Most importantly, we never say "It’s in POSIX; we’ll happily ignore your needs should your system not conform to it." We live in the real world.

However, we often say "Let’s stay away from that construct, it’s not even in POSIX".

In spite of the above two rules, we sometimes say "Although this is not in POSIX, it (is so convenient | makes the code much more readable | has other good characteristics) and practically all the platforms we care about support it, so let’s use it".

Again, we live in the real world, and it is sometimes a
judgement call, the decision based more on real world
constraints people face than what the paper standard says.

Fixing style violations while working on a real change as a preparatory clean-up step is good, but otherwise avoid useless code churn for the sake of conforming to the style.

"Once it _is_ in the tree, it's not really worth the patch noise to
go and fix it up."
Cf. http://article.gmane.org/gmane.linux.kernel/943020

We use tabs for indentation.

Case arms are indented at the same depth as case and esac lines, like this:

case "$variable" in
pattern1)
	do this
	;;
pattern2)
	do that
	;;
esac

Redirection operators should be written with space before, but no space after them. In other words, write echo test >"$file" instead of echo test> $file or echo test > $file. Note that even though it is not required by POSIX to double-quote the redirection target in a variable (as shown above), our code does so because some versions of bash issue a warning without the quotes.

(incorrect)
cat hello > world < universe
echo hello >$world

(correct)
cat hello >world <universe
echo hello >"$world"

We prefer $( …​ ) for command substitution; unlike ``, it properly nests. It should have been the way Bourne spelled it from day one, but unfortunately isn’t.

If you want to find out if a command is available on the user’s $PATH, you should use type <command>, instead of which <command>. The output of which is not machine parseable and its exit code is not reliable across platforms.

We use POSIX compliant parameter substitutions and avoid bashisms; namely:

We use ${parameter-word} and its [-=?+] siblings, and their colon’ed "unset or null" form.

We use ${parameter#word} and its [#%] siblings, and their doubled "longest matching" form.

No "Substring Expansion" ${parameter:offset:length}.

No shell arrays.

No strlen ${#parameter}.

No pattern replacement ${parameter/pattern/string}.

We use Arithmetic Expansion $…​.

Inside Arithmetic Expansion, spell shell variables with $ in front of them, as some shells do not grok $x while accepting $$x just fine (e.g. dash older than 0.5.4).

We do not use Process Substitution <(list) or >(list).

Do not write control structures on a single line with semicolon. "then" should be on the next line for if statements, and "do" should be on the next line for "while" and "for".

(incorrect)
if test -f hello; then
	do this
fi

(correct)
if test -f hello
then
	do this
fi

We prefer "test" over "[ …​ ]".

We do not write the noiseword "function" in front of shell functions.

We prefer a space between the function name and the parentheses, and no space inside the parentheses. The opening "{" should also be on the same line.

(incorrect)
my_function(){
	...

(correct)
my_function () {
	...

As to use of grep, stick to a subset of BRE (namely, no \{m,n\}, [::], [==], or [..]) for portability.

We do not use \{m,n\};

We do not use -E;

We do not use ? or + (which are \{0,1\} and \{1,\} respectively in BRE) but that goes without saying as these are ERE elements not BRE (note that \? and \+ are not even part of BRE — making them accessible from BRE is a GNU extension).

Use Git’s gettext wrappers in git-sh-i18n to make the user interface translatable. See "Marking strings for translation" in po/README.

We do not write our "test" command with "-a" and "-o" and use "&&" or "||" to concatenate multiple "test" commands instead, because the use of "-a/-o" is often error-prone. E.g.

test -n "$x" -a "$a" = "$b"

is buggy and breaks when $x is "=", but

test -n "$x" && test "$a" = "$b"

does not have such a problem.

We use tabs to indent, and interpret tabs as taking up to 8 spaces.

We try to keep to at most 80 characters per line.

As a Git developer we assume you have a reasonably modern compiler and we recommend you to enable the DEVELOPER makefile knob to ensure your patch is clear of all compiler warnings we care about, by e.g. "echo DEVELOPER=1 >>config.mak".

We try to support a wide range of C compilers to compile Git with, including old ones. That means that you should not use C99 initializers, even if a lot of compilers grok it.

Variables have to be declared at the beginning of the block.

NULL pointers shall be written as NULL, not as 0.

When declaring pointers, the star sides with the variable name, i.e. "char string", not "char string" or "char * string". This makes it easier to understand code like "char *string, c;".

Use whitespace around operators and keywords, but not inside parentheses and not around functions. So:

      while (condition)
func(bar + 1);

and not:

      while( condition )
func (bar+1);

We avoid using braces unnecessarily. I.e.

if (bla) {
	x = 1;
}

is frowned upon.  A gray area is when the statement extends
over a few lines, and/or you have a lengthy comment atop of
it.  Also, like in the Linux kernel, if there is a long list
of "else if" statements, it can make sense to add braces to
single line blocks.

We try to avoid assignments in the condition of an "if" statement.

Try to make your code understandable. You may put comments in, but comments invariably tend to stale out when the code they were describing changes. Often splitting a function into two makes the intention of the code much clearer.

Multi-line comments include their delimiters on separate lines from the text. E.g.

