The FCM 1 extract system is deprecated. The documentation for the current extract system can be found at FCM Make.
The extract system provides an interface between the revision control system (currently Subversion) and the build system. Where appropriate, it extracts code from the repository and other user-defined locations to a directory tree suitable for feeding into the build system. In this chapter, we shall use many examples to explain how to use the extract system. At the end of this chapter, you will be able to extract code from the local file system as well as from different branches of different repository URLs. You will also learn how to mirror code to an alternate destination. Finally, you will be given an introduction on how to specify configurations for the build system via the extract configuration file. (For further information on the build system, please see the next chapter The Build System.) The last section of the chapter tells you what you can do in the case when Subversion is not available.
To invoke the extract system, simply issue the command:
fcm extract
By default, the extract system searches for an extract configuration file ext.cfg in $PWD and then $PWD/cfg. If an extract configuration file is not found in these directories, the command fails with an error. If an extract configuration file is found, the system will use the configuration specified in the file to perform the current extract.
If the destination of the extract does not exist, the system performs a
new full extract to the destination. If a previous extract already exists at
the destination, the system performs an incremental extract, updating any
modifications if necessary. If a full (fresh) extract is required for
whatever reason, you can invoke the extract system using the -f
option, (i.e. the command becomes fcm extract -f). If you simply
want to remove all the items generated by a previous extract in the
destination, you can invoke the extract system using the --clean
option.
For further information on the extract command, please see FCM Command Reference > fcm extract.
The extract configuration file is the main user interface of the extract system. It is a line based text file. For a complete set of extract configuration file declarations, please refer to the Annex: Declarations in FCM extract configuration file.
A simple example of a basic extract configuration file is given below:
# Example 1
# ----------------------------------------------------------------------
cfg::type ext # line 1
cfg::version 1.0 # line 2
# line 3
dest $PWD # line 4
# line 5
repos::var::user $HOME/var # line 6
# line 7
expsrc::var::user code # line 8
The above demonstrates how to use the extract system to extract code from a local user directory. Here is an explanation of what each line does:
CFG::TYPE declares the type
of the configuration file. The value ext tells the system that
it is an extract configuration file.CFG::VERSION declares the
version of the extract configuration file. The current default is
1.0. Although it is not currently used, if we have to change
the format of the configuration file at a later stage, we shall be able to
use this number to determine whether we are reading a file with an older
format or one with a newer format.# is a comment, and is ignored by the interpreter.DEST declares the destination
root directory of this extract. The value $PWD expands to the
current working directory.REPOS::<pck>::<branch> declares the top level URL
or path of a repository. The package name of the repository is given by
<pck>. In our example, we choose var as the name of the
package. (You can choose any name you like, however, it is usually sensible
to use a package name that matches the name of the project or system you
are working with.) The branch name in the repository is given by
<branch>. (Again, you can choose any name you like, however, it is
usually sensible to use a name such as base, user
or something that matches your branch name.) In our example, the word
user is normally used to denote a local user directory. Hence
the statement declares that the repository path for the var
package in the user branch can be found at
$HOME/var.EXPSRC::<pck>::<branch> declares an
expandable source directory for the package <pck> in the
branch <branch>. In our example, the package name is
var, and the branch name is user. These match the
package and the branch names of the repository declaration in line 6. It
means that the source directory declaration is associated with the path
$HOME/var. The value of the declaration code is
therefore a sub-directory under $HOME/var. By declaring a
source directory using an EXPSRC label, the system
automatically searches for all sub-directories (recursively) under the
declared source directory.Invoking the extract system using the above configuration file will extract all sub-directories under $HOME/var/code to $PWD/src/var/code. Note: the extract system ignores symbolic links and hidden files, (i.e. file names beginning with a .). It will write a build configuration file to $PWD/cfg/bld.cfg. The configuration used for this extract will be written to the configuration file at $PWD/cfg/ext.cfg.
