/contrib/cvs/doc/cvs-paper.ms

   1.\" soelim cvs.ms | pic | tbl | troff -ms
   2.\" @(#)cvs.ms 1.2 92/01/30
   3.\"
   4.\"	troff source to the cvs USENIX article, Winter 1990, Washington, D.C.
   5.\"	Copyright (c) 1989, Brian Berliner
   6.\"
   7.\"	This program is free software; you can redistribute it and/or modify
   8.\"	it under the terms of the GNU General Public License as published by
   9.\"	the Free Software Foundation; either version 1, or (at your option)
  10.\"	any later version.
  11.\"
  12.\"	This program is distributed in the hope that it will be useful,
  13.\"	but WITHOUT ANY WARRANTY; without even the implied warranty of
  14.\"	MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15.\"	GNU General Public License for more details.
  16.\"
  17.\"	The author can be reached at:  berliner@prisma.com
  18.\"
  19.de SP
  20.if n .sp
  21.if t .sp .5
  22..
  23.de hl
  24.br
  25.in +0.5i
  26\l'\\n(LLu-1i'
  27.in -0.5i
  28.sp
  29..
  30.OH ""
  31.nr PS 11
  32.nr PO 1.25i
  33.pl -0.2i
  34.TL
  35.ps 14
  36.ft B
  37.nf
  38CVS II:
  39Parallelizing Software Development
  40.fi
  41.ft
  42.ps
  43.AU
  44.ps 12
  45.ft I
  46Brian Berliner
  47.ft
  48.ps
  49.AI
  50.ps 12
  51.ft I
  52Prisma, Inc.
  535465 Mark Dabling Blvd.
  54Colorado Springs, CO  80918
  55berliner@prisma.com
  56.ft
  57.ps
  58.AB
  59The program described in this paper fills a need in the UNIX
  60community for a freely available tool to manage software revision and
  61release control in a multi-developer, multi-directory, multi-group
  62environment.
  63This tool also addresses the increasing need for tracking third-party vendor
  64source distributions while trying to maintain local modifications to
  65earlier releases.
  66.AE
  67.NH
  68Background
  69.PP
  70In large software development projects, it is usually necessary for more
  71than one software developer to be modifying (usually different) modules of the
  72code at the same time.
  73Some of these code modifications are done in an
  74experimental sense, at least until the code functions correctly, and some
  75testing of the entire program is usually necessary.
  76Then, the modifications are returned to a master source repository
  77so that others in the project can
  78enjoy the new bug-fix or functionality.
  79In order to manage such a project, some sort of revision control system is
  80necessary.
  81.PP
  82Specifically, UNIX\**
  83.FS
  84UNIX is a registered trademark of AT&T.
  85.FE
  86kernel development is an excellent example of the
  87problems that an adequate revision control system must address.
  88The SunOS\**
  89.FS
  90SunOS is a trademark of Sun Microsystems, Inc.
  91.FE
  92kernel is composed of over a thousand files spread across a
  93hierarchy of dozens of directories.\**
  94.FS
  95Yes, the SunOS 4.0 kernel is composed of over a \fIthousand\fP files!
  96.FE
  97Pieces of the kernel must be edited
  98by many software developers within an organization.
  99While undesirable in
 100theory, it is not uncommon to have two or more people making
 101modifications to the same file within the kernel sources in
 102order to facilitate a desired change.
 103Existing revision control systems like
 104.SM
 105RCS
 106.LG
 107[Tichy] or
 108.SM
 109SCCS
 110.LG
 111[Bell] serialize file modifications by
 112allowing only one developer to have a writable copy of a particular file at
 113any one point in time.
 114That developer is said to
 115have \*Qlocked\*U the file for his exclusive use, and no other developer is
 116allowed to check out a writable copy of the file until the locking
 117developer has finished impeding others' productivity.
 118Development pressures of productivity and deadlines
 119often force organizations to require that multiple developers be able to
 120simultaneously edit
 121copies of the same revision controlled file.
 122.PP
 123The necessity for multiple developers to modify the same file concurrently
 124questions the value of serialization-based policies in traditional revision
 125control.
