B2 is an open-source project. This means that we welcome and appreciate all contributions — be it ideas, bug reports, or patches. This document contains guidelines which helps to assure that development goes on smoothly, and changes are made quickly.
The guidelines are not mandatory, and you can decide for yourself which one to follow. But note, the 10 mins that you spare writing a comment, for example, might lead to significantly longer delay for everyone.
Both bugs and patches can be submitted to the GitHub tracker.
When reporting a bug, please try to provide the following information:
-
What you did.
-
A minimal reproducible test case is very much appreciated.
-
Shell script with some annotations is much better than verbose description of the problem.
-
A regression test is the best (see test/test_system.html).
-
What you got.
-
What you expected.
-
What version of B2 did you use. If possible, please try to test with the main branch state.
When submitting a patch, please:
-
Make a single patch for a single logical change
-
Follow the policies and coding conventions below
-
Send patches as pull requests to the main branch
-
Provide a good PR message together with the patch
The purpose of message serves to communicate what was changed, and why. Without a good message, you might spend a lot of time later, wondering where a strange piece of code came from and why it was necessary.
The good message mentions each changed file and each rule/method, saying what happened to it, and why. Consider, the following log message
Better direct request handling. * new/build-request.jam (directly-requested-properties-adjuster): Redo. * new/targets.jam (main-target.generate-really): Adjust properties here. * new/virtual-target.jam (register-actual-name): New rule. (virtual-target.actualize-no-scanner): Call the above, to detected bugs, where two virtual target correspond to one Jam target name.
The messages for the last two files are good. They tell what was changed. The change to the first file is clearly under-commented.
It’s okay to use terse messages for uninteresting changes, like ones induced by interface changes elsewhere.
All serious changes must be tested. New rules must be tested by the module where they are declared. The test system (test/test_system.html) should be used to verify user-observable behavior.
It turns out that it’s hard to have too much comments, but it’s easy to have too little. Please prepend each rule with a comment saying what the rule does and what arguments mean. Stop for a minute and consider if the comment makes sense for anybody else, and completely describes what the rules does. Generic phrases like "adjusts properties" are really not enough.
When applicable, make changes to the user documentation as well.
-
All names of rules and variables are lowercase with "-" to separate words.
rule call-me-ishmael ( ) ...
-
Names with dots in them are "intended globals". Ordinary globals use a dot prefix:
.foobar $(.foobar)
-
Pseudofunctions or associations are <parameter>.<property>:
$(argument).name = hello ; $($(argument).name)
-
Class attribute names are prefixed with "self.":
self.x $(self.x)
-
Builtin rules are called via their ALL_UPPERCASE_NAMES:
DEPENDS $(target) : $(sources) ;
-
Opening and closing braces go on separate lines:
if $(a) { # } else { # }
Developing in the b2
engine, the C++ part, requires two steps to be
effective: building the "stable" engine, and developing the
"in-progress" engine.
Caution
|
The b2 only supports building when the version of the engine
matches the version of the build files. This is an important consideration
to take when deciding what to choose as "stable" and "in-progress" engine.
These instructions are written assuming the engine and the build versions match.
|
What is the "stable" engine is up to you. It only refers to a build of the
engine you know is at a good working state. When you are at a point the
source is stable you can run bootstrap.sh/bat
from the root. That will
create the b2
executable at the root. You can then use this version to run
regular B2 builds as needed both within the B2 tree and in other projects.
The "in-progress" engine is whatever build you happen to be testing at the
moment. There are two ways to build this be engine. You can either
(a) run b2 b2
at the root, or (b) run build.sh/bat
in src/engine
.
Using (a) will place, by default, a fully debuggable b2
in the .build
directories. You can run that one from a debugger with full symbols and
stepping features. This should be the first choice in developing in the
engine.
After using (a) to implement functionality you can use (b) to fully test
that functionality. The engine built from (b) is fully optimized and
is the one used, by default, by the test system when running in the test
directory. Before submitting patches it’s required to build this way and
run the tests in at least one toolset version (but preferably at least two).