aepconf - project configuration file

projectCW/baseline/aegis.conf (default)

projectCW/baseline/config (obsolete)

A project configuration file is used to store information about a project. This file is under source control, and is one of the project's source files. Developers may thus modify this file as part of a change.

As of aegis.4.17, it is possible to assign any arbitrary name to the project configuration file or files. See aenf(1) for more information.

This file contains a number of commands to be executed by Aegis. There are times when the substitutions in these commands may contain shell special characters, which would change the meaning of the commands in unintended ways. There are two main sources of these problems: filenames and architecture names. In order to have shell special characters in filenames, you must set the shell_%safe_%filenames field (see below) to false. If you do this, you will need to use the CWquote substitution (see aesub(5)) to quote them, so that the shell does not abuse them. Other things which may need quoting include architecture names if you get creative, and edit numbers if unusual ones are generated by your history tool.

Getting Started

$ CB]aenf aegis.conf
$

$ CB]aecp aegis.conf
$

$ CB]aenf -config project.configuration
$

This file contains the following fields:

configuration_directory = string;

This field names a directory which will be searched for additional configuration files. (This directive is only legal or meaningful in the master project config file.)

All source files (change source files and project source files) present in this directory will be read in as if they were added to the end of the project "aegis.conf" file.

The usual priority of files (development directory, branch baseline, etc, project trunk baseline) is observed when these files are read.

Please note that the physical directories are never searched, only the Aegis concept of the change and project files is consulted (i.e. files created and modified in the usual way with aenf(1) and aecp(1) commands). Placing additional files in the physical directories will have no effect.

It is recommended that if you use this field at all, that your top level project aegis.conf file should only contain this one field. This is to avoid overly-large re-reading of this file when it is joined to all the others.

build_command = string;

This field describes how to build the project (actually, how to do an integration build). This field is mandatory. Used by the aeb(1) command. All of the substitutions described by aesub(5) are available.

Executed as: the integrator (for integration builds) or the developer (for development builds). Current directory: the integration directory of the change (for integration builds) the development directory of the change (for development builds). Exit status: zero is considered success, non-zero is a failure and a subsequent successful (exit zero) build will be required.

development_build_command = string;

This field describes how to do a development build. If this field is absent, it defaults to the above. Used by the aeb(1) command. All of the substitutions described by aesub(5) are available.

Executed as: the developer. Current directory: the development directory of the change. Exit status: zero is considered success, non-zero is a failure and a subsequent successful (exit zero) build will be required.

development_directory_style = { ... };

This field encapsulates a set of parameters controlling the appearance of the development directory. It has significant implications for the way the DMT is used, and the directory appearance presented to the DMT.

source_file_link = boolean;

This field is true if hard links are to be used for project source files (which are not part of the change) so that the work area has a complete set of source files.

Defaults to false if not set.

If the host system does not have hard links, this field will be ignored.

Maintaining the hard links can be time consuming for large projects, and add quite a noticeable delay before builds start doing anything. If possible, change your build system to use the $search_path substitution instead and avoid links.

source_file_symlink = boolean;

This field is true if symbolic links are to be used for project source files (which are not part of the change) so that the work area has a complete set of source files.

Defaults to false if not set. [If the obsolete create_%symlinks_%before_%build field is set, defaults to the value of that field, with a warning.]

If (source_file_link == true and hard links are available) this field will be ignored. If the host system does not have symbolic links, this field will be ignored.

Maintaining the symbolic links can be time consuming for large projects, and add quite a noticeable delay before builds start doing anything. If possible, change your build system to use the $search_path substitution instead and avoid symbolic links.

source_file_copy = boolean;

This field is true if copies are to be used for project source files (which are not part of the change) so that the work area has a complete set of source files. File modification time attributes will be preserved.

Defaults to false if not set.

If ((source_file_link == true and hard links are available) OR (source_file_symlink == true and symbolic links are available)) this field will be ignored.

Maintaining the copies can be time consuming (and space consuming) for large projects, and add quite a noticeable delay before builds start doing anything. If possible, change your build system to use the $search_path substitution instead and avoid file copies.

source_file_whiteout = boolean;

The source_file_whiteout field mat be used to sprecify the presence (true) or absence (false) of white-out files, used to "cover up" files being removed by a change set. These files contain 1KB of random data, intended to cause a syntax error should be build reference them.

It is rarely necessary to explicitly set this field. It defaults to false if you set any of the source_file_link, source_file_symlink or source_file_copy to true; it defaults to false only if none of them are true.

Not meaningful (always false) for integration builds.

derived_file_link = boolean;

This field is true if hard links are to be used for non-source files which are present in the project baseline(s) but which are not present in the work area, so that the work area has a complete set of derived files. This allows work areas to take advantage of "precompiled" object files (etc) in the baseline(s).

Defaults to false if not set.

If the host system does not have hard links, this field will be ignored.

Maintaining the links can be time consuming for large projects, and add quite a noticeable delay before builds start doing anything. If possible, change your build system to use the $search_path substitution instead and avoid hard links. Alternatively, set derived_at_start_only = true; and your work area will get a "head start" but the derived files will not be checked for every build, but this will occasionally result in long build times after integrations.

See also the integrate_%begin_%exceptions and symlink_%exceptions fields (they apply to hard links as well as symbolic links).

derived_file_symlink = boolean;

This field is true if symbolic links are to be used for non-source files which are present in the project baseline(s) but which are not present in the work area, so that the work area has a complete set of derived files. This allows work areas to take advantage of "precompiled" object files (etc) in the baseline(s).

Defaults to false if not set. [If the obsolete create_%symlinks_%before_%build field is set, defaults to the value of that field, with a warning.]

If (derived_file_link == true and hard links are available) this field will be ignored. If the host system does not have symbolic links, this field will be ignored.