The next example demonstrates how to extract from a Subversion repository URL:
# Example 2
# ----------------------------------------------------------------------
cfg::type ext # line 1
cfg::version 1.0 # line 2
# line 3
dest $PWD # line 4
# line 5
repos::var::base svn://server/var/trunk # line 6
revision::var::base 1234 # line 7
# line 8
expsrc::var::base code # line 9
REVISION::<pck>::<branch> declares the revision of
the repository associated with the package <pck> in the branch
<branch>. The current line tells the extract system to use revision
1234 of svn://server/var/trunk. It is worth noting that the
declared revision must be a revision when the declared branch exists. The
actual revision used is the last changed revision of the declared one. If
the revision is not declared, the default is to use the last changed
revision at the HEAD of the branch.Invoking the extract system using the above configuration file will extract all sub-directories under svn://server/var/trunk/code to $PWD/src/var/code. It will write a build configuration file to $PWD/cfg/bld.cfg. The configuration used for this extract will be written to the configuration file at $PWD/cfg/ext.cfg.
So far, we have only declared source directories using the
EXPSRC statement, which stands for expandable source
directory. A source directory declared using this statement will
trigger the system to search recursively for any sub-directories under
the declared one. Any sub-directories containing regular source files
will be included in the extract. Symbolic links, hidden files and empty
directories (or those containing only symbolic links and/or hidden files)
are ignored.
If you do not want the system to search for sub-directories underneath
your declared source directory, you can declare your source directory
using the SRC statement. The SRC statement is
essentially the same as EXPSRC except that it does not
trigger the automatic recursive search for source directories. In fact,
the system implements the EXPSRC statement by expanding it
into a list of SRC statements.
The second field of a repository, revision or source directory declaration label is the name of the container package. It is a name selected by the user to identify the system or project he/she is working on. (Therefore, it is often sensible to choose an identifier that matches the name of the project or system.) The package name provides a unique namespace for a file container. Source directories are automatically arranged into sub-packages, using the names of the sub-directories as the names of the sub-packages. For example, the declaration at line 9 in example 2 will put the source directory in the var/code sub-package automatically.
Note that, in additional to slash /, double colon
:: and double underscore __ (internal only)
also act as delimiters for package names. Please avoid using them for
naming your files and directories.
You can declare a sub-package name explicitly in your source directory statement. For example, the following two lines are equivalent:
src::var::base code/VarMod_Surface src::var/code/VarMod_Surface::base code/VarMod_Surface
Explicit sub-package declaration should not be used normally, as it requires a lot more typing (although there are some situations where it can be useful, e.g. if you need to re-define the package name).
Currently, the extract system only supports non-space characters in the package name, as the space character is used as a delimiter between the declaration label and its value. If there are spaces in the path name to a file or directory, you should explicity re-define the package name of that path to a package name with no space using the above method. However, we recommend that only non-space characters are used for naming directories and files to make life simpler.
At the end of a successful extract, the configuration used by the current extract is written in cfg/ext.cfg under the extract destination root. This file is an expanded version of the original, with changes in the following declarations:
EXPSRC declarations are expanded into
SRC declarations.With this file, it should be possible for a later extract to re-create the current configuration even if the contents of the repository have changed. (This applies only to code stored in the repository.)
The next example demonstrates how to extract from a repository and mirror the code to an alternate location. It is essentially the same as example 2, except that it has three new lines to describe how the system can mirror the extracted code to an alternate location.
# Example 3 # ---------------------------------------------------------------------- cfg::type ext cfg::version 1.0 dest $PWD rdest::machine tx01 # line 6 rdest::logname frva # line 7 rdest /scratch/frva/extract/example3 # line 8 repos::var::base svn://server/var/trunk revision::var::base 1234 expsrc::var::base code
Here is an explanation of what each line does:
RDEST::MACHINE declares the target
machine to which the code will be mirrored. The example mirrors the code to
the machine named tx01.RDEST::LOGNAME declares the user name
of the target machine, to which the user has login access. If this is not
declared, the system uses the login name of the current user on the local
machine.RDEST declares the root directory of
the alternate destination, where the mirror version of the extract will be
sent.Invoking the extract system on the above configuration will trigger an extract similar to that given in example 2, but it will also attempt to mirror the contents at $PWD/src/var/code to /scratch/frva/extract/example3/src on the alternate destination. It will also mirror the expanded extract configuration file $PWD/cfg/ext.cfg to /scratch/frva/extract/example3/cfg/ext.cfg and $PWD/cfg/bld.cfg to /scratch/frva/extract/example3/cfg/bld.cfg. It is also worth noting that the content of the build configuration file will be slightly different, since it will include directory names appropriate for the alternate destination.
