ソースコードのサブディレクトリー¶

There are a lot of subdirectories in the source tree! Which ones should you concentrate on first? lib and daemon which contain the source code of the core library. generator is the code generator described above, so that is important. The Makefile.am in the root directory will tell you in which order the subdirectories get built. And then if you are looking at a particular tool (eg. v2v) or language binding (eg. python), go straight to that subdirectory, but remember that if you didn't run the generator yet, then you may find files which appear to be missing.

align

virt-alignment-scan(1) のコマンドおよびドキュメント。

appliance

libguestfs アプライアンス、スクリプトなどを構築します。

bash

Bash tab-completion scripts.

build-aux

autotools により使用されるさまざまなビルドスクリプト。

builder

virt-builder(1) コマンドおよびドキュメント。

cat

The virt-cat(1), virt-filesystems(1), virt-log(1), virt-ls(1) and virt-tail(1) commands and documentation.

common

Various libraries of internal code can be found in the common subdirectory:

contrib

外部の貢献、実験的な部分です。

customize

virt-customize(1) command and documentation.

daemon

libguestfs アプライアンスの中で実行され、アクションを実行しているデーモン。

df

virt-df(1) のコマンドおよびドキュメント。

dib

virt-dib(1) command and documentation.

diff

virt-diff(1) command and documentation.

docs

Miscellaneous manual pages.

edit

virt-edit(1) のコマンドおよびドキュメント。

examples

C API のコード例。

fish

guestfish(1) コマンドラインシェル、および virt-copy-in(1), virt-copy-out(1), virt-tar-in(1), virt-tar-out(1) のようなさまざまなシェルスクリプト。

format

virt-format(1) のコマンドおよびドキュメント。

fuse

guestmount(1), libguestfs の上に組み立てられた FUSE (ユーザー空間ファイルシステム)。

generator

The crucially important generator, used to automatically generate large amounts of boilerplate C code for things like RPC and bindings.

get-kernel

virt-get-kernel(1) command and documentation.

gnulib

gnulib はポートブルなライブラリとして使用されます。gnulib のコピーがこの下に含まれます。

inspector

virt-inspector(1), 仮想マシンイメージ検査ツール。

lib

C ライブラリーのソースコード。

logo

Logo used on the website. The fish is called Arthur by the way.

m4

M4 macros used by autoconf. See "THE BUILD SYSTEM".

make-fs

virt-make-fs(1) command and documentation.

p2v

virt-p2v(1) command, documentation and scripts for building the virt-p2v ISO or disk image.

po

シンプルな gettext 文字列の翻訳。

po-docs

The build infrastructure and PO files for translations of manpages and POD files. Eventually this will be combined with the po directory, but that is rather complicated.

rescue

virt-rescue(1) のコマンドおよびドキュメント。

resize

virt-resize(1) のコマンドおよびドキュメント。

sparsify

virt-sparsify(1) のコマンドおよびドキュメント。

sysprep

virt-sysprep(1) コマンドおよびドキュメント。

tests

テストします。

test-data

Files and other test data used by the tests.

test-tool

エンドユーザーが QEMU/カーネルの組み合わせが libguestfs で動作するかどうかを確認するためのテストツールです。

tmp

Used for temporary files when running the tests (instead of /tmp etc). The reason is so that you can run multiple parallel tests of libguestfs without having one set of tests overwriting the appliance created by another.

tools

Perl で書かれたコマンドラインツール (virt-win-reg(1) および他の多くのもの)。

utils

Miscellaneous utilities, such as "boot-benchmark".

v2v

virt-v2v(1) command and documentation.

website

The http://libguestfs.org website files.

csharp

erlang

gobject

golang

haskell

java

lua

ocaml

php

perl

python

ruby

言語バインディング。

THE BUILD SYSTEM¶

Libguestfs uses the GNU autotools build system (autoconf, automake, libtool).

The ./configure script is generated from configure.ac and m4/guestfs-*.m4. Most of the configure script is split over many m4 macro files by topic, for example m4/guestfs-daemon.m4 deals with the dependencies of the daemon.

The job of the top level Makefile.am is mainly to list the subdirectories ("SUBDIRS") in the order they should be compiled.

common-rules.mk is included in every Makefile.am (top level and subdirectories). subdir-rules.mk is included only in subdirectory Makefile.am files.

There are many make targets. Use this command to list them all:

 make help

ADDING A NEW API¶

Because large amounts of boilerplate code in libguestfs are generated, this makes it easy to extend the libguestfs API.

To add a new API action there are two changes:

1.