Maintaining the symbolic links can be time consuming for large projects, and add quite a noticeable delay before builds start doing anything. If possible, change your build system to use the $search_path substitution instead and avoid symbolic links. Alternatively, set derived_at_start_only = true; and your work area will get a "head start" but the derived files will not be checked for every build, occasionally resulting in long build times after integrations.

See also the integrate_%begin_%exceptions and symlink_%exceptions fields.

derived_file_copy = boolean;

This field is true if copies are to be used for non-source files which are present in the project baseline(s) but which are not present in the work area, so that the work area has a complete set of derived files. This allows work areas to take advantage of "precompiled" object files (etc) in the baseline(s).

Defaults to false if not set.

If ((derived_file_link == true and hard links are available) or (derived_file_symlink == true and symbolic links are available)) this field will be ignored.

Maintaining the copies can be time consuming (and space consuming) for large projects, and add quite a noticeable delay before builds start doing anything. If possible, change your build system to use the $search_path substitution instead and avoid symbolic links. Alternatively, set derived_at_start_only = true; and your work area will get a "head start" but the derived files will not be checked for every build, occasionally resulting in long build times after integrations.

See also the integrate_begin_exceptions and symlink_exceptions fields (they apply to copies as well as symbolic links).

during_build_only = boolean;

This field is set to true if you want the symbolic links, hard links and/or copies removed again after each build. This allows the user to maintain the illusion of using a search path, without actually doing so. This option is not especially efficient.

Defaults to false if not set. [If the obsolete remove_%symlinks_%after_%build field is set, defaults to the value of that field, with a warning.]

If this field is false, the development directory will be populated by the develeop begin (aedb) command, and the integratin directory will be populated by the integrate begin (aeib) command.

derived_at_start_only = boolean;

This field controls whether the above fields controling the appearance of derived files are acted upon before every build (false) or only when the work area is created (true).

Defaults to false if not set.

This field is ignored if the during_build_only field is true.

This field can be complex. Here are a few examples; but much, much more is possible. The first example will get you a development directory very similar to one presented by CVS:

development_directory_style =
{
    source_file_copy = true;
};

development_directory_style =
{
    source_file_link = true;
    source_file_symlink = true;
    source_file_copy = true;
};

integration_directory_style = { ... };

This field encapsulates a set of parameters controlling the appearance of the integration directory. It has significant implications for the way the DMT is used, and the directory appearance presented to the DMT.

Defaults to the value of the development_directory_style field if not set. Note that the obsolete create_%symlinks_%before_%integration_%build and remove_%symlinks_%after_%integration_%build fields affect this default (with a warning) but only if they are explicitly set.

Note that the link_integration_directory field is still relevant. That field controls how the baseline is cloned to form the integration directory. This field operates after that operation.

build_time_adjust_notify_command = string;

This command is run when Aegis adjusts the last-time-modified time-stamp on files in the integration directory. If the build tool uses additional information to supplement file modification times, this command gives you the opportunity to re-sync the associated database.

Executed as: the project owner.

Current directory: the project baseline. Note that the contents of the project baseline reflect content before the currently integrating change. In particular note that for the first delta in a project or branch, the baseline directory is empty. If data is required from the change just completed one must use the ${INTegration_Directory} substitution.

Exit status: NOT ignored. Note that a failure here puts the change in a partial state from which recovery may be difficult. Best to define this command with a set+e so that errors are ignored at the command level.

build_covers_all_architectures = boolean;

This field is set to true if the build command, when executed on any architecture, results in all architectures being built. This may be accomplished, for example, by using cross-compilation techniques, or Cook's ability to nominate hosts on which to execute each build rule.

symlink_exceptions = [ string ];

This field is used to list filename patterns for which symbolic links must not be made between the development directory and the baseline. These are usually state files for various tools. The patterns are matched against the whole filename; naming only the last filename path element will not work (unless the pattern starts with ``CW]*'').

change_file_command = string;

This field contains a command to be executed whenever a ' -CoPy_file', ' -New_File' ' -New_Test' ' -MoVe_file' or ' -ReMove_file' command is successful. See also command-specific overrides. If this field is absent, nothing is done. Used by the aecp(1), aenv(1), aenf(1), aerm(1), and aemv(1) commands. All of the substitutions described by aesub(5) are available; in addition,

${File_List}

Space separated list of files named.

Executed as: the developer. Current directory: the development directory of the change. Exit status: ignored.

change_file_undo_command = string;

This field contains a command to be executed whenever a ' -CoPy_file_Undo', ' -MoVe_file_Undo' ' -New_File_Undo', ' -New_Test_Undo', or ' -ReMove_file_Undo' command is successful. Default to change_%file_%command if absent. See also command-specific overrides. If both fields are absent, nothing is done. Used by the aecpu(1), aemvu(1), aenfu(1), aentu(1) or aermu(1), commands. All of the substitutions described by aesub(5) are available; in addition,

${File_List}

Space separated list of files named.

Executed as: the developer. Current directory: the development directory of the change. Exit status: ignored.

new_file_command = string;

Executed whenever the aegis -new_file command is run successfully. Defaults to `change_file_command' if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_List}

Space separated list of files named (at times, can be empty).

Executed as: the developer. Current directory: the development directory of the change. Exit status: ignored.

new_test_command = string;

Executed whenever the aegis -new_test command is run successfully. Defaults to `change_file_command' if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_List}

Space separated list of files named (at times, can be empty).

Executed as: the developer. Current directory: the development directory of the change. Exit status: ignored.

copy_file_command = string;

Executed whenever the aegis -copy_file command is run successfully. Defaults to `change_file_command' if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_List}

Space separated list of files named (at times, can be empty).