The extract system currently supports rdist and
rsync as its mirroring tool. The default is
rsync. To use rdist instead of
rsync, add the following line to your extract configuration
file:
rdest::mirror_cmd rdist
If rsync is used to mirror an extract, the system needs to
issue a separate remote shell command to create the container directory of
the mirror destination. The default is to issue a shell command in the
form ssh -n -oBatchMode=yes LOGNAME@MACHINE mkdir -p DEST.
The following declarations can be used to modify the command:
# Examples using the default settings: rdest::rsh_mkdir_rsh ssh rdest::rsh_mkdir_rshflags -n -oBatchMode=yes rdest::rsh_mkdir_mkdir mkdir rdest::rsh_mkdir_mkdirflags -p
In addition, the default rsync shell command is
rsync -a --exclude='.*' --delete-excluded --timeout=900 --rsh='ssh
-oBatchMode=yes' SOURCE DEST. The following declarations can be
used to modify the command:
# Examples using the default settings:
rdest::rsync rsync
rdest::rsyncflags -a --exclude='.*' --delete-excluded --timeout=900 \
--rsh='ssh -oBatchMode=yes'
So far, we have only extracted from a single location. The extract system is not much use if that is the only thing it can do. In fact, the extract system supports extract of multiple source directories from multiple branches in multiple repositories. The following configuration file is an example of how to extract from multiple repositories:
# Example 4 # ---------------------------------------------------------------------- cfg::type ext cfg::version 1.0 dest $PWD repos::var::base fcm:var_tr # line 6 repos::ops::base fcm:ops_tr # line 7 repos::gen::base fcm:gen_tr # line 8 revision::gen::base 2468 # line 10 expsrc::var::base src/code # line 12 expsrc::var::base src/scripts # line 13 expsrc::ops::base src/code # line 14 src::gen::base src/code/GenMod_Constants # line 15 src::gen::base src/code/GenMod_Control # line 16 src::gen::base src/code/GenMod_FortranIO # line 17 src::gen::base src/code/GenMod_GetEnv # line 18 src::gen::base src/code/GenMod_ModelIO # line 19 src::gen::base src/code/GenMod_ObsInfo # line 20 src::gen::base src/code/GenMod_Platform # line 21 src::gen::base src/code/GenMod_Reporting # line 22 src::gen::base src/code/GenMod_Trace # line 23 src::gen::base src/code/GenMod_UMConstants # line 24 src::gen::base src/code/GenMod_Utilities # line 25
Here is an explanation of what each line does:
We shall end up with a directory tree such as:
$PWD
|
|--- cfg
| |
| |--- bld.cfg
| |--- ext.cfg
|
|--- src
|
|--- gen
| |
| |--- code
| |
| |--- GenMod_Constants
| |--- GenMod_Control
| |--- GenMod_FortranIO
| |--- GenMod_GetEnv
| |--- GenMod_ModelIO
| |--- GenMod_ObsInfo
| |--- GenMod_Platform
| |--- GenMod_Reporting
| |--- GenMod_Trace
| |--- GenMod_UMConstants
| |--- GenMod_Utilities
|
|--- ops
| |
| |--- code
| |
| |--- ...
|
|--- var
|
|--- code
| |
| |--- ...
|
|--- scripts
|
|--- ...
