sccsfile - format of an SCCS history file
An SCCS file is an ASCII file consisting of six logical parts (SCCS v4 checksum, delta table, usernames, flags, comments, body) or eight logical parts (SCCS v6 checksum, delta table, usernames, v4 flags, v6 flags, v6 metadata, comments, body):
|checksum||Character sum computed on the rest of the file, used for error detection.|
|delta table||Log containing version info and statistics about each delta.|
|usernames||Login names and/or group IDs of users who may add deltas.|
|flags||Definitions of internal keywords and configuration for user defined behavior (SCCS v4).|
|v6 flags||Optional further internal definitions. SCCS v6 only.|
|v6 metadata||Optional global metadata. SCCS v6 only.|
|extensions||This is the area where future versions of SCCS will add extensions.|
|comments||An arbitrary descriptive information block about the file.|
|body||the Actual text lines intermixed with control lines. This body is also known as interleaved deltas or the weave.|
This document describes the SCCS v4 history file format that has been introduced in February 1977 as well as the SCCS v6 enhancements that have been introduced in September 2011.
Throughout an SCCS file there are lines which begin with the ASCII SOH (start of heading) character (octal 001). This character is hereafter referred to as the control character, and will be represented as ‘^A’. If a line described below is not depicted as beginning with the control character, it cannot do so and still be within SCCS file format.
Entries of the form ddddd represent a decimal number. In the statistics line from the delta table, this represents a five digit string (a number between 00000 and 99999). Other numbers are not artificially limited to a specific range. Serial numbers and SID number components may be any non negative number representable by a signed 32 bit integer.
The usable range on a specific machine is however limited by the available virtual memory. This implementation needs virtual memory that is approx. 100 times the highest serial number from the file. If a SCCS history file contains one million deltas, this implementation needs 100 MBytes of virtual memory to process the deltas.
The checksum is the first line of an SCCS file. The form of the line is:
The value of the checksum is the low 16 bits from the signed sum of all characters, except those contained in the first line. When reading files, values computed as the unsigned sum of all characters are accepted too. The ^Ah provides a 16 bit magic number of (octal) 064001 in PDP-byteorder and 000550 in Motorola-byteorder ("\001\150").
SCCS v6 Checksum
SCCS v6 introduces a different checksum line:
The checksum line was changed in order to prevent historic SCCS implementations from accidentally ignoring project-related (project-global) locks. It permits future versions to decide on different checksum algorithms without a need to again introduce a new history file format.
In SCCS v6, the ^Ah magic is not directly followed by the checksum but by the letter V that is followed by the SCCS version number. The version number is followed by a comma and the checksum algorithm name. The string sum is interpreted as the SCCS v4 checksum method. The SCCS v4 checksum entry must directly follow the SCCS version number.
Further checksums may be added in the future, e.g.:
but the sum= entry is mandatory. SCCS v6 currently does not implement checksum algorithms other than sum. Other entries are currently ignored when reading and silently discarded when copying or modifying files.
The delta table consists of a variable number of entries of the form:
^As inserted/deleted/unchanged ^Ad type sid yr/mo/da hr:mi:se username serial-number \ predecessor-sn ^Ai include-list ^Ax exclude-list ^Ag ignored-list ^Am mr-number ... ^Ac comments... ... ^Ae
The lines with the entry type ^As, ^Ad and ^Ae are mandatory.
The first line (^As) contains the number of lines inserted/deleted/unchanged respectively. The actual values for inserted/deleted/unchanged are five digit numbers. If an actual value is greater than 99999, then it is replaced by 99999.
The second line (^Ad) contains the type of the delta in the second word:
|D||for a normal delta|
|R||for a removed delta|
|U||for a delta that causes to unlink the g-file
If this delta type is used for the initial release of a SCCS history file, the default SID is set to 1.0 to permit the first delta with data to have the typical 1.1.
The delta table type is followed by the SCCS ID (SID) of the delta, the date and time of creation of the delta as local time, the user-name corresponding to the real user ID at the time the delta was created, and the serial numbers of the delta and its predecessor, respectively.
The year is either represented by a two digit year in the range 69..99 for 1969..1999 or 00..68 for 2000..2068 or by a four digit year number. Year numbers before 1969 are currently not supported. Older SCCS versions may not be able to understand four digit year numbers.
The user-name must not contain a space character, as space is the field separator in this file format. Since some non-POSIX platforms permit a space in the username, it is converted into a ‘_’ by the SCCS software before storing it in the history file.