Executed as: the developer. Current directory: the development directory of the change. Exit status: ignored.

remove_file_command = string;

Executed whenever the aegis -remove_file command is run successfully. Defaults to `change_file_command' if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_List}

Space separated list of files named (at times, can be empty).

Executed as: the developer. Current directory: the development directory of the change. Exit status: ignored.

new_file_undo_command = string;

Executed whenever the aegis -new_file_undo command is run successfully. Defaults to change_file_undo_command if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_List}

Space separated list of files named (at times, can be empty).

Executed as: the developer. Current directory: the development directory of the change. Exit status: ignored.

new_test_undo_command = string;

Executed whenever the aegis -new_test_undo command is run successfully. Defaults to change_file_undo_command if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_List}

Space separated list of files named (at times, can be empty).

Executed as: the developer Current directory: the development directory of the change Exit status: ignored

copy_file_undo_command = string;

Executed whenever the aegis -copy_file_undo command is run successfully. Defaults to change_file_undo_command if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_List}

Space separated list of files named (at times, can be empty).

Executed as: the developer Current directory: the development directory of the change Exit status: ignored

remove_file_undo_command = string;

Executed whenever the aegis -remove_file_undo command is run successfully. Defaults to change_file_undo_command if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_List}

Space separated list of files named (at times, can be empty).

Executed as: the developer Current directory: the development directory of the change Exit status: ignored

make_transparent_command = string;

The make_transparent_command is executed whenever the aegis -make_transparent command is run successfully. Defaults to change_file_command if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_List}

Space separated list of files named (at times, can be empty).

Executed as: the developer Current directory: the development directory of the change Exit status: ignored

make_transparent_undo_command = string;

The make_transparent_undo_command is executed whenever the aegis -make_transparent_undo command is run successfully. Defaults to change_file_undo_command if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_List}

Space separated list of files named (at times, can be empty).

Executed as: the developer Current directory: the development directory of the change Exit status: ignored

project_file_command = string;

This field contains a command to be executed during a development build before the development build command above, when (a) it is the first build after a develop begin, or (b) some other change has been integrated into the baseline since the last build. If this field is absent, nothing is done. Used by the aeb(1) command. All of the substitutions described by aesub(5) are available.

develop_begin_command = string;

This field contains a command to be executed whenever a ' -Develop_Begin' command is successful. If this field is absent, nothing is done. Used by the aedb(1) command. All of the substitutions described by aesub(5) are available.

Executed as: the developer. Current directory: the development directory of the change. Exit status: ignored.

develop_begin_undo_command = string;

This field contains a command to be executed whenever a ' -Develop_Begin_Undo' command is successful. If this field is absent, nothing is done. Used by the aedbu(1) command. All of the substitutions described by aesub(5) are available.

Executed as: the developer. Current directory: wherever the command was executed from. Exit status: ignored.

integrate_begin_command = string;

This field contains a command to be executed whenever a ' -Integrate_Begin' command is successful. If this field is absent, nothing is done. Used by the aeib(1) command. All of the substitutions described by aesub(5) are available.

Executed as: the project owner. Current directory: the integration directory. Exit status: ignored.

link_integration_directory = boolean;

This flag is true if Aegis should link the files from the baseline into the integration directory, rather than copy them (the default). This has risks, as the build script (e.g. Howto.cook or Makefile, etc) must unlink targets before rebuilding them; if this is not done the baseline will be corrupted. Used by the aeib(1) command.

integrate_begin_exceptions = [ string ];

This field may be used to specify a list of file names (and file name patterns) which are to be omitted from the copy (link) of the baseline when creating the integration directory. Used by the aeib(1) command. This field only applies to derived files, it does not apply to source files. The patterns are matched against the whole filename; naming only the last filename path element will not work (unless the pattern starts with ``CW]*'').

history_create_command = string;

This field is used to create a new history. The command is always executed as the project owner. Used by the aeipass(1) command.

It is strongly recommended that the history_%create_%command and history_%put_%command fields are identical. If not set, the history_%create_%command field defaults to the same value as the history_%put_%command field.

All of the substitutions described by aesub(5) are available; in addition,

${Input}

Absolute path of the source file.

${History}

Absolute path of the history file. This may need to be reworked with the Dirname and Basename substitutions to yield a string suitable for the history tool in question.

${File_Name}

The base relative file name of the file for this check-in. Note that the file name can vary over the lifetime of the file as it is renamed, but the history file name (above) never varies. Do not use this as the name of the history file. (Optional)

${UUID}

The universally unique identifier of the source file. This is invariant for the lifetime of the file. Do not use use this as the name of the history file. (Optional)

See also the history_put_trashes_file field, below.

Executed as: the project owner. Current directory: the base of the history tree. Exit status: zero indicates success, all non-zero exits indicate failure (the integrate pass will fail).

history_get_command = string;

This field is used to get a file from history. The command may be executed by developers. Used by the aeipass(1) and aecp(1) commands. All of the substitutions described by aesub(5) are available; in addition,

${History}

The absolute path of the history file. This may need to be reworked with the Dirname and Basename substitutions to yield a string suitable for the history tool in question.

${Edit}

The edit number to be extracted. It may be an arbitrary string, varying on the particular history tool.

${Output}

The absolute path of the destination file.

Executed as: the developer (or the executing user, in the case of the -independent option). Current directory: the base of the history tree Exit status: zero indicates success, all non-zero exits indicate failure (the aecp will fail).

history_put_command = string;

This field is used to add a new change to the history. The command is always executed as the project owner. Used by the aeipass(1) command.

It is strongly recommended that the history_%put_%command and history_%create__%command fields are identical. If not set, the history_%put_%command field defaults to the same value as the history_%create_%command field.

All of the substitutions described by aesub(5) are available; in addition,

${Input}

The absolute path of the source file.

${History}

The absolute path of the history file. This may need to be reworked with the Dirname and Basename substitutions to yield a string suitable for the history tool in question.

${File_Name}

The base relative file name of the file for this check-in. Note that the file name can vary over the lifetime of the file as it is renamed, but the history file name (above) never varies. Do not use this as the name of the history file. (Optional)

${UUID}

The universally unique identifier of the source file. This is invariant for the lifetime of the file. Do not use use this as the name of the history file. (Optional)

See also the history_put_trashes_file field, below.

Executed as: the project owner. Current directory: the base of the history tree. Exit status: zero indicates succes, all non-zero exits indicate failure (the integrate pass will fail).

history_transaction_begin_command = string;

The history_transaction_begin_command field is used to specify a command to be run by aeipass(1) before any history create or history put commands are run. The default is to do nothing.

All of the substitutions described in aesub(5) are available. If you need a transaction ID, use the $version substitution.

Executed as: the project owner. Current directory: the base of the history tree. Exit status: zero indicates succes, all non-zero exits indicate failure (the integrate pass will fail).

history_transaction_end_command = string;

The history_transaction_end_command field is used to specify a command to be run by aeipass(1) after any history create or history put commands are run, but before any history query commands are run. The default is to do nothing.

All of the substitutions described in aesub(5) are available. If you need a transaction ID, use the $version substitution.

Executed as: the project owner. Current directory: the base of the history tree. Exit status: zero indicates succes, all non-zero exits indicate failure (the integrate pass will fail).

history_transaction_abort_command = string;

The history_transaction_abort_command field is used to specify a command to be run by aeipass(1) to indicate that a transaction has been abandoned. The default is to do nothing.

All of the substitutions described in aesub(5) are available. If you need a transaction ID, use the $version substitution.

Executed as: the project owner. Current directory: the base of the history tree. Exit status: ignored (the integrate pass has already failed).

history_query_command = string;

This field is used to query the topmost edit of a history file. Result to be printed on the standard output. This command may be executed by developers. Used by the aeipass(1) and aecp(1) commands. All of the substitutions described by aesub(5) are available; in addition,

${History}

The absolute path of the history file. This may need to be reworked with the Dirname and Basename substitutions to yield a string suitable for the history tool in question.

Executed as: the project owner. Current directory: the base of the history tree. Exit status: zero indicates succes, all non-zero exits indicate failure (the integrate pass will fail).

history_label_command = string;

This field contains a command to be executed whenever a aeipass(1) or aedn(1) command is successful. This command is invoked for every file in the project. So using it incurs a performance penalty. If this field is absent, nothing is done. All of the substitutions described by aesub(5) are available; in addition,

${History}

The absolute path of the history file.

${Edit}

The edit number to be labeled. It may be an arbitrary string, varying on the particular history tool.

${Label}

The label to be attached to the history. When executed from aeipass(1) this value is the same as ${Version}, which may need to be reworked with the ${Subst} substitutions to yield a string suitable for the history tool in question. When executed from aedn(1) it is set to the value passed in from the command line.

Executed as: the project owner. Current directory: the base of the history tree. Exit status: zero indicates success, all non-zero exits indicate failure (a warning will be issued).

Labeling does not scale, so the use of this command is not encouraged. If you have a project with 10,000 files, and a change modified exactly one of them, only one history_%put_%command execution is required, which operates on one history file. If you have labeling turned on, it will also be necessary to execute 10,000 history_%label_%commands, to add information Aegis will never use.

history_put_trashes_file = (fatal, warn, ignore);

Many history tools (e.g. RCS) can modify the contents of the file when it is committed. While there are usually options to turn this off, they are seldom used. The problem is: if the commit changes the file, the source in the repository now no longer matches the object file in the repository - i.e. the history tool has compromised the referential integrity of the repository.

fatal

Emit a fatal error if one or more source files are modified by a history_put_command or history_create_command. This is the default.

warn

Emit a warning if a source file is modified.

ignore

Ignore a source file changing. You sure better hope it was only in a comment!

history_content_limitation = (ascii_text, international_text, binary_capable);

This field describes the content style which the history tool is capable of working with.

ascii_text

The history tool can only cope with files which contain printable ASCII characters, plus space, tab and newline. The file must end with a newline. This is the default.

international_text

The history tool can only cope with files which do not contain the NUL character. The file must end with a newline.

binary_capable

The history tool can cope with all files without any limitation on the form of the contents.

When a file is added to the history (by either the history_%create_%command or the history_%put_%command field) it is examined for conformance to this limitation. If there is a problem, the file is encoded in either quoted printable for MIME64, whichever is smaller, before being given to the history tool. This encoding is transparent, the file in the baseline is unchanged.

On extract (the history_get_command field) the encoding is reversed, using information attached to the change file information. This is because each put could use a different encoding (although in practice, file contents rarely change that dramatically, and the same encoding is likely to be deduced every time).

Please note that this field does not apply to the diff_%command or merge_%command fields.

diff_command = string;

This field is used to difference of 2 files. The command is always executed by developers. Used by the aed(1) command. All of the substitutions described by aesub(5) are available; in addition,

${ORiginal}

The absolute path of the original file copied into the change. Usually in the baseline, but not always.

${Input}

The absolute path of the file in the development directory.

${Output}

The absolute path of the file in which to write the difference listing.

Executed as: the project owner (for integration diffs), or the developer (for development diffs). Current directory: the integration directory (for integration diffs), or the development directory (for development diffs). Exit status: zero indicates success, all non-zero exits indicate failure (the aed will fail).

merge_command = string;

This field is used to merge two competing edits to a file. The command is always executed by developers. The current directory will be the development directory. This field is used by the aed(1) command. All of the substitutions described by aesub(5) are available; in addition,

${ORiginal}

The absolute path of the original file copied into the change. Usually not in the baseline, often a temporary file.

${Most_Recent}

The absolute path of the competing edit, usually in the baseline.

${Input}

The absolute path of the file in the development directory. This is the ``preferred'' edit, if the tool has this concept when highlighting conflicting edits.

${Output}

The absolute path of the file in which to write the merged result. This will usually be the name if a change source file in the development directory.

It is important that this command does not move files around. (See the obsolete diff3_command field, below, for some history.)

Executed as: the project owner (for integration diffs), or the developer (for development diffs). Current directory: the integration directory (for integration diffs), or the development directory (for development diffs). Exit status: zero indicates succes, all non-zero exits indicate failure (the aed will fail).

patch_diff_command = string;

The difference of 2 files, to send around as a patch. (This isn't the same as diff_command, because it's aimed at GNU Patch, not at humans.) The command is always executed by developers. Used by the aepatch(1) command.

Defaults to CW]"set +e; diff -c -L $index -L $index $original $input > $output; test $? -le 1" if not set.

All of the substitutions described by aesub(5) are available; in addition,

${ORiginal}

The absolute path of the original file copied into the change. Usually in the baseline, but not always.

${Input}

The absolute path of the file in the development directory.

${Output}

The absolute path of the file in which to write the difference listing.

${INDex}

The project-relative name of the file, for use when the file name is embedded in the output. (Optional.)

Executed as: the project owner (for integration diffs), or the developer (for development diffs). Current directory: the integration directory (for integration diffs), or the development directory (for development diffs). Exit status: zero indicates succes, all non-zero exits indicate failure (the aed will fail).

annotate_diff_command = string;

The difference of 2 files, for the use of the aeannotate(1) command. (This isn't the same as the diff_%command field, because it's aimed at aeannotate(1), not at humans.) The command is always executed by developers. Used by the aeannotate(1) command.

Extreme care should be taken if you are considering setting this field, otherwise the result reported by aeannotate(1) may bear little relation to reality. The most useful option is GNU diff's --ignore-all-space option, which will have the effect of ignoring the majority of indenting and code formatting changes. The --ignore-case option could also be useful for case insensitive languages such as FORTRAN or PL/1. Avoid options which would alter the number of lines, such as - -ignore-blank-lines or --context as these will produce misleading results.

Defaults to CW]"set +e; diff $option $original $input > $output; test $? -le 1" if not set.

All of the substitutions described by aesub(5) are available; in addition,

${ORiginal}

The absolute path of the original file copied into the change. Usually in the baseline, but not always.

${Input}

The absolute path of the file in the development directory.

${Output}

The absolute path of the file in which to write the difference listing.

${INDex}

The project-relative name of the file, for use when the file name is embedded in the output. (Optional.)

${OPTion}

Extra options to be passed to the diff command, as set by the aeannotate(1) -diff-option command line option. Use with extreme care.

Executed as: the project owner (for integration diffs), or the developer (for development diffs). Current directory: the integration directory (for integration diffs), or the development directory (for development diffs). Exit status: zero indicates succes, all non-zero exits indicate failure (the aed will fail).

review_policy_command = string;

This field is used to set the command to be executed by the aerpass(1) command. This command is useful in cases where the enterprise has determined that more than one review is necessary or that the reviewer must be senior to the developer, etc. Defaults to "CW]exit 0" if not set.

The exit status is examined. An zero exit status (success) means that the change will proceed to the awaiting integration state; a non-zero exit status (failure) means that the change requires further review state, and the develop_%end_%action is consulted to determine the appropriate state (awaiting_%review or being_%reviewed) for the change to move to.

All of the substitutions described by aesub(5) are available. Of particular interest are CW]${Change_Developer_List} and CW]${Change_Reviewer_List} for passing the specific staff involved with the change.

Executed as: the current reviewer. Current directory: the development directory. Exit status: zero indicates success, non-zero indicates failure.

For example, to have a script which is a project source file to be used to gate the code review process, a setting such as the following may be used: review_policy_command = "$sh ${source script/reviewpolicy.sh} " "-p $project -c $change " "-d ${developer_list} " "-r ${reviewer_list}" ; This is only one of many ways to implement a project specific review policy.

develop_end_policy_command = string;

This field is used to set the command to be executed by the aede(1) command. This command is useful in cases where the enterprise has determined that additional pre-conditions must be met (in addition to those already imposed by the saedegP(1) command) before a change may leave the being developed state. Defaults to "CW]exit 0" if not set.

The exit status is examined. An zero exit status (success) means that the change may leave to the being developed state; a non-zero exit status (failure) means that the change requires further development.

All of the substitutions described by aesub(5) are available.

Executed as: the developer. Current directory: the development directory. Exit status: zero indicates success, non-zero indicates failure.

test_command = string;

This field is used to set the command to be executed by the aet(1) command. Defaults to "$shell $file_name" if not set.

All of the substitutions described in aesub(5) are available. In addition:

${File_Name}

The absolute path of the test to be executed.

${Search_Path}

Colon separated list of directories to search for tests and test support files. (This is a normal aesub(5) substitution.)

${Search_Path_Executable}

Colon separated list of directories to search for executable files and executable support files. Usually it is the same as the above, except during an ``aet -bl'' command.

Note that tests are source files, and thus never have the execute bit set.

Executed as: the project owner (for integration tests) or the developer (for development tests), or the executing user (for -independent tests). Current directory: the integration directory (for integration tests), the development directory (for development tests), the project baseline (for -bl tests), or the current directory (for -independent tests). Exit status: zero indicates success, one indicates failure, anything else indicates "no result".

development_test_command = string;

This field is used to set the command to be executed by the aet(1) command when a change is in the being developed state. Defaults to be the same as the test_command field if not set.