 126This paper discusses the approach that
 127Prisma took in adapting a standard revision control system,
 128.SM
 129RCS\c
 130.LG
 131, along with an existing public-domain collection of shell scripts that sits
 132atop
 133.SM
 134RCS
 135.LG
 136and provides the basic conflict-resolution algorithms.
 137The resulting
 138program, \fBcvs\fP, addresses not only the issue of conflict-resolution in
 139a multi-developer open-editing environment, but also the issues of
 140software release control and vendor source support and integration.
 141.NH
 142The CVS Program
 143.PP
 144\fBcvs\fP
 145(Concurrent Versions System)
 146is a front end to the
 147.SM
 148RCS
 149.LG
 150revision control system which extends
 151the notion of revision control from a collection of files in a single
 152directory to a hierarchical collection of directories each containing
 153revision controlled files.
 154Directories and files in the \fBcvs\fP system can be combined together in
 155many ways to form a software release.
 156\fBcvs\fP
 157provides the functions necessary to manage these software releases and to
 158control the concurrent editing of source files among multiple software
 159developers.
 160.PP
 161The six major features of \fBcvs\fP are listed below, and will be
 162described in more detail in the following sections:
 163.RS
 164.IP 1.
 165Concurrent access and conflict-resolution algorithms to guarantee that
 166source changes are not \*Qlost.\*U
 167.IP 2.
 168Support for tracking third-party vendor source distributions while
 169maintaining the local modifications made to those sources.
 170.IP 3.
 171A flexible module database that provides a symbolic mapping of names to
 172components of a larger software distribution.
 173This symbolic mapping provides for location independence within the software
 174release and, for example, allows one to check out a copy of the \*Qdiff\*U
 175program without ever knowing that the sources to \*Qdiff\*U actually reside
 176in the \*Qbin/diff\*U directory.
 177.IP 4.
 178Configurable logging support allows all \*Qcommitted\*U source file changes
 179to be logged using an arbitrary program to save the log messages in a file,
 180notesfile, or news database.
 181.IP 5.
 182A software release can be symbolically tagged and checked out at any time
 183based on that tag.
 184An exact copy of a previous software release can be checked out at
 185any time, \fIregardless\fP of whether files or directories have been
 186added/removed from the \*Qcurrent\*U software release.
 187As well,
 188a \*Qdate\*U can be used to check out the \fIexact\fP version of the software
 189release as of the specified date.
 190.IP 6.
 191A \*Qpatch\*U format file [Wall] can be produced between two software
 192releases, even if the releases span multiple directories.
 193.RE
 194.PP
 195The sources maintained by \fBcvs\fP are kept within a single directory
 196hierarchy known as the \*Qsource repository.\*U
 197This \*Qsource repository\*U holds the actual
 198.SM
 199RCS
 200.LG
 201\*Q,v\*U files directly, as well as a special per-repository directory
 202(\c
 203.SM
 204CVSROOT.adm\c
 205.LG
 206) which contains a small number of administrative files that describe the
 207repository and how it can be accessed.
 208See Figure 1 for a picture of the \fBcvs\fP tree.