You need to add a description of the call (name, parameters, return type, tests, documentation) to generator/actions_*.ml and possibly generator/proc_nr.ml.

There are two sorts of API action, depending on whether the call goes through to the daemon in the appliance, or is serviced entirely by the library (see "ARCHITECTURE" in guestfs-internals(1)). "guestfs_sync" in guestfs(3) is an example of the former, since the sync is done in the appliance. "guestfs_set_trace" in guestfs(3) is an example of the latter, since a trace flag is maintained in the handle and all tracing is done on the library side.

Most new actions are of the first type, and get added to the "daemon_functions" list. Each function has a unique procedure number used in the RPC protocol which is assigned to that action when we publish libguestfs and cannot be reused. Take the latest procedure number and increment it.

For library-only actions of the second type, add to the "non_daemon_functions" list. Since these functions are serviced by the library and do not travel over the RPC mechanism to the daemon, these functions do not need a procedure number, and so the procedure number is set to "-1".

2.

Implement the action (in C):

For daemon actions, implement the function "do_<name>" in the "daemon/" directory.

For library actions, implement the function "guestfs_impl_<name>" in the "lib/" directory.

In either case, use another function as an example of what to do.

3.

As an alternative to step 2: Since libguestfs 1.38, daemon actions can be implemented in OCaml. You have to set the "impl = OCaml ..." flag in the generator. Take a look at daemon/file.ml for an example.

これらの変更をした後、コンパイルするために "make" を使用してください。

Note that you don’t need to implement the RPC, language bindings, manual pages or anything else. It’s all automatically generated from the OCaml description.

Adding tests for an API

You can supply zero or as many tests as you want per API call. The tests can either be added as part of the API description (generator/actions_*.ml), or in some rarer cases you may want to drop a script into "tests/*/". Note that adding a script to "tests/*/" is slower, so if possible use the first method.

The following describes the test environment used when you add an API test in actions_*.ml.

テスト環境は 4 個のブロックデバイスを持ちます:

/dev/sda 2 GB

テスト用の一般的なブロックデバイス。

/dev/sdb 2 GB

/dev/sdb1 is an ext2 filesystem used for testing filesystem write operations.

/dev/sdc 10 MB

2 つのブロックデバイスが必要となるいくつかのテストにおいて使用されます。

/dev/sdd

ISO with fixed content (see images/test.iso).

To be able to run the tests in a reasonable amount of time, the libguestfs appliance and block devices are reused between tests. So don't try testing "guestfs_kill_subprocess" in guestfs(3) :-x

Each test starts with an initial scenario, selected using one of the "Init*" expressions, described in generator/types.ml. These initialize the disks mentioned above in a particular way as documented in types.ml. You should not assume anything about the previous contents of other disks that are not initialized.

You can add a prerequisite clause to any individual test. This is a run-time check, which, if it fails, causes the test to be skipped. Useful if testing a command which might not work on all variations of libguestfs builds. A test that has prerequisite of "Always" means to run unconditionally.

In addition, packagers can skip individual tests by setting environment variables before running "make check".

 SKIP_TEST_<CMD>_<NUM>=1

eg: "SKIP_TEST_COMMAND_3=1" skips test #3 of "guestfs_command" in guestfs(3).

または:

 SKIP_TEST_<CMD>=1

eg: "SKIP_TEST_ZEROFREE=1" skips all "guestfs_zerofree" in guestfs(3) tests.

Packagers can run only certain tests by setting for example:

 TEST_ONLY="vfs_type zerofree"

See tests/c-api/tests.c for more details of how these environment variables work.

Debugging new APIs

Test new actions work before submitting them.

新しいコマンドを試すために guestfish を使うことができます。

Debugging the daemon is a problem because it runs inside a minimal environment. However you can fprintf messages in the daemon to stderr, and they will show up if you use "guestfish -v".

ADDING A NEW LANGUAGE BINDING¶

All language bindings must be generated by the generator (see the generator subdirectory).

There is no documentation for this yet. We suggest you look at an existing binding, eg. generator/ocaml.ml or generator/perl.ml.

Adding tests for language bindings

Language bindings should come with tests. Previously testing of language bindings was rather ad-hoc, but we have been trying to formalize the set of tests that every language binding should use.

Currently only the OCaml and Perl bindings actually implement the full set of tests, and the OCaml bindings are canonical, so you should emulate what the OCaml tests do.