As seen in the above example, if a revision number is not specified for a repository URL, it defaults to the last changed revision at the HEAD of the branch. The revision number can also be declared in other ways:
If a revision number is specified for a branch, the actual revision used by the extract system is the last changed revision of the branch, which may differ from the declared revision. While this behaviour is useful in most situations, some users may find it confusing to work with. It is possible to alter this behaviour so that extract will fail if the declared revision does not correspond to a changeset of the declared branch. Make the following declaration to switch on this checking:
revmatch true
We have so far dealt with a single branch in any package. The extract system can be used to combine changes from different branches of a package. An example is given below:
# Example 5 # ---------------------------------------------------------------------- cfg::type ext cfg::version 1.0 dest $PWD repos::var::base fcm:var_tr repos::ops::base fcm:ops_tr repos::gen::base fcm:gen_tr revision::gen::base 2468 expsrc::var::base src/code expsrc::var::base src/scripts expsrc::ops::base src/code src::gen::base src/code/GenMod_Constants src::gen::base src/code/GenMod_Control src::gen::base src/code/GenMod_FortranIO src::gen::base src/code/GenMod_GetEnv src::gen::base src/code/GenMod_ModelIO src::gen::base src/code/GenMod_ObsInfo src::gen::base src/code/GenMod_Platform src::gen::base src/code/GenMod_Reporting src::gen::base src/code/GenMod_Trace src::gen::base src/code/GenMod_UMConstants src::gen::base src/code/GenMod_Utilities repos::var::branch1 fcm:var_br/frva/r1234_new_stuff # line 27 repos::var::branch2 fcm:var_br/frva/r1516_bug_fix # line 28 repos::ops::branch1 fcm:ops_br/opsrc/r3188_good_stuff # line 29
The configuration file in example 5 is similar to that of example 4 except for the last three lines. Here is an explanation of what they do:
When we invoke the extract system, it will attempt to extract from the first declared branch of a package, if the last changed revision of the source directory is the same in all the branches. However, if the last changed revision of the source directory differs for different branches, the system will attempt to obtain an extract priority list for each source directory, using the following logic:
For the var package in the above example, let us assume that we have three source directory packages X, Y and Z under code, and their last changed revisions under base are 100. Let's say we have committed some changes to X and Z in the branch1 branch at revision 102, and other changes to Y and Z in the branch2 branch at revision 104, the extract priority lists for X, Y and Z will look like:
Once we have an extract priority list for a source directory, we can begin extracting source files in the source directory. The source directory of the base branch is extracted first, followed by that in the subsequent branches. If a source file in a subsequent branch has the same content as the that in the base branch, it is discarded. Otherwise, the following logic determines the branch to use:
fail, merge (default)
and override. If the conflict mode is fail, the
extract fails. If the conflict mode is merge, the system will
attempt to merge the changes using a tool such as diff3. The
result of the merge will be used to update the destination. The extract
fails only if there are unresolved conflicts in the merge. (In which case,
the conflict should be resolved using the version control system before
re-running the extract system.) If the conflict mode is
override, the change in the latest declared branch takes
precedence, and the changes in all other branches will be ignored. The
conflict mode can be changed using the CONFLICT declaration in
the extract configuration file. E.g:
conflict fail
Once the system has established which source files to use, it determines whether the destination file is out of date or not. The destination file is out of date if it does not exist or if its content differs from the version of the source file we are using. The system only updates the destination if it is considered to be out of date.
The extract system can also combine changes from branches in the Subversion repository and the local file system. This is demonstrated in the next example.
# Example 6 # ---------------------------------------------------------------------- cfg::type ext cfg::version 1.0 dest $PWD repos::var::base fcm:var_tr repos::ops::base fcm:ops_tr repos::gen::base fcm:gen_tr revision::gen::base 2468 expsrc::var::base src/code expsrc::var::base src/scripts expsrc::ops::base src/code src::gen::base src/code/GenMod_Constants src::gen::base src/code/GenMod_Control src::gen::base src/code/GenMod_FortranIO src::gen::base src/code/GenMod_GetEnv src::gen::base src/code/GenMod_ModelIO src::gen::base src/code/GenMod_ObsInfo src::gen::base src/code/GenMod_Platform src::gen::base src/code/GenMod_Reporting src::gen::base src/code/GenMod_Trace src::gen::base src/code/GenMod_UMConstants src::gen::base src/code/GenMod_Utilities repos::var::branch1 fcm:var_br/frva/r1234_new_stuff repos::var::branch2 fcm:var_br/frva/r1516_bug_fix repos::ops::branch1 fcm:ops_br/opsrc/r3188_good_stuff repos::var::user $HOME/var # line 31 repos::gen::user $HOME/gen # line 32
Example 6 is similar to example 5 except that it is also extracting from local directories. Here is an explanation of the lines:
INC declarations to include other extract
configuration files. For example, if the configuration file of example 5
is stored in the file $HOME/example5/ext.cfg, line 1 to 29
of example 6 can be replaced with an
INC declaration. Example 6 can then
be written as:
inc $HOME/example5/ext.cfg repos::var::user $HOME/var repos::gen::user $HOME/gen
Note: the INC declaration supports the special
environment variable $HERE. If this variable is already set in
the environment, it acts as a normal environment variable. However, if it
is not set, it will be expanded into the container directory of the
current extract configuration file. This feature is particularly useful
if you are including a hierarchy of extract configurations from files in
the same container directory in a repository.