 209.KF
 210.hl
 211.DS B
 212.PS
 213line from 4.112,9.200 to 5.550,8.887
 214line from 5.447,8.884 to 5.550,8.887 to 5.458,8.933
 215line from 4.112,9.200 to 4.550,8.950
 216line from 4.451,8.978 to 4.550,8.950 to 4.476,9.021
 217line from 4.112,9.200 to 3.737,8.887
 218line from 3.798,8.971 to 3.737,8.887 to 3.830,8.932
 219line from 3.612,8.762 to 4.737,8.137
 220line from 4.638,8.164 to 4.737,8.137 to 4.662,8.208
 221line from 3.612,8.762 to 3.737,8.137
 222line from 3.693,8.231 to 3.737,8.137 to 3.742,8.240
 223line from 3.612,8.762 to 2.612,8.200
 224line from 2.687,8.271 to 2.612,8.200 to 2.712,8.227
 225line from 2.362,9.262 to 2.737,8.950
 226line from 2.645,8.995 to 2.737,8.950 to 2.677,9.033
 227line from 2.362,9.262 to 1.925,8.950
 228line from 1.992,9.028 to 1.925,8.950 to 2.021,8.988
 229line from 3.362,9.762 to 4.050,9.387
 230line from 3.950,9.413 to 4.050,9.387 to 3.974,9.457
 231line from 3.362,9.762 to 2.487,9.387
 232line from 2.570,9.450 to 2.487,9.387 to 2.589,9.404
 233.ps 11
 234"newfs.c,v" at 4.487,8.043 ljust
 235.ps 11
 236"mkfs.c,v" at 3.487,8.043 ljust
 237.ps 11
 238"Makefile,v" at 2.237,8.043 ljust
 239.ps 11
 240"newfs" at 3.487,8.793 ljust
 241.ps 11
 242"halt.c,v" at 5.487,8.793 ljust
 243.ps 11
 244"Makefile,v" at 4.237,8.793 ljust
 245.ps 11
 246"modules,v" at 2.487,8.793 ljust
 247.ps 11
 248"loginfo,v" at 1.488,8.793 ljust
 249.ps 11
 250"etc" at 3.987,9.293 ljust
 251.ps 11
 252"CVSROOT.adm" at 1.988,9.293 ljust
 253.ps 11
 254"/src/master" at 2.987,9.793 ljust
 255.PE
 256.DE
 257.hl
 258.ce 100
 259.LG
 260\fBFigure 1.\fP
 261.SM
 262\fBcvs\fP Source Repository
 263.ce 0
 264.sp
 265.KE
 266.NH 2
 267Software Conflict Resolution\**
 268.FS
 269The basic conflict-resolution algorithms
 270used in the \fBcvs\fP program find their roots
 271in the original work done by Dick Grune at Vrije Universiteit in Amsterdam
 272and posted to \fBcomp.sources.unix\fP in the volume 6 release sometime in 1986.
 273This original version of \fBcvs\fP was a collection of shell scripts that
 274combined to form a front end to the
 275.SM
 276RCS
 277.LG
 278programs.
 279.FE
 280.PP
 281\fBcvs\fP allows several software developers to edit personal copies of a
 282revision controlled file concurrently.
 283The revision number of each checked out file is maintained independently
 284for each user, and \fBcvs\fP forces the checked out file to be current with
 285the \*Qhead\*U revision before it can be \*Qcommitted\*U as a permanent change.
 286A checked out file is brought up-to-date with the \*Qhead\*U revision using
 287the \*Qupdate\*U command of \fBcvs\fP.
 288This command compares the \*Qhead\*U revision number with that of the user's
 289file and performs an
 290.SM
 291RCS
 292.LG
 293merge operation if they are not the same.
 294The result of the merge is a file that contains the user's modifications
 295and those modifications that were \*Qcommitted\*U after the user
 296checked out his version of the file (as well as a backup copy of the
 297user's original file).
 298\fBcvs\fP points out any conflicts during the merge.
 299It is the user's responsibility to resolve these conflicts
 300and to \*Qcommit\*U his/her changes when ready.
 301.PP
 302Although the \fBcvs\fP conflict-resolution algorithm was defined in 1986,
 303it is remarkably similar to the \*QCopy-Modify-Merge\*U scenario included
 304with NSE\**
 305.FS
 306NSE is the Network Software Environment, a product of Sun Microsystems, Inc.
 307.FE
 308and described in [Honda] and [Courington].
 309The following explanation from [Honda] also applies to \fBcvs\fP:
 310.QP
 311Simply stated, a developer copies an object without locking it, modifies
 312the copy, and then merges the modified copy with the original.
 313This paradigm allows developers to work in isolation from one another since
 314changes are made to copies of objects.
 315Because locks are not used, development is not serialized and can proceed
 316in parallel.
 317Developers, however, must merge objects after the changes have been made.
 318In particular, a developer must resolve conflicts when the same object has
 319been modified by someone else.
 320.PP
 321In practice, Prisma has found that conflicts that occur when the same
 322object has been modified by someone else are quite rare.