/*
 * A very long
 * multi-line comment.
 */

Note however that a comment that explains a translatable string to
translators uses a convention of starting with a magic token
"TRANSLATORS: " immediately after the opening delimiter, even when
it spans multiple lines.  We do not add an asterisk at the beginning
of each line, either.  E.g.

/* TRANSLATORS: here is a comment that explains the string
   to be translated, that follows immediately after it */
_("Here is a translatable string explained by the above.");

Double negation is often harder to understand than no negation at all.

There are two schools of thought when it comes to comparison, especially inside a loop. Some people prefer to have the less stable value on the left hand side and the more stable value on the right hand side, e.g. if you have a loop that counts variable i down to the lower bound,

while (i > lower_bound) {
	do something;
	i--;
}

Other people prefer to have the textual order of values match the
actual order of values in their comparison, so that they can
mentally draw a number line from left to right and place these
values in order, i.e.

while (lower_bound < i) {
	do something;
	i--;
}

Both are valid, and we use both.  However, the more "stable" the
stable side becomes, the more we tend to prefer the former
(comparison with a constant, "i > 0", is an extreme example).
Just do not mix styles in the same part of the code and mimic
existing styles in the neighbourhood.

There are two schools of thought when it comes to splitting a long logical line into multiple lines. Some people push the second and subsequent lines far enough to the right with tabs and align them:

      if (the_beginning_of_a_very_long_expression_that_has_to ||
span_more_than_a_single_line_of ||
the_source_text) {
              ...

while other people prefer to align the second and the subsequent
lines with the column immediately inside the opening parenthesis,
with tabs and spaces, following our "tabstop is always a multiple
of 8" convention:

   if (the_beginning_of_a_very_long_expression_that_has_to ||
span_more_than_a_single_line_of ||
the_source_text) {
           ...

Both are valid, and we use both.  Again, just do not mix styles in
the same part of the code and mimic existing styles in the
neighbourhood.

When splitting a long logical line, some people change line before a binary operator, so that the result looks like a parse tree when you turn your head 90-degrees counterclockwise:

   if (the_beginning_of_a_very_long_expression_that_has_to
|| span_more_than_a_single_line_of_the_source_text) {

while other people prefer to leave the operator at the end of the
line:

   if (the_beginning_of_a_very_long_expression_that_has_to ||
span_more_than_a_single_line_of_the_source_text) {

Both are valid, but we tend to use the latter more, unless the
expression gets fairly complex, in which case the former tends to
be easier to read.  Again, just do not mix styles in the same part
of the code and mimic existing styles in the neighbourhood.

When splitting a long logical line, with everything else being equal, it is preferable to split after the operator at higher level in the parse tree. That is, this is more preferable:

if (a_very_long_variable * that_is_used_in +
    a_very_long_expression) {
	...

than

if (a_very_long_variable *
    that_is_used_in + a_very_long_expression) {
	...

Some clever tricks, like using the !! operator with arithmetic constructs, can be extremely confusing to others. Avoid them, unless there is a compelling reason to use them.

Use the API. No, really. We have a strbuf (variable length string), several arrays with the ALLOC_GROW() macro, a string_list for sorted string lists, a hash map (mapping struct objects) named "struct decorate", amongst other things.

When you come up with an API, document it.

The first #include in C files, except in platform specific compat/ implementations, must be either "git-compat-util.h", "cache.h" or "builtin.h". You do not have to include more than one of these.

A C file must directly include the header files that declare the functions and the types it uses, except for the functions and types that are made available to it by including one of the header files it must include by the previous rule.

If you are planning a new command, consider writing it in shell or perl first, so that changes in semantics can be easily changed and discussed. Many Git commands started out like that, and a few are still scripts.

Avoid introducing a new dependency into Git. This means you usually should stay away from scripting languages not already used in the Git core command set (unless your command is clearly separate from it, such as an importer to convert random-scm-X repositories to Git).

When we pass <string, length> pair to functions, we should try to pass them in that order.

Use Git’s gettext wrappers to make the user interface translatable. See "Marking strings for translation" in po/README.

Most of the C guidelines above apply.

We try to support Perl 5.8 and later ("use Perl 5.008").

use strict and use warnings are strongly preferred.

Don’t overuse statement modifiers unless using them makes the result easier to follow.

We try to avoid assignments inside "if ()" conditions.

Learn and use Git.pm if you need that functionality.

For Emacs, it’s useful to put the following in GIT_CHECKOUT/.dir-locals.el, assuming you use cperl-mode:

;; note the first part is useful for C editing, too
((nil . ((indent-tabs-mode . t)
              (tab-width . 8)
              (fill-column . 80)))
 (cperl-mode . ((cperl-indent-level . 8)
                (cperl-extra-newline-before-brace . nil)
                (cperl-merge-trailing-else . t))))

We follow PEP-8 (http://www.python.org/dev/peps/pep-0008/).

As a minimum, we aim to be compatible with Python 2.6 and 2.7.

Where required libraries do not restrict us to Python 2, we try to also be compatible with Python 3.1 and later.

When you must differentiate between Unicode literals and byte string literals, it is OK to use the b prefix. Even though the Python documentation for version 2.6 does not mention this prefix, it has been supported since version 2.6.0.

Do not end error messages with a full stop.

Do not capitalize ("unable to open %s", not "Unable to open %s")

Say what the error is first ("cannot open %s", not "%s: cannot open")

For configuration variable names, follow the existing convention: