regexps.com
When a single development path splits into two paths, that's called a branch . Typically, a branch is followed by a merge -- adding the changes made in a branch back to the branch from which it diverged.
This chapter explains how to create a branch and the simplest way to merge changes from two branches. Later chapters will explain fancier techniques, useful in more complex situations.
In arch
, branches are "tags".
A
tag
is an alternative name for some revision. For example, one
could use the tag release-candidate
to name whatever revision people
should download for testing purposes, regardless of what branch,
version, or patch level the release candidate happens to be on.
In arch
, tags are revisions on ordinary branches. For example,
suppose we are developing arch--devo--0.5
and want to create a tag
release-candidate
to mark revisions which "early adopters"
should test. First, create a development path for the tag:
% larch make-branch arch--release-candidate % larch make-version arch--release-candidate--0.5
To tag a specific revision, use the command:
% larch tag SOURCE-REVISION TAG-REVISION
Note that source-revision
and tag-revision
may be abbreviated.
For example, to tag the most recent revision of arch--devo
:
% larch tag arch--devo arch--release-candidate
or
% larch tag arch--devo--0.5 arch--release-candidate--0.5
or
% larch tag arch--devo--0.5--patch-37 \ arch--release-candidate--0.5
After such a command, you can retrieve the tagged revision in the ordinary way:
% larch get arch--release-candidate--0.5
Note that when you get
a tag that way, the default version of the
resulting project tree is the tag's version, not the tagged version
(see Labelling Project Trees).
You can always update a tag, making it point to a later revision,
again using the tag
command:
% larch tag arch--devo--0.5--patch-53 \ arch--release-candidate--0.5
You can see the history of a tag in the usual way, too:
% larch revisions --summary arch--release-candidate--0.5 base-0 tag of joe.hacker@gnu.org--arch/arch--devo--0.5--patch-37 patch-1 tag of joe.hacker@gnu.org--arch/arch--devo--0.5--patch-53
There is another way to create a branch. This technique is slower,
but has the side effect of leaving you with a working directory for
the new branch. The effect on the repository is the same as when
using the tag
command.
Creating a new branch from an existing branch can be accomplished by this sequence of commands:
% larch prepare-branch OLD-REVISION NEW-BRANCH-VERSION WORKING-DIR % cd WORKING-DIR [...edit log message...] % larch finish-branch
but it is worth deeply understanding that prepare-branch
is
equivalent to:
# This command gets the source tree for the revision we # are branching from: # % larch get OLD-REVISION WORKING-DIR
# These commands switch the working directory to the # new branch: # % cd WORKING-DIR % larch add-log NEW-BRANCH-VERSION % larch set-tree-version NEW-BRANCH-VERSION
# These commands create a "continuation" base-revision # for the new branch: # % larch make-log
and that finish-branch
is the same as:
% larch make-branch NEW-BRANCH % larch make-version NEW-BRANCH-VERSION % larch commit --continuation OLD-REVISION
For example, to create a branch hello--experimental--1.0
from the
latest revision of hello--devo--1.0
, use:
% larch prepare-branch hello--devo--1.0 \ hello--experimental--1.0 \ wd
% cd wd
[...edit log message...]
% larch finish-branch
There is no requirement that a branch be stored in the same archive as the revision from which it branched. For example, you can create a private archive, and store some branches there -- only merging those changes back into the shared archive when they are ready.
Here's a tip: make your private archive your default archive. Use fully-qualified version and revision names when getting or committing revisions in the shared archive. This makes it less likely that you'll accidently make unintended changes to the shared archive.
If you have a project tree for a branch, you might want to know what has happened in the version from which you branched.
The whats-missing
is used for this. In a working directory for a
branch, use:
% larch whats-missing --summary ORIGINAL-VERSION
where ORIGINAL-VERSION
is the version name of the version from which
you branched. Actually, ORIGINAL-VERSION
can be any version for
which your project tree has a patch log.
The whats-missing
command is explained in greater detail in the next
chapter (see Patch Logs and ChangeLogs).
Similarly, update
and replay
work for any version for which a
project tree has a patch log, such as a version from which a branch
occurred:
% larch update OLD-DIR NEW-DIR [archive/]VERSION
% larch replay OLD-DIR NEW-DIR [archive/]VERSION
The simplest use of branching and merging is this: you have one development path, call it the "trunk". You form a branch from that development path, which we'll just call "the branch".
To make some changes, you do your work on the branch: check out the latest revision from the branch, make changes, commit, make more changes, commit again, etc.
As you work, you might sometimes need to "catch up" to changes made
to the trunk. You can do that by using update
or replay
.
When you're done, and the branch is fully up-to-date with the trunk, you can check out the latest branch revision, then commit that version to the trunk. All of the changes that you made on the branch will be summarized into a single patch.
There are more complicated and more realistic uses of branches. These are the subjects of the next several chapters.
arch: The arch Revision Control Systemregexps.com