MKSQUASHFS ACTIONS
                           ==================

The new Mksquashfs Actions code allows an "action" to be executed
on a file if one or more "tests" succeed.  If you're familiar
with the "find" command, then an action is similar to "-print",
and a test is similar to say "-name" or "-type".

Actions add greater flexibility when building images from sources.
They can be used to optimise compression, I/O performance, and they
also allow more control on the exclusion of files from the source, and
allow uid/gid and mode to be changed on a file basis.

1. Specification
================

Actions can be specified on the command line with the -action option.
They can also be put into a file, and added  with the -action-file
option.  If put into a file, there is one action per line.  But, lines
can be extended over many lines with continuation (\).

If you want to get a log of what actions were performed, and the values
returned by the tests for each file, you can use the -log-action option
for the command line and -log-action-file for action files.

Similarly there are -true-action (-true-action-file) and -false-action
(-false-action-file) options which log if the tests evaluated to TRUE,
and vice-versa:

2. Syntax
=========

An action consists of two parts, separated by an "@".  The action to
be executed is placed before the @, and one or more tests are
placed afer the @.  If the action or tests has an argument, it is
given in brackets.  Brackets are optional if no argument is needed,
e.g.

compressed()@name("filename")

compressed@name("filename")

do exactly the same thing.

Arguments can be either numeric or string, depending on the
action and test.

String arguments can be enclosed in double-quotes ("), to prevent the
parser from treating characters within it specially.  Within double-quotes
only '\' is treatedly specially, and only at the end of a line.  Special
characters can also be backslashed (\) to prevent interpretation by the
parser, e.g. the following is equivalent:

compressed@name(file\ name\ with\ \&&\ and\ spaces)

compressed@name("file name with && and spaces")

Numeric arguments are of the form [range]number[size], where
[range] is either

	'<' or '-', match on less than number
	'>' or '+', match on greater than number
	"" (nothing), match on exactly number

[size] is either:
	"" (nothing), number
	'k' or 'K',  number * 2^10
	'm' or 'M', number * 2^20
	'g' or 'G', number * 2^30

e.g. the following is equivalent:

compressed@filesize(-81920)
compressed@filesize(<80K)

Both will match on files less than 80K in size.

Characters which are treated specially by the parser are * ( ) && ||
! , @.  Plus whitespace (spaces and tabs).

Note: if the action is typed on the command line, then many special
characters will be evaluated by the shell, and you should always
check what is actually being passed to Mksquashfs.  If in doubt use
-action-file where the additional complexities of shell evaluation is
avoided.

For example this action line will work in an action file

compressed()@name("file name")

But, if typed on the command line, it will need to be:

% mksquashfs source image -action "compressed()@name(\"file name\")"


3. Logical operators
====================

Tests can be combined with the logical operators && (and), || (or) and
can be negated with the unary ! (not).  Expressions thus formed can also
be bracketed with "(" and ")", to create nested expressions.

Operators do not have precedence and are evaluated strictly left to
right.  To enforce precedence use brackets, e.g.

test1 && test2 || test3

will be evaluated as

(test1 && test2) || test3

&& and || are short-circuit operators, where the rhs (right hand side)
is only evaluated if the lhs (left hand side) has been insufficient
to determine the value.  For example in the above, test3 will only be
evaluated if (test1 && test2) evaluates to FALSE.

4. Test operators
=================

The following test operators are supported:

4.1 name(pattern)

Returns TRUE if the filename (basename without leading directory components)
matches pattern.  Pattern can have wildcards.

4.2 pathname(pattern)
---------------------

Returns TRUE if the full pathname of the file matches pattern.
Pattern can have wildcards.

4.3 subpathname(pattern)
------------------------

Returns TRUE if the <n> directory components of pattern match the first <n>
directory components of the pathname.

For example, if pattern has one component:

subpathname(dir1) will match "dir1/somefile", "dir1/dir2/somefile" etc.

If pattern had two components:

subpathname(dir1/dir2) will match ""dir1/dir2/somefile" etc.

Pattern can have wildcards.

4.4 filesize(value)
-------------------

Return TRUE if the size of the file is less than, equal to, or larger than
<value>, where <value> can be [<-]number, number, [>+]number.  Returns FALSE
on anything not a file.

4.5 dirsize(value)
------------------

Return TRUE if the size of the directory is less than, equal to, or larger than
<value>, where <value> can be [<-]number, number, [>+]number.  Returns FALSE
on anything not a directory.