 323When they do happen, the changes made by the other developer are usually
 324easily resolved.
 325This practical use has shown that the \*QCopy-Modify-Merge\*U paradigm is a
 326correct and useful one.
 327.NH 2
 328Tracking Third-Party Source Distributions
 329.PP
 330Currently, a large amount of software is based on source
 331distributions from a third-party distributor.
 332It is often the case that local modifications are to be made to this
 333distribution, \fIand\fP that the vendor's future releases should be
 334tracked.
 335Rolling your local modifications forward into the new vendor release is a
 336time-consuming task, but \fBcvs\fP can ease this burden somewhat.
 337The \fBcheckin\fP program of \fBcvs\fP initially sets up a source
 338repository by integrating the source modules directly from the vendor's
 339release, preserving the directory hierarchy of the vendor's distribution.
 340The branch support of
 341.SM
 342RCS
 343.LG
 344is used to build this vendor release as a branch of the main
 345.SM
 346RCS
 347.LG
 348trunk.
 349Figure 2 shows how the \*Qhead\*U tracks a sample vendor
 350branch when no local modifications have been made to the file.
 351.KF
 352.hl
 353.DS B
 354.PS
 355ellipse at 3.237,6.763 wid 1.000 ht 0.500
 356dashwid = 0.050i
 357line dashed from 3.237,7.513 to 3.737,7.513 to 3.737,9.762 to 4.237,9.762
 358line from 4.138,9.737 to 4.237,9.762 to 4.138,9.787
 359line dashed from 2.237,8.262 to 3.237,8.262 to 3.237,7.013
 360line from 3.212,7.112 to 3.237,7.013 to 3.262,7.112
 361line from 3.737,6.763 to 4.237,6.763
 362line from 4.138,6.737 to 4.237,6.763 to 4.138,6.788
 363line from 2.237,6.763 to 2.737,6.763
 364line from 2.637,6.737 to 2.737,6.763 to 2.637,6.788
 365line from 1.738,6.013 to 1.738,6.513
 366line from 1.762,6.413 to 1.738,6.513 to 1.713,6.413
 367line from 1.238,7.013 to 2.237,7.013 to 2.237,6.513 to 1.238,6.513 to 1.238,7.013
 368line from 4.237,9.012 to 5.237,9.012 to 5.237,8.512 to 4.237,8.512 to 4.237,9.012
 369line from 4.237,8.012 to 5.237,8.012 to 5.237,7.513 to 4.237,7.513 to 4.237,8.012
 370line from 4.237,7.013 to 5.237,7.013 to 5.237,6.513 to 4.237,6.513 to 4.237,7.013
 371line from 4.737,7.013 to 4.737,7.513
 372line from 4.763,7.413 to 4.737,7.513 to 4.712,7.413
 373line from 4.737,8.012 to 4.737,8.512
 374line from 4.763,8.412 to 4.737,8.512 to 4.712,8.412
 375line from 4.237,10.012 to 5.237,10.012 to 5.237,9.512 to 4.237,9.512 to 4.237,10.012
 376line from 4.737,9.012 to 4.737,9.512
 377line from 4.763,9.412 to 4.737,9.512 to 4.712,9.412
 378line from 5.987,5.013 to 5.987,6.013 to 0.988,6.013 to 0.988,5.013 to 5.987,5.013
 379.ps 11
 380"\"HEAD\"" at 1.550,8.231 ljust
 381.ps 11
 382"'SunOS'" at 2.987,6.293 ljust
 383.ps 11
 384"1.1.1" at 3.050,6.793 ljust
 385.ps 11
 386"1.1" at 1.613,6.793 ljust
 387.ps 11
 388"1.1.1.1" at 4.487,6.793 ljust
 389.ps 11
 390"1.1.1.2" at 4.487,7.793 ljust
 391.ps 11
 392"1.1.1.3" at 4.487,8.793 ljust
 393.ps 11
 394"1.1.1.4" at 4.487,9.793 ljust
 395.ps 11
 396"'SunOS_4_0'" at 5.487,6.793 ljust
 397.ps 11
 398"'SunOS_4_0_1'" at 5.487,7.793 ljust
 399.ps 11
 400"'YAPT_5_5C'" at 5.487,8.793 ljust
 401.ps 11
 402"'SunOS_4_0_3'" at 5.487,9.793 ljust
 403.ps 11
 404"rcsfile.c,v" at 2.987,5.543 ljust
 405.PE
 406.DE
 407.hl
 408.ce 100
 409.LG
 410\fBFigure 2.\fP
 411.SM
 412\fBcvs\fP Vendor Branch Example
 413.ce 0
 414.sp .3
 415.KE
 416Once this is done, developers can check out files and make local changes to
 417the vendor's source distribution.