This is the numbering scheme used by the tests:

 - 000+ basic tests:
 
   010  load the library
   020  create
   030  create-flags
   040  create multiple handles
   050  test setting and getting config properties
   060  explicit close
   065  implicit close (in GC'd languages)
   070  optargs
   080  version
   090  retvalues
 
 - 100  launch, create partitions and LVs and filesystems
 
 - 400+ events:
 
   410  close event
   420  log messages
   430  progress messages
 
 - 800+ regression tests (specific to the language)
 
 - 900+ any other custom tests for the language

To save time when running the tests, only 100, 430, 800+, 900+ should launch the handle.

FORMATTING CODE¶

Our C source code generally adheres to some basic code-formatting conventions. The existing code base is not totally consistent on this front, but we do prefer that contributed code be formatted similarly. In short, use spaces-not-TABs for indentation, use 2 spaces for each indentation level, and other than that, follow the K&R style.

If you use Emacs, add the following to one of your start-up files (e.g., ~/.emacs), to help ensure that you get indentation right:

 ;;; In libguestfs, indent with spaces everywhere (not TABs).
 ;;; Exceptions: Makefile and ChangeLog modes.
 (add-hook 'find-file-hook
     '(lambda () (if (and buffer-file-name
                          (string-match "/libguestfs\\>"
                              (buffer-file-name))
                          (not (string-equal mode-name "Change Log"))
                          (not (string-equal mode-name "Makefile")))
                     (setq indent-tabs-mode nil))))
 
 ;;; When editing C sources in libguestfs, use this style.
 (defun libguestfs-c-mode ()
   "C mode with adjusted defaults for use with libguestfs."
   (interactive)
   (c-set-style "K&R")
   (setq c-indent-level 2)
   (setq c-basic-offset 2))
 (add-hook 'c-mode-hook
           '(lambda () (if (string-match "/libguestfs\\>"
                               (buffer-file-name))
                           (libguestfs-c-mode))))

変更のテスト方法¶

Turn warnings into errors when developing to make warnings hard to ignore:

 ./configure --enable-werror

有用なターゲットは次のとおりです:

"make check"

一般的なテスト群を実行します。

This is implemented using the regular automake "TESTS" target. See the automake documentation for details.

"make check-valgrind"

valgrind にあるテスト群のサブセットを実行します。

See "VALGRIND" below.

"make check-valgrind-local-guests"

ローカルにインストールされた libvirt 仮想マシン (読み込み専用) を使用して、valgrind にあるテスト群のサブセットを実行します。

"make check-direct"

Runs all tests using default appliance back-end. This only has any effect if a non-default backend was selected using "./configure --with-default-backend=..."

"make check-valgrind-direct"

Run a subset of the test suite under valgrind using the default appliance back-end.

"make check-uml"

Runs all tests using the User-Mode Linux backend.

As there is no standard location for the User-Mode Linux kernel, you have to set "LIBGUESTFS_HV" to point to the kernel image, eg:

 make check-uml LIBGUESTFS_HV=~/d/linux-um/vmlinux
    

"make check-valgrind-uml"

Runs all tests using the User-Mode Linux backend, under valgrind.

As above, you have to set "LIBGUESTFS_HV" to point to the kernel.

"make check-with-upstream-qemu"

Runs all tests using a local qemu binary. It looks for the qemu binary in QEMUDIR (defaults to $HOME/d/qemu), but you can set this to another directory on the command line, eg:

 make check-with-upstream-qemu QEMUDIR=/usr/src/qemu
    

"make check-with-upstream-libvirt"

Runs all tests using a local libvirt. This only has any effect if the libvirt backend was selected using "./configure --with-default-backend=libvirt"

It looks for libvirt in LIBVIRTDIR (defaults to $HOME/d/libvirt), but you can set this to another directory on the command line, eg:

 make check-with-upstream-libvirt LIBVIRTDIR=/usr/src/libvirt
    

"make check-slow"

Runs some slow/long-running tests which are not run by default.

To mark a test as slow/long-running:

"sudo make check-root"

Runs some tests which require root privileges. These are supposed to be safe, but take care. You have to run this as root (eg. using sudo(8) explicitly).

To mark a test as requiring root:

"make check-all"

Equivalent to running all "make check*" rules except "check-root".

"make check-release"

Runs a subset of "make check*" rules that are required to pass before a tarball can be released. Currently this is:

"make installcheck"

Run "make check" on the installed copy of libguestfs.

The version of installed libguestfs being tested, and the version of the libguestfs source tree must be the same.

Do:

 ./autogen.sh
 make clean ||:
 make
 make installcheck
    

VALGRIND¶

When you do "make check-valgrind", it searches for any Makefile.am in the tree that has a "check-valgrind:" target and runs it.