The ^Ai, ^Ax, and ^Ag lines contain the serial numbers of deltas included, excluded, and ignored, respectively. These lines do not always appear.
The ^Am lines (optional) each contain one MR number associated with the delta.
The ^Ac lines contain comments associated with the delta. If there is more than one comment line, each comment line appears after a separate ^Ac lead in.
The ^Ae line ends the current delta table entry.
SCCS v6 Delta Table
SCCS v6 uses an enhanced delta table:
^As inserted/deleted/unchanged ^Ad type sid yr/mo/da hr:mi:se [.ss]+-hhmmusername\ serial-number predecessor-sn ^Ai include-list ^Ax exclude-list ^Ag ignored-list ^Am mr-number ... ^AS sid-specificmetadata ... ^Ac comments... ... ^Ae
The second line (^Ad) must have a four digit year number, may add sub-second time stamp granularity and must have a time zone offset.
Optional sub-second time stamp granularity is introduced by a dot ‘.’ and adds one to nine decimal digits that represent a fraction of a second up to nanosecond granularity. This number must be non-negative.
The time zone offset starts with a ‘+’ or a ‘-’, the value 0000 starts with a ‘+’, negative values start with a ‘-’. Positive values are east to GMT. The first two decimal digits represent the hour part of the GMT offset, the last two decimal digits represent the minute part of the GMT offset. A granularity less than a minute cannot be represented.
The date and time part represents local time as in SCCS v4 entries, but the mandatory timezone offset makes the time unique. The time stamp:
represents 2012, the first of February 12:00 GMT which is 13:00 MET.
The ^AS lines introduce SID specific SCCS v6 extensions. SID specific extension lines are in name/value format and take the form:
^AS name optional value text
The following name parameters are defined:
|ppath||Record the current path of the g-file relatively to the project’s file set home directory. A path entry only appears in case a rename or remove operation has been applied to the file.
The project set home is a directory that holds a directory .sccs for project specific SCCS metadata. The location of this directory .sccs is searched for by scanning the filesystem towards the root directory, starting from the current working directory. All files that belong to a project must be below the project’s file set home directory.
See also the description for the same keyword in the section for global meta data, where the the initial file name is recorded.
|snnnnn||A SID specific additional checksum. The value of the checksum is the low 16 bits from the unsigned sum of all characters in the SID specific file content as retrieved by ‘get -e’ (without keyword expansion). If the file is archived UU-encoded (see ‘e’ flag), the checksum is computed on the UU-encoded text and not on the original file.|
The data format in the extended SCCS delta entry (^Ad) and the SCCS SID specific metadata (^AS) is not accepted by historic SCCS implementations. When converting a SCCS v6 history file back to a SCCS v4 history file, these entries are converted into special comment at the beginning of the comment section. While converting, a copy of the unmodified ^Ad entry is kept as ^Ac_d and ^AS is turned into ^Ac_S.
The list of user-names and/or numerical group IDs of users who may add deltas to the file, separated by NEWLINE characters. The lines containing these login names and/or numerical group IDs are surrounded by the bracketing lines ^Au and ^AU. An empty list allows anyone to make a delta.
Flags are keywords that are used internally (see sccs-admin(1) for more information on their use). Each flag line takes the form:
^Af flag optional text
Flags may be selected from the set of 26 lower case characters in the range ‘a’..‘z’. Historical SCCS implementations will dump core in case a character outside the specified range appears as flag character.
The following flags are defined in order of appearance:
|^Af t type-of-program||Defines the replacement for the %Y% ID keyword.|
|^Af v program-name||Controls prompting for MR numbers in addition to comments. If the optional text is present, it defines an MR number validity checking program.
The v flag and the z flag are mutually exclusive.
|^Af i [value]||Indicates that the ‘No id keywords’ message is to generate an error that terminates the SCCS command. Otherwise, the message is treated as a warning only.
If the parameter value to the ‘i’ flag is not empty, then it holds a line fragment with keywords starting with a ‘%’, e.g.