 418These local changes form a new branch to the tree which is then used as the
 419source for future check outs.
 420Figure 3 shows how the \*Qhead\*U moves to the main
 421.SM
 422RCS
 423.LG
 424trunk when a local modification is made.
 425.KF
 426.hl
 427.DS B
 428.PS
 429ellipse at 3.237,6.763 wid 1.000 ht 0.500
 430dashwid = 0.050i
 431line dashed from 2.800,9.075 to 1.738,9.075 to 1.738,8.012
 432line from 1.713,8.112 to 1.738,8.012 to 1.762,8.112
 433line from 1.738,7.013 to 1.738,7.513
 434line from 1.762,7.413 to 1.738,7.513 to 1.713,7.413
 435line from 1.238,8.012 to 2.237,8.012 to 2.237,7.513 to 1.238,7.513 to 1.238,8.012
 436line from 3.737,6.763 to 4.237,6.763
 437line from 4.138,6.737 to 4.237,6.763 to 4.138,6.788
 438line from 2.237,6.763 to 2.737,6.763
 439line from 2.637,6.737 to 2.737,6.763 to 2.637,6.788
 440line from 1.738,6.013 to 1.738,6.513
 441line from 1.762,6.413 to 1.738,6.513 to 1.713,6.413
 442line from 1.238,7.013 to 2.237,7.013 to 2.237,6.513 to 1.238,6.513 to 1.238,7.013
 443line from 4.237,9.012 to 5.237,9.012 to 5.237,8.512 to 4.237,8.512 to 4.237,9.012
 444line from 4.237,8.012 to 5.237,8.012 to 5.237,7.513 to 4.237,7.513 to 4.237,8.012
 445line from 4.237,7.013 to 5.237,7.013 to 5.237,6.513 to 4.237,6.513 to 4.237,7.013
 446line from 4.737,7.013 to 4.737,7.513
 447line from 4.763,7.413 to 4.737,7.513 to 4.712,7.413
 448line from 4.737,8.012 to 4.737,8.512
 449line from 4.763,8.412 to 4.737,8.512 to 4.712,8.412
 450line from 4.237,10.012 to 5.237,10.012 to 5.237,9.512 to 4.237,9.512 to 4.237,10.012
 451line from 4.737,9.012 to 4.737,9.512
 452line from 4.763,9.412 to 4.737,9.512 to 4.712,9.412
 453line from 5.987,5.013 to 5.987,6.013 to 0.988,6.013 to 0.988,5.013 to 5.987,5.013
 454.ps 11
 455"1.2" at 1.613,7.793 ljust
 456.ps 11
 457"\"HEAD\"" at 2.862,9.043 ljust
 458.ps 11
 459"'SunOS'" at 2.987,6.293 ljust
 460.ps 11
 461"1.1.1" at 3.050,6.793 ljust
 462.ps 11
 463"1.1" at 1.613,6.793 ljust
 464.ps 11
 465"1.1.1.1" at 4.487,6.793 ljust
 466.ps 11
 467"1.1.1.2" at 4.487,7.793 ljust
 468.ps 11
 469"1.1.1.3" at 4.487,8.793 ljust
 470.ps 11
 471"1.1.1.4" at 4.487,9.793 ljust
 472.ps 11
 473"'SunOS_4_0'" at 5.487,6.793 ljust
 474.ps 11
 475"'SunOS_4_0_1'" at 5.487,7.793 ljust
 476.ps 11
 477"'YAPT_5_5C'" at 5.487,8.793 ljust
 478.ps 11
 479"'SunOS_4_0_3'" at 5.487,9.793 ljust
 480.ps 11
 481"rcsfile.c,v" at 2.987,5.543 ljust
 482.PE
 483.DE
 484.hl
 485.ce 100
 486.LG
 487\fBFigure 3.\fP
 488.SM
 489\fBcvs\fP Local Modification to Vendor Branch
 490.ce 0
 491.sp
 492.KE
 493.PP
 494When a new version of the vendor's source distribution arrives, the
 495\fBcheckin\fP program adds the new and changed vendor's files to the
 496already existing source repository.