4.6 size(value)
---------------

Return TRUE if the size of the file is less than, equal to, or larger than
<value>, where <value> can be [<-]number, number, [>+]number.  Works on any
file type.

4.7 inode(value)
----------------

Return TRUE if the inode number is less than, equal to, or larger than
<value>, where <value> can be [<-]number, number, [>+]number.

4.8 nlink(value)
----------------

Return TRUE if the nlink count is less than, equal to, or larger than
<value>, where <value> can be [<-]number, number, [>+]number.

4.9 fileblocks(value)
---------------------

Return TRUE if the size of the file in blocks (512 bytes) is less than, equal
to, or larger than <value>, where <value> can be [<-]number, number, [>+]number.
Returns FALSE on anything not a file.

4.10 dirblocks(value)
---------------------

Return TRUE if the size of the directory in blocks (512 bytes) is less than,
equal to, or larger than <value>, where <value> can be [<-]number, number,
[>+]number.  Returns FALSE on anything not a directory.

4.11 blocks(value)
------------------

Return TRUE if the size of the file in blocks (512 bytes) is less than, equal
to, or larger than <value>, where <value> can be [<-]number, number, [>+]number.
Works on any file type.

4.12 uid(value)
---------------
Return TRUE if the uid value is less than, equal to, or larger than
<value>, where <value> can be [<-]number, number, [>+]number.

4.13 gid(value)
---------------

Return TRUE if the gid value is less than, equal to, or larger than
<value>, where <value> can be [<-]number, number, [>+]number.

4.14 user(string)
-----------------

Return TRUE if the file owner matches <string>.

4.15 group(string)
------------------

Return TRUE if the file group matches <string>.

4.16 depth(value)
-----------------

Return TRUE if file is at depth less than, equal to, or larger than <value>,
where <value> can be [<-]number, number, [>+]number.  Top level directory is
depth 1.

4.17 dircount(value)
--------------------

Return TRUE if the number of files in the directory is less than, equal to, or
larger than <value>, where <value> can be [<-]number, number, [>+]number.
Returns FALSE on anything not a directory.

4.18 filesize_range(minimum, maximum)
-------------------------------------

Return TRUE if the size of the file is within the range [<minimum>, <maximum>]
inclusive.  Returns FALSE on anything not a file.

4.19 dirsize_range(minimum, maximum)
------------------------------------

Return TRUE if the size of the directory is within the rang
[<minimum>, <maximum>] inclusive.  Returns FALSE on anything not a directory.

4.20 size_range(minimum, maximum)
---------------------------------

Return TRUE if the size of the file is within the range [<minimum>, <maximum>]
inclusive.  Works on any file type.

4.21 inode_range(minimum, maximum)
----------------------------------

Return TRUE if the inode number is within the range [<minimum>, <maximum>]
inclusive.

4.22 fileblocks_range(minimum, maximum)
---------------------------------------

Return TRUE if the size of the file in blocks (512 bytes) is within the range
[<minimum>, <maximum>] inclusive.  Returns FALSE on anything not a file.

4.23 dirblocks_range(minimum, maximum)
--------------------------------------

Return TRUE if the size of the directory in blocks (512 bytes) is within the
range [<minimum>, <maximum>] inclusive.  Returns FALSE on anything not a
directory.

4.24 blocks_range(minimum, maximum)
-----------------------------------

Return TRUE if the size of the file in blocks (512 bytes) is within the range
[<minimum>, <maximum>] inclusive.  Works on any file type.

4.25 uid_range(minimum, maximum)
--------------------------------

Return TRUE if the file uid is within the range [<minimum>, <maximum>]
inclusive.

4.26 gid_range(minimum, maximum)
--------------------------------

Return TRUE if the file gid is within the range [<minimum>, <maximum>]
inclusive.

4.27 depth_range(minimum, maximum)
----------------------------------

Return TRUE if file is at depth within the range [<minimum>, <maximum>].
Top level directory is depth 1.

4.28 dircount_range(minimum, maximum)
-------------------------------------

Returns TRUE is the number of files in the directory is within the range
[<minimum>, <maximum>].  Returns FALSE on anything not a directory.