Note: It is a significantly bad idea to make tests behave differently in being development and being integrated states; avoid this at all costs.

All of the substitutions described in aesub(5) are available. In addition:

${File_Name}

The absolute path of the test to be executed.

${File_Name}

The absolute path of the test to be executed.

${Search_Path}

Colon separated list of directories to search for tests and test support files. (This is a normal aesub(5) substitution.)

${Search_Path_Executable}

Colon separated list of directories to search for executable files and executable support files. Usually it is the same as the above, except during an ``aet -bl'' command.

Note that tests are source files, and thus never have the execute bit set.

Executed as: the developer. Current directory: the development directory (for development tests), the project baseline (for -bl tests). Exit status: zero indicates success, one indicates failure, anything else indicates "no result".

batch_test_command = string;

This field is used to set the command to be executed by the aet(1) command, in preference to the test_command or development_test_command, if set. It is capable of running more than one test at once.

All of the substitutions described in aesub(5) are available. In addition:

${Output}

This is the name of the file to be generated to hold the test results. See aetest(5) for the format of this file.

A space separated list of absolute paths of the tests to be executed.

${File_Names}

The absolute path of the tests to be executed.

${File_Name}

The absolute path of the test to be executed.

${Search_Path}

Colon separated list of directories to search for tests and test support files. (This is a normal aesub(5) substitution.)

${Search_Path_Executable}

Colon separated list of directories to search for executable files and executable support files. Usually it is the same as the above, except during an ``aet -bl'' command.

${Current}

Number of first test in the batch.

${Total}

Total number of tests. If this is 0 then no progress messages should be issued.

Note that tests are source files, and thus never have the execute bit set.

It is strongly recommended that you design your test scripts so that they may be executed by either batch or non-batch methods. This permits simple migration when your environment changes.

Executed as: the project owner (for integration tests) or the developer (for development tests), or the executing user (for -independent tests). Current directory: the integration directory (for integration tests), the development directory (for development tests), the project baseline (for -bl tests), or the current directory (for -indenpendent tests). Exit status: zero indicates succes, one indicates failure, anything else indicates "no result".

architecture_discriminator_command = string;

If this field is present it is used as a command to be executed in order to further identify the platform architecture (see below). All of the substitutions described by aesub(5) are available;

Executed as: the developer. Current directory: the development directory of the change. Exit status: zero indicates success, all non-zero exits indicate failure.

architecture = [{ ... }];

This field is a list of system and machine architectures on which each change must successfully build and test. May be assigned more than once. The structures listed have fields as follows:

name = string;

The name of the architecture. This name is available in the ${ARCHitecture} substitution (see aesub(5) for more information), as well as being used internally by Aegis. You may use almost any name for your architecture, but it is best to avoid shell special characters and white space, because it may be substituted into commands to be executed by Aegis.

pattern = string;

The system and machine architecture are determined by using the uname(2) system call. The uname(2) return value is assembled into a string of the form "sysname-release-version-machine", or "sysname-release-version-machine-disc" if architecture_discriminator_command is used.

The pattern field must match this uname result string. The first match found is used. The pattern is a shell file name pattern, see sh(1) for more information.

For example, the pattern SunOS-4.1*-*-sun4* matches a machine the author commonly uses, which returns SunOS-4.1.3-8-sun4m from the uname(2) system call.

mode = (required, optional, forbidden);

The mode field is used to control how the architecture information is used.

required

Architectures of thus mode will be copied into changes as their required architectures when the change is created. This is the default.

optional

Architectures of thus mode will not be copied into changes as their required architectures when the change is created. However, if you add them subsequently, they become required for that change.

forbidden

Aegis will refuse to build or test on architectures of this mode.

When a change is created, the required architecture names are copied into the change's architecture list. Once names are in this list, they are required for the change, and the project attributes are less relevant.

If the architecture field is not set, it defaults to

architecture =
[
	{
		name = "unspecified";
		pattern = "*";
		mode = required;
	}
];

file_template = [ { ... } ];

The file template is consulted whenever a new file is created, by one of the aenf(1) or aent(1) commands. May be assigned more than once. Each list item has the form:

pattern = [ string ];

The name of the file, relative to the development directory. Each string is a shell file name pattern; see sh(1) for more information. The patterns are matched against the whole filename; naming only the last filename path element will not work (unless the pattern starts with ``CW]*'').

body_command = string;

Command to run to initialize the body of the file.

Executed as: the developer. Current directory: the development directory of the change. Exit status: ignored.

body = string;

What to initialize the body of the file to.

All of the substitutions described in aesub(5) are available for the body and body_command strings. (Only specify one of them.) In addition:

${File_Name}

will be replaced by the name of the new file.

whiteout_template = [ { ... } ];

The file template is consulted whenever a file is removed, by one of the aerm(1) or aemv(1) commands. It is used to place a ``whiteout'' entry in the development directory, in order to induce compile errors of the removed file is referenced during the build. Each list item has the form:

pattern = [ string ];

The name of the file, relative to the development directory. Each string is a shell file name pattern; see sh(1) for more information. The patterns are matched against the whole filename; naming only the last filename path element will not work (unless the pattern starts with ``CW]*'').

body = string;

What to initialize the body of the file to. If not present, no whiteout file will be created; if the empty string, a zero-length whiteout file will be created.

All of the substitutions described in aesub(5) are available for the body string. In addition:

${File_Name}

will be replaced by the name of the removed file.

If the name of the file being removed does not match any of the filename patterns, a file consisting of 1KB of very ugly garbage will be generated. The idea is that it will produce a syntax error for most languages if you try to run it, compile it, or include it.

maximum_filename_length = integer;

This field is used to limit the length of filenames. All new files may not have path components longer than this. Existing files are not affected. The last component must also allow for the ",D" suffix of difference files. Where this value is larger than the file system allows, the file system limit will be imposed. Defaults to 255 if not set. Legal values range from 9 to 255.