 497For files that have not been changed locally, the new file from the
 498vendor becomes the current \*Qhead\*U revision.
 499For files that have been modified locally, \fBcheckin\fP warns that the
 500file must be merged with the new vendor release.
 501The \fBcvs\fP \*Qjoin\*U command is a useful tool that aids this process by
 502performing the necessary
 503.SM
 504RCS
 505.LG
 506merge, as is done above when performing an \*Qupdate.\*U
 507.PP
 508There is also limited support for \*Qdual\*U derivations for source files.
 509See Figure 4 for a sample dual-derived file.
 510.KF
 511.hl
 512.DS B
 513.PS
 514ellipse at 2.337,8.575 wid 0.700 ht 0.375
 515ellipse at 2.312,9.137 wid 0.700 ht 0.375
 516line from 1.225,9.012 to 1.225,9.363
 517line from 1.250,9.263 to 1.225,9.363 to 1.200,9.263
 518line from 0.875,9.725 to 1.600,9.725 to 1.600,9.363 to 0.875,9.363 to 0.875,9.725
 519line from 0.875,9.012 to 1.600,9.012 to 1.600,8.650 to 0.875,8.650 to 0.875,9.012
 520line from 4.050,10.200 to 4.775,10.200 to 4.775,9.850 to 4.050,9.850 to 4.050,10.200
 521line from 4.050,9.475 to 4.775,9.475 to 4.775,9.113 to 4.050,9.113 to 4.050,9.475
 522line from 4.050,8.762 to 4.775,8.762 to 4.775,8.400 to 4.050,8.400 to 4.050,8.762
 523line from 4.425,8.762 to 4.425,9.113
 524line from 4.450,9.013 to 4.425,9.113 to 4.400,9.013
 525line from 4.425,9.475 to 4.425,9.850
 526line from 4.450,9.750 to 4.425,9.850 to 4.400,9.750
 527line from 3.050,10.000 to 3.775,10.000 to 3.775,9.637 to 3.050,9.637 to 3.050,10.000
 528line from 3.050,9.312 to 3.775,9.312 to 3.775,8.950 to 3.050,8.950 to 3.050,9.312
 529line from 0.713,7.325 to 0.713,8.075 to 4.925,8.075 to 4.925,7.325 to 0.713,7.325
 530line from 1.238,8.075 to 1.238,8.637
 531line from 1.262,8.537 to 1.238,8.637 to 1.213,8.537
 532line from 1.613,8.825 to 1.975,8.575
 533line from 1.878,8.611 to 1.975,8.575 to 1.907,8.652
 534line from 2.675,8.575 to 4.050,8.575
 535line from 3.950,8.550 to 4.050,8.575 to 3.950,8.600
 536line from 2.675,9.137 to 3.050,9.137
 537line from 2.950,9.112 to 3.050,9.137 to 2.950,9.162
 538line from 3.425,9.325 to 3.425,9.637
 539line from 3.450,9.537 to 3.425,9.637 to 3.400,9.537
 540line from 1.613,8.825 to 1.925,9.137
 541line from 1.872,9.049 to 1.925,9.137 to 1.837,9.084
 542.ps 11
 543"'BSD'" at 2.138,9.481 ljust
 544.ps 11
 545"1.2" at 1.113,9.543 ljust
 546.ps 11
 547"1.1" at 1.125,8.831 ljust
 548.ps 11
 549"1.1.1.1" at 4.175,8.543 ljust
 550.ps 11
 551"1.1.1.2" at 4.175,9.281 ljust
 552.ps 11
 553"1.1.1.3" at 4.175,9.993 ljust
 554.ps 11
 555"1.1.2.2" at 3.175,9.793 ljust
 556.ps 11
 557"1.1.2.1" at 3.175,9.106 ljust
 558.ps 11
 559"rcsfile.c,v" at 2.425,7.706 ljust
 560.ps 11
 561"1.1.1" at 2.175,8.568 ljust
 562.ps 11
 563"'SunOS'" at 2.125,8.243 ljust
 564.ps 11
 565"1.1.2" at 2.163,9.131 ljust
 566.PE
 567.DE
 568.hl
 569.ce 100
 570.LG
 571\fBFigure 4.\fP
 572.SM
 573\fBcvs\fP Support For \*QDual\*U Derivations
 574.ce 0
 575.sp
 576.KE
 577This example tracks the SunOS distribution but includes major changes from
 578Berkeley.