Writing the Makefile.am and tests correctly to use valgrind and working with automake parallel tests is subtle.

If your tests are run via a shell script wrapper, then in the wrapper use:

 $VG virt-foo

and in the Makefile.am use:

 check-valgrind:
     make VG="@VG@" check

However, if your binaries run directly from the "TESTS" rule, you have to modify the Makefile.am like this:

 LOG_COMPILER = $(VG)
 
 check-valgrind:
     make VG="@VG@" check

In either case, check that the right program is being tested by examining the tmp/valgrind* log files carefully.

SUBMITTING PATCHES¶

パッチをメーリングリストに提出します: http://www.redhat.com/mailman/listinfo/libguestfs および rjones@redhat.com (Cc)。

You do not need to subscribe to the mailing list if you don’t want to. There may be a short delay while your message is moderated.

DAEMON CUSTOM PRINTF FORMATTERS¶

In the daemon code we have created custom printf formatters %Q and %R, which are used to do shell quoting.

%Q

シンプルなシェルクオート文字列。すべての空白と他のシェル文字がエスケープされます。

%R

Same as %Q except the string is treated as a path which is prefixed by the sysroot.

例:

 asprintf (&cmd, "cat %R", path);

"cat /sysroot/some\ path\ with\ spaces" を生成します

Note: Do not use these when you are passing parameters to the "command{,r,v,rv}()" functions. These parameters do NOT need to be quoted because they are not passed via the shell (instead, straight to exec). You probably want to use the "sysroot_path()" function however.

国際化 (i18n) サポート¶

We support i18n (gettext anyhow) in the library.

However many messages come from the daemon, and we don’t translate those at the moment. One reason is that the appliance generally has all locale files removed from it, because they take up a lot of space. So we'd have to readd some of those, as well as copying our PO files into the appliance.

Debugging messages are never translated, since they are intended for the programmers.

HOW OCAML PROGRAMS ARE COMPILED AND LINKED¶

Mostly this section is "how we make automake & ocamlopt work together" since OCaml programs themselves are easy to compile.

Automake has no native support for OCaml programs, ocamlc nor ocamlopt. What we do instead is to treat OCaml programs as C programs which happen to contain these "other objects" ("DEPENDENCIES" in automake-speak) that happen to be the OCaml objects. This works because OCaml programs usually have C files for native bindings etc.

So a typical program is described as just its C sources:

 virt_v2v_SOURCES = ... utils-c.c xml-c.c

For programs that have no explicit C sources, we create an empty dummy.c file, and list that instead:

 virt_resize_SOURCES = dummy.c

The OCaml objects which contain most of the code are listed as automake dependencies (other dependencies may also be listed):

 virt_v2v_DEPENDENCIES = ... cmdline.cmx v2v.cmx

The only other special thing we need to do is to provide a custom link command. This is needed because automake won't assemble the ocamlopt command, the list of objects and the "-cclib" libraries in the correct order otherwise.

 virt_v2v_LINK = \
     $(top_srcdir)/ocaml-link.sh -cclib '-lutils -lgnu' -- ...

The actual rules, which you can examine in v2v/Makefile.am, are a little bit more complicated than this because they have to handle:

• Compiling for byte code or native code.
• The pattern rules needed to compile the OCaml sources to objects.

  These are now kept in subdir-rules.mk at the top level, which is included in every subdirectory Makefile.am.

• Adding OCaml sources files to "EXTRA_DIST".

  Automake isn't aware of the complete list of sources for a binary, so it will not add them all automatically.

VIRT-V2V¶

First a little history. Virt-v2v has been through at least two complete rewrites, so this is probably about the third version (but we don't intend to rewrite it again). The previous version was written in Perl and can be found here: https://git.fedorahosted.org/git/virt-v2v.git

The current version started out as almost a line-for-line rewrite of the Perl code in OCaml + C, and it still has a fairly similar structure. Therefore if there are details of this code that you don't understand (especially in the details of guest conversion), checking the Perl code may help.

The files to start with when reading this code are:

• types.mli
• v2v.ml

types.mli defines all the structures used and passed around when communicating between different bits of the program. v2v.ml controls how the program runs in stages.

After studying those files, you may want to branch out into the input modules (input_*), the output modules (output_*) or the conversion modules (convert_*). The input and output modules define -i and -o options (see the manual). The conversion modules define what guest types we can handle and the detailed steps involved in converting them.

Every other file in this directory is a support module / library of some sort. Some code is written in C, especially where we want to use an external C library such as libxml2.