4.29 type(c)
------------

Returns TRUE if the file matches type <c>.  <c> can be
	f - regular file
	d - directory
	l - symbolic link
	c - character device
	b - block device
	p - Named Pipe / FIFO
	s - socket


4.30 perm(mode)
---------------

Return TRUE if file permissions match <mode>.  <Mode> is the same as
find's -perm option:

	perm(mode) - TRUE if file's permission bits are exactly <mode>.
		   <mode> can be octal or symbolic.

	perm(-mode) - TRUE if all <mode> permission bits are set for this file.
		    <mode> can be octal or symbolic.

	perm(/mode) - TRUE if any <mode> permission bits are set for this file.
		    <mode> can be octal or symbolic.

	The symbolic mode is of the format [ugoa]*[[+-=]PERMS]+
		PERMS = [rwxXst]+ or [ugo]
		and can be repeated separated with commas.

Examples:

perm(0644) match on a file with permissions exactly rw-r--r--.
perm(u=rw,go=r) as above, but expressed symbolically.

perm(/222) match on a file which is writable for any of user, group, other,
perm(/u=w,g=w,o=w) as above but expressed symbolically,
perm(/ugo=w) as above but specified more concisely.

4.31 file(string)
-----------------

Execute "file command" on file, and return TRUE if the output
matches the substring <string>, for example

file(ASCII text) will return TRUE if the file is ASCII text.

Note, this is an expensive test, and should only be run if the file
has matched a number of other short-circuit tests.

4.32 exists()
-------------

Test if the file pointed to by the symbolic link exists within the
output filesystem, that is, whether the symbolic link has a relative
path and the relative path can be resolved to an entry within the
output filesystem.

If the file isn't a symbolic link then the test always returns TRUE.

4.33 absolute()
---------------

Test if the symbolic link is absolute, which by definition means
it points outside of the output filesystem (unless it is to be mounted
as root).  If the file isn't a symbolic link then the test always returns
FALSE.

4.34 readlink(expression)
-------------------------

Follow or dereference the symbolic link, and evaluate <expression> in
the context of the file pointed to by the symbolic link.  All inode
attributes, pathname, name and depth all refer to the dereferenced
file.

If the symbolic link cannot be dereferenced because it points to something
not in the output filesystem (see exists() function above), then FALSE is
returned.  If the file is not a symbolic link, the result is the same as
<expression>, i.e. readlink(<expression>) == <expression>.

Examples:

readlink("name(*.[ch])") returns TRUE if the file referenced by the symbolic
link matches *.[ch].

Obviously, expressions created with && || etc. can be specified.

readlink("depth(1) && filesize(<20K)") returns TRUE if the file referenced
by the symbolic link is a regular file less than 20K in size and in the
top level directory.

Note: in the above tests the embedded expression to be evaluated is enclosed
in double-quotes ("), this is to prevent the special characters being
evaluated by the parser when parsed at the top-level.  Readlink causes
re-evaluation of the embedded string.

4.36 eval(path, expression)
---------------------------

Follow <path> (arg1), and evaluate the <expression> (arg2) in the
context of the file discovered by following <path>.  All inode attributes,
pathname, name and depth all refer to the file discovered by following
<path>.

This test operation allows you to add additional context to the evaluation
of the file being scanned, such as "if the current file is XXX, test if the
parent is YYY, and then do ...".  Often times you need or want to test
a combination of file status.

The <path> can be absolute (in which case it is from the root directory of the
output filesystem), or it can be relative to the current file.  Obviously
relative paths are more useful.

If the file referenced by <path> does not exist in the output filesystem,
then FALSE is returned.

Examples of usage:

1. If a directory matches pattern, check that it contains a ".git" directory.
   This allows you to exclude git repositories, with a double check that it is
   a git repository by checking for the .git subdirectory.

   prune@name(*linux*) && type(d) && eval(.git, "type(d)")

   This action will match on any directory named *linux*, and exclude it if
   it contains a .git subdirectory.


2. If a file matches a pattern, check that the parent directory matches
   another pattern.  This allows you to delete files if and only if they
   are in a particular directory.

   prune@name(*.[ch]) && eval(.., "name(*linux*)")

   This action will delete *.[ch] files, but, only if they are in a directory
   matching *linux*.

4.37 false
----------

Always returns FALSE.

4.38 true
---------

Always returns TRUE.

5. Actions
==========

An action is something which is done (or applied) to a file if the expression
(made up of the above test operators) returns TRUE.