 579These BSD files are saved directly in the
 580.SM
 581RCS
 582.LG
 583file off a new branch.
 584.NH 2
 585Location Independent Module Database
 586.PP
 587\fBcvs\fP contains support for a simple, yet powerful, \*Qmodule\*U database.
 588For reasons of efficiency, this database is stored in \fBndbm\fP\|(3) format.
 589The module database is used to apply names to collections of directories
 590and files as a matter of convenience for checking out pieces of a large
 591software distribution.
 592The database records the physical location of the sources as a form of
 593information hiding, allowing one to check out whole directory hierarchies
 594or individual files without regard for their actual location within the
 595global source distribution.
 596.PP
 597Consider the following small sample of a module database, which must be
 598tailored manually to each specific source repository environment:
 599.DS
 600\f(CW		#key      [-option argument] directory [files...]
 601		diff      bin/diff
 602		libc      lib/libc
 603		sys       -o sys/tools/make_links sys
 604		modules   -i mkmodules CVSROOT.adm modules
 605		kernel    -a sys lang/adb
 606		ps        bin Makefile ps.c\fP
 607.DE
 608.PP
 609The \*Qdiff\*U and \*Qlibc\*U modules refer to whole directory hierarchies that
 610are extracted on check out.
 611The \*Qsys\*U module extracts the \*Qsys\*U hierarchy, and runs the
 612\*Qmake_links\*U program at the end of the check out process (the \fI-o\fP
 613option specifies a program to run on check\fIo\fPut).
 614The \*Qmodules\*U module allows one to edit the module database file and
 615runs the \*Qmkmodules\*U program on check\fIi\fPn to regenerate the
 616\fBndbm\fP database that \fBcvs\fP uses.
 617The \*Qkernel\*U module is an alias (as the \fI-a\fP option specifies)
 618which causes the remaining arguments after the \fI-a\fP to be interpreted
 619exactly as if they had been specified on the command line.
 620This is useful for objects that require shared pieces of code from far away
 621places to be compiled (as is the case with the kernel debugger, \fBkadb\fP,
 622which shares code with the standard \fBadb\fP debugger).
 623The \*Qps\*U module shows that the source for \*Qps\*U lives in the \*Qbin\*U
 624directory, but only \fIMakefile\fP and \fIps.c\fP are required to build the
 625object.
 626.PP
 627The module database at Prisma is now populated for the entire UNIX
 628distribution and thereby allows us to issue the
 629following convenient commands to check out components of the UNIX
 630distribution without regard for their actual location within the master source
 631repository:
 632.DS
 633\f(CW		example% cvs checkout diff
 634		example% cvs checkout libc ps
 635		example% cd diff; make\fP
 636.DE
 637.PP
 638In building the module database file, it is quite possible to have name
 639conflicts within a global software distribution.
 640For example, SunOS provides two \fBcat\fP programs:
 641one for the standard environment, \fI/bin/cat\fP, and one for the System V
 642environment, \fI/usr/5bin/cat\fP.
 643We resolved this conflict by naming the standard \fBcat\fP module
 644\*Qcat\*U, and the System V \fBcat\fP module \*Q5cat\*U.