All the examples above dealt with standalone extract, that is, the current extract is independent of any other extract. If a previous extract exists in another location, the extract system can inherit from this previous extract in your current extract. This works like a normal incremental extract, except that your extract will only contain the changes you have specified (compared with the inherited extract) instead of the full source directory tree. This type of incremental extract is useful in several ways. For instance:
The following example is based on example 4 and example 6. The assumption is that an extract has already been performed at the directory ~frva/var/vn22.0 based on the configuration file in example 4.
# Example 7 # ---------------------------------------------------------------------- cfg::type ext cfg::version 1.0 dest $PWD use ~frva/var/vn22.0 # line 6 repos::var::branch1 fcm:var_br/frva/r1234_new_stuff # line 8 repos::var::branch2 fcm:var_br/frva/r1516_bug_fix # line 9 repos::ops::branch1 fcm:ops_br/opsrc/r3188_good_stuff # line 10 repos::var::user $HOME/var # line 12 repos::gen::user $HOME/gen # line 13
Running the extract system using the above configuration will trigger an incremental extract, as if you are running an incremental extract having modified the configuration file in example 4 to that of example 6. The only difference is that the original extract using the example 4 configuration will be left untouched at ~frva/var/vn22.0, and the new extract will contain only the changes in the branches declared from line 8 to 13.
Note: extract inheritance allows you to add more branches to a package,
but you should not redefine the REPOS, REVISION,
EXPSRC or SRC declarations of a branch that is
already declared (and already extracted) in the inherited extract. Although
the system will not stop you from doing so, you may end up with an extract
that does not quite do what it is supposed to do. For example, if the
base branch in the foo package
(repos::foo::base) is already defined and extracted in an extract
you are inheriting from, you should not redefine any of the
*::foo::base declarations in your current extract. However, you are
free to add more branches for the same package with new labels (e.g.
repos::foo::b1), and indeed new packages that are not already
defined in the inherited extract (e.g. repos::bar::base).
If you are setting up an extract to be inherited, you do not have to perform a build. If you don't you will still gain the benefit of incremental file extract, but you will be performing a full build of the code.
It is worth bearing in mind that rdest::* settings are not inherited. If mirroring is required in the inheriting extract, it will require its own set of rdest::* declarations.
The system will, however, assume that a mirrored version of the inherited extract is available for inheritance from the mirrored destination of the current extract.
E.g.: Consider an extract at /path/to/inherited/ and an inheriting extract at /path/to/current/. If the former does not have a mirror, the latter should not have one either. If the former mirrors to machine@/path/to/inherited/mirror/ and the latter mirrors to machine@/path/to/current/mirror/, the system will assume that the subsequent build at machine@/path/to/current/mirror/ can inherit from the build at machine@/path/to/inherited/mirror/. This is illustrated below:
/path/to/current/ => at machine: /path/to/current/mirror/ use /path/to/inherited/ => at machine: use /path/to/inherited/mirror/
Configuration settings for feeding into the build system can be declared
through the extract configuration file using the BLD:: prefix.