Different actions are applied in separate phases or gated, rather than being
applied all at once.  This is to ensure that you know what the overall
state of the filesystem is when an action is applied.  Or to put it another
way, if you have an action that depends on another action having already been
processed (for the entire filesystem), you'll want to know that is how
they will be applied.

5.1 Actions applied at source filesystem reading (stage 1)
----------------------------------------------------------

5.1.1 exclude()
---------------

This action excludes all files and directories where the expression
returns TRUE.

Obviously this action allows much greater control over which files are
excluded than the current name/pathname matching.

Examples:

1. Exclude any files/directories belonging to user phillip

exclude@user(phillip)

2. Exclude any regular files larger than 1M

exclude@filesize(>1M)

3. Only archive files/directories to a depth of 3

exclude@depth(>3)

4. As above but also exclude directories at the depth of 3
   (which will be empty due to the above exclusion)

exclude@depth(>3) || (depth(3) && type(d))

Which obviously reduces to

exclude@depth(3) && type(d)

Note: the following tests do not work in stage 1, and so they can't be
used in the exclude() action (see prune() action for explanation and
alternative).

	dircount()
	dircount_range()
	exists()
	absolute()
	readlink()
	eval()

5.2 Actions applied at directory scanning (stage 2)
---------------------------------------------------

5.2.1 fragment(name)
--------------------

Place all files matching the expression into a specialised fragment
named <name>.  This can increase compression and/or improve
I/O by placing similar fragments together.

Examples:

1. fragment(cfiles)@name(*.[ch])

Place all C files into special fragments reserved for them.

2. fragment(phillip)@user(phillip)

Place all files owned by user Phillip into special fragments.

5.2.2 fragments()
-----------------

Tell Mksquashfs to use fragment packing for the files matching the
expression.

For obvious reasons this should be used in conjunction with the global
Mksquashfs option -no-fragments.  By default all files are packed into
fragments if they're less than the block size.

5.2.3 no-fragments()
--------------------

Tell Mksquashfs to not pack the files matching the expression into
fragments.

This can be used where you want to optimise I/O latency by not packing
certain files into fragments.

5.2.4 tailend()
---------------

Tell Mksquashfs to use tail-end packing for the files matching the
expression.  Normally Mksquashfs does not pack tail-ends into fragments,
as it may affect I/O performance because it may produce more disk head
seeking.

But tail-end packing can increase compression.  Additionally with modern
solid state media, seeking is not such a major issue anymore.

5.2.5. no-tailend()
-------------------

Tell Mksquashfs not to use tail-end packing for the files matching the
exppression.

For obvious reasons this should be used in conjuction with the global
Mksquashfs option -always-use-fragments.  By default tail-ends are not
packed into fragments.

5.2.6 compressed()
------------------

Tell Mksquashfs to compress the fies matching the expression.

For obvious reasons this should be used in conjunction with the global
Mksquashfs options -noD and -noF.  File are by default compressed.

5.2.7 uncompressed()
--------------------

Tell Mksquashfs to not compress the files matching the expression.

This action obviously can be used to avoid compressing already compressed
files (XZ, GZIP etc.).

5.2.8 uid(uid or user)
----------------------

Set the ownership of the files matching the expression to uid (if arg1
is a number) or user (if arg1 is a string).

5.2.9 gid(gid or group)
-----------------------

Set the group of the files matching the expression to gid (if arg1
is a number) or group (if arg1 is a string).

5.2.10 guid(uid/user, gid/group)
--------------------------------

Set the uid/user and gid/group of the files matching the expression.

5.2.11 chmod(mode)
------------------

Mode can be octal, or symbolic.

If Mode is Octal, the permission bits are set to the octal value.

If Mode is Symbolic, permissions can be Set, Added or Removed.

The symbolic mode is of the format [ugoa]*[[+-=]PERMS]+
	PERMS = [rwxXst]+ or [ugo]
	and the above sequence can be repeated separated with commas.

A combination of the letters ugoa, specify which permission bits will
be affected, u means user, g means group, o means other, and a
means all or ugo.

The next letter is +, - or =.  The letter + means add to the existing
permission bits, - means remove the bits from the existing permission
bits, and = means set the permission bits.

The permission bits (PERMS) are a combination of [rwxXst] which
sets/adds/removes those bits for the specified ugoa combination. They
can alternatively be u, g or o, which takes the permission bits from the
user, group or other of the file respectively.

Examples:

1. chmod(u+r)

Adds the read permission to user.