‘%Z%%M% %I% %E%’
This line fragment needs to exactly match a part of a line in the file and to result in expanded keywords. Otherwise an attempt to check in a new delta will fail. The parameter to the ‘i’ flag is a SUN extension.
|^Af b||Indicates that the -b option may be used with the SCCS get command to create a branch in the delta tree.|
|^Af m module-name||Defines the first choice for the replacement text of the %M% ID keyword.|
|^Af f floor||Defines the "floor" release, that is, the release below which no deltas may be added.|
|^Af c ceiling||Defines the "ceiling" release, that is, the release above which no deltas may be added.|
|^Af d default-sid||The d flag defines the default SID to be used when none is specified on an SCCS get command.|
|^Af n||The n flag enables the SCCS delta command to insert a "null" delta (a delta that applies no changes) in those releases that are skipped when a delta is made in a new release (for example, when delta 5.1 is made after delta 2.7, releases 3 and 4 are skipped).|
|^Af j||Enables concurrent get -e calls for editing on the same base SID of an SCCS file. This allows multiple concurrent updates to take place on the same version of the SCCS file.|
|^Af l lock-releases||Defines a list of releases that are locked against editing.|
|^Af q user-defined||Defines the replacement for the %Q% ID keyword.|
|^Af e 0|1||The e flag indicates whether a source file is encoded or not. A 1 indicates that the file is encoded. Source files need to be encoded when they contain control characters, or when they do not end with a NEWLINE. The e flag allows files that contain binary data to be checked in.|
|^Af z application||The name of an application for the CMF enhancements. CMF enhancements are currently undocumented and it is not known how they are expected to work.
The v flag and the z flag are mutually exclusive.
|^Af s nnn||The number of lines to be scanned for keyword expansion.
This flag is a SUN extension that does not exist in historic sccs implementations.
|^Af x SCHILY|0||If the value of the ‘x’ flag is set to ‘SCHILY’, enable sccs extensions that are not implemented in classical sccs variants. If the ‘x’ flag is enabled, the keywords %d%, %e%, %g% and %h% are expanded even though not explicitly enabled by the ‘y’ flag.
This flag is a SCHILY extension that does not exist in historic sccs implementations.
This version of SCCS implements read only compatibility support for a SCO SCCS extension that sets the executable bit in the file permissions of a gotten file if the x-flag was set in the history file with no parameter. This version of SCCS does not allow to set this variant of the x-flag in the history file. If you like to get executable files from SCCS, set the executable bit in the file permissions of the history file.
If this version of SCCS is used to create the history file and the executable bit was set in the original file, SCCS automatically sets the executable bit in the history file and thus retains the executable bit in the gotten file.
|^Af y val...||The list of sccs keywords to be expanded. If the y-flag is missing in the sccs historyfile, all sccs keywords are expanded. If the flag is present but the list is empty, no keyword is expanded and no ‘No id keywords’ message is generated. The value ‘*’ controls the expansion of the %sccs.include.filename% keyword.
This flag is a SUN/SCHILY extension that does not exist in historic sccs implementations.
SCCS v6 Flags
SCCS v6 flags are additional keywords that are used internally (see sccs-admin(1) for more information on their use). Each flag line takes the form:
^AF flag optional text
No SCCS v6 flags are currently defined.
Historical SCCS implementations do not complain about SCCS v6 flags when reading SCCS history files and retain SCCS v6 flags when modifying history files. This is why SCCS v6 flags may be kept unmodified when converting a SCCS v6 history file back to a SCCS v4 history file.
SCCS v6 Metadata
SCCS v6 metadata is additional global metadata that is used internally (see sccs-admin(1) for more information on it’s use). Each metadata line takes the form:
^AG keyword optional text
The following keywords are defined:
|ppath||Record the initial path of the g-file relatively to the project’s file set home directory. Once set up, this name must never change.
See also the description for the project set home in the documentation for the same keyword in the section for SID specific meta data of the delta table. In case of a rename, the new file name is recorded in in the SID specific meta data of the delta table.
|rrandom||A unified pseudo random number (see also changeset(4)) that is used to help to identify single files across renames during the lifetime of the sccs history.
The pseudo random number is a hexadecimal string that represents the microseconds since Jul 13 11:01:20 2012 GMT when initially creating the sccs history for a specific file. Including microseconds gives sufficient randomness to make clashes rare.
With a 32 bit signed time_t, 52 bits in the pseudo random number are sufficient. With a 64 bit pseudo random number, more than 500000 years are covered.
The minimal length for the pseudo random number is thirteen hexadecimal characters. If the number could be represented with less digits, it is left filled with zeroes. This allows to have a unique length for this number until Mar 31 10:55:07 2155 GMT.
The random metadata is mandatory for SCCS v6 history files. The initial path tag may be recorded later but before the changeset file is created. The value for this metadata tags must not change.