The file name lengths of project files will be checked at develop end if the project aegis.conf file is in the change. See aede (1) for more information.

posix_filename_charset = boolean;

This field may be used to limit the characters allowed in filenames to only those explicitly allowed by POSIX. Defaults to false if not set.

For a filename to be portable across conforming implementations of IEEE Std 1003.1-1988, it shall consist only of alphanumeric characters, dot, hyphen or underscore. Hyphen shall not be used as the first character of a portable filename.

If this field is false, all characters are allowed except non-printing characters, space characters and leading hyphens.

dos_filename_required = boolean;

This field may be used to limit filenames so that they conform to the DOS 8+3 filename limits and to the DOS filename character set. Also denies filenames which look like devices (AUX, etc). Defaults to false if not set. This field is used in combination with the other filename fields, it does not replace them.

windows_filename_required = boolean;

This field may be used to limit filenames so that they conform to the Windows98 and WindowsNT filename limits and character set. Also denies filenames which look like devices (AUX, etc). Defaults to false if not set. This field is used in combination with the other filename fields, it does not replace them.

shell_safe_filenames = boolean;

This field may be used to limit filenames so that they may not contain shell special characters. If you do not set this to true, you will need to use the ${quote} substitution around filenames in commands, or risk unexpected errors. This field defaults to true if not set.

filename_pattern_accept = [ string ];

This field is used to specify a list of patterns of acceptable filenames. The patterns are matched against each filename path element. The patterns are constructed from the usual shell filename wildcards. Defaults to "*" if not set.

filename_pattern_reject = [ string ];

This field is used to specify a list of patterns of unacceptable filenames. The patterns are matched against each filename path element. The patterns are constructed from the usual shell filename wildcards. Defaults to "*,D" if not set. The pattern "*,D" is always appended. Where the filename_pattern_accept and filename_pattern_reject fields conflict, the reject takes precedence.

new_test_filename = string;

This field is used to form the filename of new tests, where the filename is not specified on the aent command line. Defaults to "test/${zpad $hundred 2}/t${zpad $number 4}${left $type 1}.sh" if not set.

All of the substitutions defined in aesub(5) are available. The following three substitutions are also available:

$Hundred

The test number divided by 100, optional

$Number

The test number, mandatory

$Type

The test type: "automatic" or "manual", optional

development_directory_template = string;

This field is used to determine the name of the development directory at develop begin. All of the substitutions defined in aesub(5) are available. The following substitutions is also available:

Default_Development_Directory

The directory within which the development directory is to be created.

Magic

A single letter, starting from ``C'', which can be inserted. This must be used, as it allows Aegis to try different names should there be a conflict.

If not set, defaults to "CW$ddd/${left $p ${expr ${namemax $ddd} - ${length .$magic$c}}}.$magic$c".

For DOS compatibility (8+3 filenames), a useful setting is "CW$ddd/${downcase ${left ${id $p} 8}.$magic${right 0$c 2}}". This ensures that the filename is always a valid 8.3 filename, that it is always lowercase, and it translates any punctuation in the project name into underscores.

metrics_filename_pattern = string;

This field is used to form the name of the metrics file, given a source file. All of the substitutions defined in aesub(5) are available. The following substitutions is also available:

File_Name

The absolute pathname of the source file.

Defaults to "$filename,S" if not set.

trojan_horse_suspect = [ string ];

This list of filename patterns is consulted by aedist --receive when it is checking for files which could be used to host Trojan horse attacks. This will be different for different projects, so you will need to update this yourself. The patterns are matched against the whole filename; naming only the last filename path element will not work (unless the pattern starts with ``CW]*'').

project_specific = [ { ... } ];

This is a list of name and value pairs for use within the ${project-specific} substitution (see aesub(5) for more information). May be assigned more than once. The sub-fields are

name = string;

The name of the value. By convention, names which start with an upper-case letter will appear in listings, and lower-case will not. Attribute names are case-insensitive.

value = string;

The value to be substituted.

There are almost no limitations on the strings which may appear in either of these fields.

When commands are executed by Aegis, it ensures that the AEGIS_PROJECT, AEGIS_CHANGE, AEGIS_ARCH, LINES and COLS environment variables are set appropriately. The project configuration file's project_%specific field is also consulted, looking for value's whose name starts with "setenv:" and sets the corresponding environment variable. All of the substitutions described by aesub(5) are available. For example: specifying a PATH and a SEARCH_PATH to be used for all commands may be set as follows:

project_specific =
[
  {
    name = "setenv:PATH";
    value = "/usr/bin:/bin";
  },
  {
    name = "setenv:SEARCH_PATH";
    value = "${search_path}";
  },
];

build_time_adjust = (...);

This field controls the adjustment of file modification times at the end of integrate-pass. File times are adjusted so that development directories are, in the main, out of date with respect to the baseline. The idea is that, at the very least, programs need to be relinked so that aet -reg does not give false negatives.

Combining this with the project_%file_%command (above) can alleviate the vast majority of file modification time inconsistencies experienced as a result of a project integration and the subsequent changes in the baseline's file modification times.

Unless you are a masochist, do not set this field. Leave it as the default.

adjust_and_sleep

Causes the file times to be adjusted, and if the file times would extend into the future, aeipass will sleep until that time has passed. This is the default.

adjust_only

Causes the file times to be adjusted. If the file time extend into the future, a warning is issued.

dont_adjust

File modification times are not adjusted. This is a really bad idea. Really. Make sure that, at the very minimum, project_%file_%command touches all of the change's files, otherwise the build problems which ensue are going to take you weeks to track down and lose you mucho productivity. You have been warned.