2. chmod(ug+rw)

Adds the read and write permissions to both user and group.

3. chmod(u=rw,go=r)

Sets the permissions to rw-r--r--, which is eqivalent to

4. chmod(644)

5. cgmod(ug=o)

Sets the user and group permissions to the permissions for other.

5.3 Actions applied at second directory scan (stage 3)
------------------------------------------------------

5.3.1 prune()

The prune() action deletes the file or directory (and everything
underneath it) that matches the expression.  In that respect it is
identical to the exclude() action, except that it takes place in the
third stage, rather than the first stage.  There are a number of
reasons to have a prune() action in addition to an exclude()
action.

1. In the first stage Mksquashfs is building an in-memory representation
   of the filesystem to be compressed.  At that point some of the tests
   don't work because they rely on an in-memory representation having been
   built.

   So the following tests don't work in stage 1, and so they can't be
   used in the exclude() action.

	dircount()
	dircount_range()
	exists()
	absolute()
	readlink()
	eval()

   If you want to use these tests, you have to use the prune() action.

2. Many exclusion/pruning operations may only be easily applied after
   transformation actions have been applied in stages 1 & 2.

   For example, you may change the ownership and permissions of
   matching files in stage 2, and then want to delete files based on
   some criteria which relies on this having taken place.

5.4. Actions applied at third directory scan (stage 4)
------------------------------------------------------

5.4.1 empty(reason)

The empty() action deletes any directory which matches the expression,
and which is also empty for <reason>.  <reason> is one of "excluded",
"source" and "all".  If no argument is given, empty() defaults to "all".

The reason "excluded" means the directory has become empty due to
the exclude() or prune() actions or by exclusion on the command line.
The reason "source" means the directory was empty in the source filesystem.
The reason "all" means it is empty for either one of the above two reasons.

This action is often useful when exclusion has produced an empty
directory, or a hierarchy of directories which are empty but for a
sub-directory which is empty but for a sub-directory until an
empty directory is reached.

Example

1. Exclude any file which isn't a directory, and then clean-up
   any directories which are empty as a result.

	exclude@!type(d)
	empty(excluded)@true

This will produce an empty filesystem, unless there were some
directories that were originally empty.

Changing the empty action to

	exclude@!type(d)
	empty@true

Will produce an empty filesystem.

5.5 Actions performed at filesystem creation (stage 6)
------------------------------------------------------

5.5.1 xattrs-exclude(regex)

The xattrs-exclude action excludes any xattr names matching <regex>.  <regex> is
a POSIX regular expression, e.g. xattrs-exclude("^user.")  excludes xattrs from
the user namespace.

5.5.2 xattrs-include(regex)

The xattrs-include action includes any xattr names matching <regex>.  <regex> is
a POSIX regular expression, e.g. -xattrs-include("^user.") includes xattrs from
the user namespace.

5.5.3 xattrs-add(name=val)

The xattrs-add action adds the xattr <name> with contents <val>.  If an user
xattr it can be added to regular files and directories (see man 7 xattr).
Otherwise it can be added to all files.

The extended attribute value by default will be treated as binary (i.e. an
uninterpreted byte sequence), but it can be prefixed with 0s, where it will be
treated as base64 encoded, or prefixed with 0x, where it will be treated as
hexidecimal.

Obviously using base64 or hexidecimal allows values to be used that cannot be
entered on the command line such as non-printable characters etc.  But it
renders the string non-human readable.  To keep readability and to allow
non-printable characters to be entered, the 0t prefix is supported.  This
encoding is similar to binary encoding, except backslashes are specially
treated, and a backslash followed by three octal digits can be used to encode
any ASCII character, which obviously can be used to encode non-printable values.

The following four actions are equivalent

-xattrs-add("user.comment=hello world")
-xattrs-add("user.comment=0saGVsbG8gd29ybGQ=")
-xattrs-add("user.comment=0x68656c6c6f20776f726c64")
-xattrs-add("user.comment=0thello world")

Obviously in the above example there are no non-printable characters and so
the 0t prefixed string is identical to the first line.  The following three
actions are identical, but where the space has been replaced by the
non-printable NUL '\0' (null character).

-xattrs-add("user.comment=0thello\000world")
-xattrs-add("user.comment=0saGVsbG8Ad29ybGQ=")
-xattrs-add("user.comment=0x68656c6c6f00776f726c64")