace.
• No space between a function name and an opening paren. puts(x), not puts (x).
• Function declarations at the start of the line.
• Use void foo(void) to declare a function with no arguments. Saying void foo() is C++ syntax.
• Use const for new APIs.
• Variables should be initialized when declared, rather than declared at the top of a scope.

If you use an editor that has plugins for editorconfig.org, the file .editorconfig will help you to conform this coding style.

We try hard to build without warnings everywhere. In particular, if you're using gcc, you should invoke the configure script with the option --enable-fatal-warnings. This will tell the compiler to make all warnings into errors.

Functions to use; functions not to use

We have some wrapper functions like tor_malloc, tor_free, tor_strdup, and tor_gettimeofday; use them instead of their generic equivalents. (They always succeed or exit.)

Specifically, Don't use malloc, realloc, calloc, free, or strdup. Use tor_malloc, tor_realloc, tor_calloc, tor_free, or tor_strdup.

Don't use tor_realloc(x, y\*z). Use tor_reallocarray(x, y, z) instead.;

You can get a full list of the compatibility functions that Tor provides by looking through src/lib/*/*.h. You can see the available containers in src/lib/containers/*.h. You should probably familiarize yourself with these modules before you write too much code, or else you'll wind up reinventing the wheel.

We don't use strcat or strcpy or sprintf of any of those notoriously broken old C functions. We also avoid strncat and strncpy. Use strlcat, strlcpy, or tor_snprintf/tor_asprintf instead.

We don't call memcmp() directly. Use fast_memeq(), fast_memneq(), tor_memeq(), or tor_memneq() for most purposes. If you really need a tristate return value, use tor_memcmp() or fast_memcmp().

Don't call assert() directly. For hard asserts, use tor_assert(). For soft asserts, use tor_assert_nonfatal() or BUG(). If you need to print debug information in assert error message, consider using tor_assertf() and tor_assertf_nonfatal(). If you are writing code that is too low-level to use the logging subsystem, use raw_assert().

Don't use toupper() and tolower() functions. Use TOR_TOUPPER and TOR_TOLOWER macros instead. Similarly, use TOR_ISALPHA, TOR_ISALNUM et. al. instead of isalpha(), isalnum(), etc.

When allocating new string to be added to a smartlist, use smartlist_add_asprintf() to do both at once.

Avoid calling BSD socket functions directly. Use portable wrappers to work with sockets and socket addresses. Also, sockets should be of type tor_socket_t.

Don't use any of these functions: they aren't portable. Use the version prefixed with tor_ instead: strtok_r, memmem, memstr, asprintf, localtime_r, gmtime_r, inet_aton, inet_ntop, inet_pton, getpass, ntohll, htonll. (This list is incomplete.)

What code can use what other code?

We're trying to simplify Tor's structure over time. In the long run, we want Tor to be structured as a set of modules with no circular dependencies.

This property is currently provided by the modules in src/lib, but not throughout the rest of Tor. In general, higher-level libraries may use lower-level libraries, but never the reverse.

To prevent new circular dependencies from landing, we have a tool that you can invoke with make check-includes, and which is run automatically as part of make check. This tool will verify that, for every source directory with a .may_include file, no local headers are included except those specifically permitted by the .may_include file. When editing one of these files, please make sure that you are not introducing any cycles into Tor's dependency graph.

Floating point math is hard

Floating point arithmetic as typically implemented by computers is very counterintuitive. Failure to adequately analyze floating point usage can result in surprising behavior and even security vulnerabilities!

General advice:

• Don't use floating point.
• If you must use floating point, document how the limits of floating point precision and calculation accuracy affect function outputs.
• Try to do as much as possible of your calculations using integers (possibly acting as fixed-point numbers) and convert to floating point for display.
• If you must send floating point numbers on the wire, serialize them in a platform-independent way. Tor avoids exchanging floating-point values, but when it does, it uses ASCII numerals, with a decimal point (".").
• Binary fractions behave very differently from decimal fractions. Make sure you understand how these differences affect your calculations.
• Every floating point arithmetic operation is an opportunity to lose precision, overflow, underflow, or otherwise produce undesired results. Addition and subtraction tend to be worse than multiplication and division (due to things like catastrophic cancellation). Try to arrange your calculations to minimize such effects.
• Changing the order of operations changes the results of many floating-point calculations. Be careful when you simplify calculations! If the order is significant, document it using a code comment.
• Comparing most floating point values for equality is unreliable. Avoid using ==, instead, use >= or <=. If you use an epsilon value, make sure it's appropriate for the ranges in question.
• Different environments (including compiler flags and per-thread state on a single platform!) can get different results from the same floating point calculations. This means you can't use floats in anything that needs to be deterministic, like consensus generation. This also makes reliable unit tests of floating-point outputs hard to write.

For additional useful advice (and a little bit of background), see What Every Programmer Should Know About Floating-Point Arithmetic.

A list of notable (and surprising) facts about floating point arithmetic is at Floating-point complexities. Most of that series of posts on floating point is helpful.