Any line in an extract configuration containing a label with such a prefix
will be considered a build system variable. At the end of a successful
extract, the system strips out the BLD:: prefix before writing
these variables to the build configuration file. Some example entries are
given between line 17 and 22 in the following configuration file:
# Example 8 # ---------------------------------------------------------------------- cfg::type ext cfg::version 1.0 dest $PWD repos::var::base fcm:var_tr repos::ops::base fcm:ops_tr repos::gen::base fcm:gen_tr revision::gen::base 2468 expsrc::var::base src/code expsrc::var::base src/scripts expsrc::ops::base src/code src::gen::base src/code/GenMod_Constants src::gen::base src/code/GenMod_Control src::gen::base src/code/GenMod_FortranIO src::gen::base src/code/GenMod_GetEnv src::gen::base src/code/GenMod_ModelIO src::gen::base src/code/GenMod_ObsInfo src::gen::base src/code/GenMod_Platform src::gen::base src/code/GenMod_Reporting src::gen::base src/code/GenMod_Trace src::gen::base src/code/GenMod_UMConstants src::gen::base src/code/GenMod_Utilities bld::target VarProg_AnalysePF.exe # line 27 bld::tool::fc sxmpif90 # line 29 bld::tool::cc sxmpic++ # line 30 bld::tool::ld sxmpif90 # line 31
The above example is essentially the same as example 4, apart from the additional build configuration. The following is a simple explanation of what the lines represent: (For detail of the build system, please see the next chapter on The Build System.)
When you start using the extract system to define compiler flags for the build system, you may end up having to make a lot of long and repetitive declarations. In this case, you may want to define variables to replace the repetitive parts of the declarations.
Environment variables whose names contain only upper case latin
alphabets, numbers and underscores can be referenced in a declaration
value via the syntax $NAME or ${NAME}. For
example:
repos::um::base ${HOME}/svn-wc/um
bld::tool::fflags $MY_FFLAGS
You can define a user variable by making a declaration with a label
that begins with a percent sign %. The value of a user
variable remains in memory until the end of the current file is reached.
You can reference a user variable in a declaration value via the syntax
%NAME or %{NAME}. For example:
# Declare a variable %fred %fred -Cdebug -eC -Wf,-init heap=nan stack=nan bld::tool::fflags %fred # bld::tool::fflags -Cdebug -eC -Wf,-init heap=nan stack=nan bld::tool::fflags::foo %fred -f0 # bld::tool::fflags::foo -Cdebug -eC -Wf,-init heap=nan stack=nan -f0 bld::tool::fflags::bar -w %fred # bld::tool::fflags::bar -w -Cdebug -eC -Wf,-init heap=nan stack=nan
Further to this, each declaration results in an internal variable of the same name and you can also refer to any of these internal variables in the same way. So, the example given above could also be written as follows:
bld::tool::fflags -Cdebug -eC -Wf,-init heap=nan stack=nan bld::tool::fflags::foo %bld::tool::fflags -f0 bld::tool::fflags::bar -w %bld::tool::fflags
If you use a hierarchy of INC declarations or variables,
you may end up with a configuration file that is difficult to understand.
To help you with this, the extract system generates an as-parsed
configuration file at cfg/parsed_ext.cfg of the destination.
The content of the as-parsed configuration file is what the extract
system actually reads. It should contain everything in your original
extract configuration file, except that all INC
declarations, environment variables and user/internal variables are
expanded.
The amount of diagnostic messages generated by the extract system is
normally set to a level suitable for normal everyday operation. This is the
default diagnostic verbose level 1. If you want a minimum amount of
diagnostic messages, you should set the verbose level to 0. If you want more
diagnostic messages, you can set the verbose level to 2 or 3. You can modify
the verbose level in two ways. The first way is to set the environment
variable FCM_VERBOSE to the desired verbose level. The second way
is to invoke the extract system with the -v <level>
option. (If set, the command line option overrides the environment
variable.)
The following is a list of diagnostic output at each verbose level:
rdist is used to mirror the code, run the command
with the -q option.rdist is used to mirror the code, run the command
without the -q option.rdist is used to mirror the code, print the
distfile supplied to the command.rsync is used to mirror the code, invoke the
command with the -v option.The extract system can still be used if Subversion is not available. Clearly, you can only use local repositories. However, you can still do incremental extract, mirror an extract to an alternate location, or combine code from multiple local repositories.
If you are using Subversion but your server is down then clearly there is
little you can do. However, if you already have an extract then you can
re-run fcm extract as long as the extract configuration file
only refers to fixed revisions. If this is not the case then you can always
use the expanded extract configuration file which can be found in
cfg/ext.cfg under the extract destination root. This means that
you can continue to makes changes to local code and do incremental extracts
even whilst your Subversion server is down.