VIRT-P2V¶

Virt-p2v is a front end on virt-v2v. ie. All it does is act as a GUI front end, and it calls out to virt-v2v to perform the actual conversion. Therefore most of the C code in the p2v/ subdirectory is Gtk (GUI) code, or supporting code for talking to the remote conversion server. There is no special support for physical machines in virt-v2v. They are converted in the same way as foreign VMs.

Running virt-p2v

You can run the p2v/virt-p2v binary directly, but it will try to convert your machine’s real /dev/sda which is unlikely to work well. However virt-p2v also has a test mode in which you can supply a test disk:

 make -C p2v run-virt-p2v-directly

This is a wrapper around the virt-p2v(1) --test-disk option. You can control the "physical machine" disk by setting "PHYSICAL_MACHINE" to point to a disk image.

A more realistic test is to run virt-p2v inside a VM on the local machine. To do that, do:

 make -C p2v run-virt-p2v-in-a-vm

This also runs qemu with the "physical machine" disk (which you can set by setting "PHYSICAL_MACHINE"), a virtual CD, and a variety of network cards for testing. You can change the qemu binary and add extra qemu options by setting "QEMU" and/or "QEMU_OPTIONS" on the make commandline.

A third way to run virt-p2v simulates fairly accurately the program being downloaded over PXE and then doing an automatic conversion of the source physical machine (the non-GUI path -- see next section below):

 make -C p2v run-virt-p2v-non-gui-conversion

Understanding the virt-p2v code

See also: "HOW VIRT-P2V WORKS" in virt-p2v(1)

There are two paths through the code, GUI or non-GUI (parsing the kernel command line):

 main.c ──────┬─────▶ gui.c ──────┬─────▶ conversion.c
              │                   │
              │                   │
              └────▶ kernel.c ────┘

but both paths call back to the conversion.c function "start_conversion" to run the remote virt-v2v.

The main task of gui.c/kernel.c is to populate the virt-v2v configuration (config.c).

During conversion, we need to establish ssh connections, and that is done using two libraries:

 conversion.c ──────▶ ssh.c ──────▶ miniexpect.c

where ssh.c is responsible for managing ssh connections overall, and miniexpect.c implements "expect-like" functionality for talking interactively to the remote virt-v2v conversion server.

(Note that miniexpect is a separate library with its own upstream, so if you patch miniexpect.c, then please make sure the changes get reflected in miniexpect’s upstream too: http://git.annexia.org/?p=miniexpect.git;a=summary)

MAINTAINER MAKEFILE TARGETS¶

These "make" targets probably won’t work and aren't useful unless you are a libguestfs maintainer.

make maintainer-commit

This commits everything in the working directory with the commit message "Version $(VERSION).". You must update configure.ac, clean and rebuild first.

make maintainer-tag

This tags the current HEAD commit with the tag "v$(VERSION)" and one of the messages:

 Version $(VERSION) stable
 Version $(VERSION) development

(See "LIBGUESTFS VERSION NUMBERS" in guestfs(3) for the difference between a stable and development release.)

make maintainer-check-authors

Check that all authors (found in git commit messages) are included in the generator/authors.ml file.

make maintainer-check-extra-dist

This rule must be run after "make dist" (so there is a tarball in the working directory). It compares the contents of the tarball with the contents of git to ensure that no files have been missed from Makefile.am "EXTRA_DIST" rules.

make maintainer-upload-website

This is used by the software used to automate libguestfs releases to copy the libguestfs website to another git repository before it is uploaded to the web server.

MAKING A STABLE RELEASE¶

When we make a stable release, there are several steps documented here. See "LIBGUESTFS VERSION NUMBERS" in guestfs(3) for general information about the stable branch policy.

•

Check "make && make check" works on at least:

• Check "./configure --without-libvirt" works.
• Finalize guestfs-release-notes.pod
• Push and pull from Zanata.

  次を実行します:

   zanata push
      

  to push the latest POT files to Zanata. Then run:

   ./zanata-pull.sh
      

  which is a wrapper to pull the latest translated *.po files.

• Consider updating gnulib to latest upstream version.
• http://libguestfs.org/download の下に新しい安定版および開発版のディレクトリーを作成します。
• Edit website/index.html.in.
• Set the version (in configure.ac) to the new stable version, ie. 1.XX.0, and commit it:

   ./localconfigure
   make distclean -k
   ./localconfigure
   make && make dist
   make maintainer-commit
   make maintainer-tag
      

• Create the stable branch in git:

   git branch stable-1.XX
   git push origin stable-1.XX
      

• Do a full release of the stable branch.
• Set the version to the next development version and commit that. Optionally do a full release of the development branch.