See also the build_%time_%adjust_%notify_%command field.

signed_off_by = boolean;

If this field is set each aedb(1), aechown(1), aede(1) and aerpass(1) will append a CW]Signed-off-by line to the change description. This field should only be set to true for open source projects.

For a description of CW]Signed-off-by see http://www.ussg.iu.edu/hypermail/linux/kernel/0405.2/1301.html and http://www.osdl.org/newsroom/press_releases/2004/2004_05_24_dco.html

Aegis has the ability to feed RSS channels when changesets transition states. See the User Guide for full details. Following is a brief description of the project-specific attributes used to control this process.

Create / Add to a channel

An RSS channel is specified with the rss:feedfilename project_specific attribute:

project_specific = [ { name = "rss:feedfilename-<filename>"; value = "<space-separated list of states>"; }

]

Specify the Description of an RSS channel

The description of an RSS channel is specified with the rss:feeddescription project_specific attribute:

project_specific = [ { name = "rss:feeddescription-<filename>"; value = "<description>"; }

]

Specify the Title of an RSS channel

The title of an RSS channel is specified with the rss:feedtitle project_specific attribute:

project_specific = [ { name = "rss:feedtitle-<filename>"; value = "<title>"; }

]

Specify the Language of an RSS channel

The language of an RSS channel is specified with the rss:feedlanguage project_specific attribute:

project_specific = [ { name = "rss:feedlanguage-<filename>"; value = "<language"; }

]

There are some obsolete fields in the file. They are provided for backwards compatibility only, and should not be used.

diff3_command = string;

This field is used to difference 3 files. The command is always executed by developers. Used by the aed(1) command. All of the substitutions described by aesub(5) are available; in addition,

${ORiginal}

The absolute path of the original file copied into the change. Usually not in the baseline.

${Most_Recent}

The absolute path of the competing edit, usually in the baseline.

${Input}

The absolute path of the file in the development directory.

${Output}

The absolute path of the file in which to write the difference listing.

Executed as: the project owner (for integration diffs), or the developer (for development diffs). Current directory: the integration directory (for integration diffs), or the development directory (for development diffs). Exit status: zero indicates success, all non-zero exits indicate failure (the aed will fail).

The problem with this field was that the default usage placed the merged source in a strange place. And subsequent aed(1) commands would over-write it. This meant that merges would be lost, causing a number of nasty problems. Some sites overcame this by adding ``mv'' commands to put the output back where the input came from, but this meant that Aegis' commentary was misleading. Use the ``merge_command'' field instead. It is almost identical, but Aegis will move the files around for you - so you get the good behavior by default (no lost merges) and the error message is consistent.

create_symlinks_before_build = boolean;

This flag is true if Aegis should create symlinks from the development directory to the baseline for all files in the baseline not in the development directory immediately before a development_build_command is issued. Usually used to trick dumb DMTs into believing the development directory contains an entire copy of the project, though sometimes the DMT is smart enough, the tools it must work with are not. Symlinks in the development directory which point to nonexistent files will be removed.

Defaults to false if not set.

create_symlinks_before_integration_build = boolean;

This flag is true if Aegis should create symlinks from the integration directory to the ancestral baseline for all files in the ancestral not in the integration directory immediately before a build_command is issued. Usually used to trick dumb DMTs into believing the integration directory contains an entire copy of the project, though sometimes the DMT is smart enough, the tools it must work with are not. Symlinks in the integration directory which point to nonexistent files will be removed.

Defaults to the same value as create_%symlinks_%before_%build if not set.

remove_symlinks_after_build = boolean;

This flag is true if Aegis should remove symlinks which point from the development directory to the baseline directory immediately after a development_%build_%command is issued. Only consulted if the create_%symlinks_%before_%build field is true, for the purpose of reversing the actions of the create_%symlinks_%before_%build field.

Defaults to false if not set.

remove_symlinks_after_integration_build = boolean;

This flag is true if Aegis should remove symlinks which point from the integration directory to the ancestral baseline directory immediately after a build_command is issued. Only consulted if the create_%symlinks_%before_%integration_%build field is true, for the purpose of reversing the actions of the create_%symlinks_%before_%integration_%build field.

Defaults to true if not set. This default is intentional. It is important that there are no symlinks in the (new) baseline, because they could go stale between integrations. If you set this field to false, caveat emptor.

cache_project_file_list_for_each_delta = boolean;

It is possible to have Aegis cache the list of project files that were present at integrate pass for each delta (integrated change set). This is used to optimize all project-history-based operations, such as aecp -delta or aepatch(1).

This cache will optimize many operations which would otherwise require time to reconstruct the prohect state using the roll-foworward data available in each change set. However, it comes at the cost of disk space, and not everyone can afford more and more disk.

This field defaults to true if not set.

aeb(1)

build a change

aecp(1)

copy a file into a change

aecpu(1)

reverse action of aecp

aed(1)

find differences between a change and the baseline

aede(1)

end development of a change

aerpass(1)

pass a review of a change

aeib(1)

begin integration of a change

aeipass(1)

pass integration of a change

aemv(1)

rename a file as part of a change

aenf(1)

add new files to be created by a change

aenfu(1)

remove new files from a change

aent(1)

add a new test to be created by a change

aentu(1)

remove new tests from a change

aet(1)

run tests

aegis(5)

file format syntax

aesub(5)

available command substitutions

aetest(5)

batch test results file

version

Copyright Peter Miller; All rights reserved.

The program comes with ABSOLUTELY NO WARRANTY; for details use the ' -VERSion License' command. This is free software and you are welcome to redistribute it under certain conditions; for details use the ' -VERSion License' command.

tab(;); l r l. Peter Miller;E-Mail:;millerp@canb.auug.org.au CW/\/\*;WWW:;http://www.canb.auug.org.au/~millerp/