For more detailed (and math-intensive) background, see What Every Computer Scientist Should Know About Floating-Point Arithmetic.

Other C conventions

The a ? b : c trinary operator only goes inside other expressions; don't use it as a replacement for if. (You can ignore this inside macro definitions when necessary.)

Assignment operators shouldn't nest inside other expressions. (You can ignore this inside macro definitions when necessary.)

Binary data and wire formats

Use pointer to char when representing NUL-terminated string. To represent arbitrary binary data, use pointer to uint8_t. (Many older Tor APIs ignore this rule.)

Refrain from attempting to encode integers by casting their pointers to byte arrays. Use something like set_uint32()/get_uint32() instead and don't forget about endianness.

Try to never hand-write new code to parse or generate binary formats. Instead, use trunnel if at all possible. See

https://gitweb.torproject.org/trunnel.git/tree

for more information about trunnel.

For information on adding new trunnel code to Tor, see src/trunnel/README

Calling and naming conventions

Whenever possible, functions should return -1 on error and 0 on success.

For multi-word identifiers, use lowercase words combined with underscores. (e.g., multi_word_identifier). Use ALL_CAPS for macros and constants.

Typenames should end with _t.

Function names should be prefixed with a module name or object name. (In general, code to manipulate an object should be a module with the same name as the object, so it's hard to tell which convention is used.)

Functions that do things should have imperative-verb names (e.g. buffer_clear, buffer_resize); functions that return booleans should have predicate names (e.g. buffer_is_empty, buffer_needs_resizing).

If you find that you have four or more possible return code values, it's probably time to create an enum. If you find that you are passing three or more flags to a function, it's probably time to create a flags argument that takes a bitfield.

What To Optimize

Don't optimize anything if it's not in the critical path. Right now, the critical path seems to be AES, logging, and the network itself. Feel free to do your own profiling to determine otherwise.

Log conventions

FAQ - Log Levels

No error or warning messages should be expected during normal OR or OP operation.

If a library function is currently called such that failure always means ERR, then the library function should log WARN and let the caller log ERR.

Every message of severity INFO or higher should either (A) be intelligible to end-users who don't know the Tor source; or (B) somehow inform the end-users that they aren't expected to understand the message (perhaps with a string like "internal error"). Option (A) is to be preferred to option (B).

Assertions In Tor

Assertions should be used for bug-detection only. Don't use assertions to detect bad user inputs, network errors, resource exhaustion, or similar issues.

Tor is always built with assertions enabled, so try to only use tor_assert() for cases where you are absolutely sure that crashing is the least bad option. Many bugs have been caused by use of tor_assert() when another kind of check would have been safer.

If you're writing an assertion to test for a bug that you can recover from, use tor_assert_nonfatal() in place of tor_assert(). If you'd like to write a conditional that incorporates a nonfatal assertion, use the BUG() macro, as in:

if (BUG(ptr == NULL))
	return -1;

Allocator conventions

By convention, any tor type with a name like abc_t should be allocated by a function named abc_new(). This function should never return NULL.

Also, a type named abc_t should be freed by a function named abc_free_(). Don't call this abc_free_() function directly -- instead, wrap it in a macro called abc_free(), using the FREE_AND_NULL macro:

void abc_free_(abc_t *obj);
#define abc_free(obj) FREE_AND_NULL(abc_t, abc_free_, (obj))

This macro will free the underlying abc_t object, and will also set the object pointer to NULL.

You should define all abc_free_() functions to accept NULL inputs:

void
abc_free_(abc_t *obj)
{
  if (!obj)
    return;
  tor_free(obj->name);
  thing_free(obj->thing);
  tor_free(obj);
}

If you need a free function that takes a void * argument (for example, to use it as a function callback), define it with a name like abc_free_void():

static void
abc_free_void_(void *obj)
{
  abc_free_(obj);
}

When deallocating, don't say e.g. if (x) tor_free(x). The convention is to have deallocators do nothing when NULL pointer is passed.

Doxygen comment conventions

Say what functions do as a series of one or more imperative sentences, as though you were telling somebody how to be the function. In other words, DO NOT say:

/** The strtol function parses a number.
 *
 * nptr -- the string to parse.  It can include whitespace.
 * endptr -- a string pointer to hold the first thing that is not part
 *    of the number, if present.
 * base -- the numeric base.
 * returns: the resulting number.
 */
long strtol(const char *nptr, char **nptr, int base);

Instead, please DO say:

/** Parse a number in radix <b>base</b> from the string <b>nptr</b>,
 * and return the result.  Skip all leading whitespace.  If
 * <b>endptr</b> is not NULL, set *<b>endptr</b> to the first character
 * after the number parsed.
 **/
long strtol(const char *nptr, char **nptr, int base);

Doxygen comments are the contract in our abstraction-by-contract world: if the functions that call your function rely on it doing something, then your function should mention that it does that something in the documentation. If you rely on a function doing something beyond what is in its documentation, then you should watch out, or it might do something else later.