Historical SCCS implementations do not complain about SCCS v6 metadata when reading SCCS history files and retain SCCS v6 metadata when modifying history files. This is why SCCS v6 metadata may be kept unmodified when converting a SCCS v6 history file back to a SCCS v4 history file.
This is the area where future versions of SCCS will add extensions.
SCCS reserves the area just before the comments section for extensions by only checking the content at this location for syntactic correctness. Unknown elements at this location are still copied and kept intact when the historyfile is modified. SCCS v6 already introduced SCCS v6 flags and global SCCS v6 metadata as extensions, so future extensions must appear past the SCCS v6 metadata.
Arbitrary text surrounded by the bracketing lines ^At and ^AT. The comments section typically will contain a description of the file’s purpose.
The body consists of text lines and control lines. Text lines do not begin with the control character, unless they are using SCCS v6 escape extensions. Control lines always begin with the control character.
There are three kinds of control lines: insert, delete, and end, represented by:
^AIddddd ^ADddddd ^AEddddd
respectively. The digit string is the serial number corresponding to the delta for the control line.
An inserted block of lines looks this way:
^AIddddd block of data ^AEddddd
A deleted block of lines looks this way:
^ADddddd block of data ^AEddddd
The block of data may contain control lines with other serial numbers.
SCCS v6 Body escape extensions
There are two additional types of text lines with a control character at the beginning that represent features introduced by SCCS v6. These features are not understood by SCCS implementations that support the SCCS v4 history file format only.
|^A^A||A line in the interleaved delta block that begins with two control characters represents a text line that begins at the second control character.|
|^AN||A line in the interleaved delta block that begins with the ^AN sequence represents a text line that does not end with a newline character. A line in the form ^AN^Atext is extracted as ^Atext without a need to add another quote.|
|‘‘g-file’’||version retrieved by get|
|d.file||temporary file created by the delta(1) program with the content of the previous version, used to compute differences to the current version|
|e.file||temporary file to hold an uuencoded version of the g-file in case of an encoded history file. Used by admin(1) and delta(1).|
|l.file||file containing extracted delta table info|
|o.file||saved original version of the history file, created by sccscvt(1).|
|p.file||permissions (lock) file for checked-out versions. The p.file has an arbitrary number of lines, each is related to a get -e call. Each line has the following structure:
The old-sid is the SID that was checked out with get -e, the new-sid is the SID that will be used for the new version when delta is called. The username is the user-name corresponding to the real user ID at the time get -e was called. The date and time fields are in the same format as used in the delta table of the s.file as described in sccsfile(4) for SCCS v4. In order to grant POSIX compatibility, a two digit year is used between 1969 and 2068. For years outside that range, a four digit year is used. The following fields are only present when one or more of the -i -x or -z options have been specified on the command line, they refer to the list of included and excluded deltas or to the CMR list from the NSE enhancements.
|q.file||temporary copy of the p.file; renamed to the p.file after completion|
|s.file||SCCS history file, see sccsfile(4).|
|SCCS/s.file||history file in SCCS subdirectory|
|x.file||temporary copy of the s.file; renamed to the s.file after completion. Used by admin(1), cdc(1), delta(1), rmdel(1), and sccscvt(1).|
|z.file||temporary lock file contains the binary process id in host byte order followed by the host name|
|dump.core||If the file dump.core exists in the current directory and a fatal signal is received, a coredump is initiated via abort(3).|
|If the environment SCCS_INCLUDEPATH does not exist, include files for the %sccs.include.filename% keyword directive are searched in this directory.|
The following SCCS programs write or modify SCCS history files:
|admin(1)||create and administer SCCS history files|
|cdc(1)||change the delta commentary of an SCCS delta|
|delta(1)||make a delta to an SCCS file|
|rmdel(1)||remove a delta from an SCCS file|
|sccscvt(1)||convert history files between SCCS v4 and SCCS v6|
SCCS v4 uses local time without a timezone offset. If a delta is created less than an hour after switching from DST to non-DST, it may look as if the clock is going backwards. SCCS v6 fixes this problem by introducing a mandatory timezone offset.
The SCCS suite was originally written by Marc J. Rochkind at Bell Labs in 1972. Release 4.0 of SCCS, introducing new versions of the programs admin(1), get(1), prt(1), and delta(1) was published on February 18, 1977; it introduced the new text based SCCS v4 history file format (previous SCCS releases used a binary history file format). The SCCS suite was later maintained by various people at AT&T and Sun Microsystems. Since 2006, the SCCS suite is maintained by J..org Schilling.