 645Similar name modifications must be applied to other conflicting names, as
 646might be found between a utility program and a library function, though
 647Prisma chose not to include individual library functions within the module
 648database at this time.
 649.NH 2
 650Configurable Logging Support
 651.PP
 652The \fBcvs\fP \*Qcommit\*U command is used to make a permanent change to the
 653master source repository (where the
 654.SM
 655RCS
 656.LG
 657\*Q,v\*U files live).
 658Whenever a \*Qcommit\*U is done, the log message for the change is carefully
 659logged by an arbitrary program (in a file, notesfile, news database, or
 660mail).
 661For example, a collection of these updates can be used to produce release
 662notices.
 663\fBcvs\fP can be configured to send log updates through one or more filter
 664programs, based on a regular expression match on the directory that is
 665being changed.
 666This allows multiple related or unrelated projects to exist within a single
 667\fBcvs\fP source repository tree, with each different project sending its
 668\*Qcommit\*U reports to a unique log device.
 669.PP
 670A sample logging configuration file might look as follows:
 671.DS
 672\f(CW	#regex      filter-program
 673	DEFAULT     /usr/local/bin/nfpipe -t %s utils.updates
 674	^diag       /usr/local/bin/nfpipe -t %s diag.updates
 675	^local      /usr/local/bin/nfpipe -t %s local.updates
 676	^perf       /usr/local/bin/nfpipe -t %s perf.updates
 677	^sys        /usr/local/bin/nfpipe -t %s kernel.updates\fP
 678.DE
 679.PP
 680This sample allows the diagnostics and performance groups to
 681share the same source repository with the kernel and utilities groups.
 682Changes that they make are sent directly to their own notesfile [Essick]
 683through the \*Qnfpipe\*U program.
 684A sufficiently simple title is substituted for the \*Q%s\*U argument before
 685the filter program is executed.
 686This logging configuration file is tailored manually to each specific
 687source repository environment.
 688.NH 2
 689Tagged Releases and Dates
 690.PP
 691Any release can be given a symbolic tag name that is stored directly in the
 692.SM
 693RCS
 694.LG
 695files.
 696This tag can be used at any time to get an exact copy of any previous
 697release.
 698With equal ease, one can also extract an exact copy of the source files as
 699of any arbitrary date in the past as well.
 700Thus, all that's required to tag the current kernel, and to tag the kernel
 701as of the Fourth of July is:
 702.DS
 703\f(CW	example% cvs tag TEST_KERNEL kernel
 704	example% cvs tag -D 'July 4' PATRIOTIC_KERNEL kernel\fP
 705.DE
 706The following command would retrieve an exact copy of the test kernel at
 707some later date:
 708.DS
 709\f(CW	example% cvs checkout -fp -rTEST_KERNEL kernel\fP
 710.DE
 711The \fI-f\fP option causes only files that match the specified tag to be
 712extracted, while the \fI-p\fP option automatically prunes empty directories.
 713Consequently, directories added to the kernel after the test kernel was
 714tagged are not included in the newly extracted copy of the test kernel.
 715.PP
 716The \fBcvs\fP date support has exactly the same interface as that provided
 717with
 718.SM
 719RCS\c
 720.LG
 721, however \fBcvs\fP must process the \*Q,v\*U files directly due to the
 722special handling required by the vendor branch support.
 723The standard
 724.SM
 725RCS
 726.LG
 727date handling only processes one branch (or the main trunk) when checking
 728out based on a date specification.
 729\fBcvs\fP must instead process the current \*Qhead\*U branch and, if a
 730match is not found, proceed to look for a match on the vendor branch.
 731This, combined with reasons of performance, is why \fBcvs\fP processes
 732revision (symbolic and numeric) and date specifications directly from the
 733\*Q,v\*U files.
 734.NH 2
 735Building \*Qpatch\*U Source Distributions
 736.PP
 737\fBcvs\fP can produce a \*Qpatch\*U format [Wall] output file which can be
 738used to bring a previously released software distribution current with the
 739newest release.
 740This patch file supports an entire directory hierarchy within a single
 741patch, as well as being able to add whole new files to the previous
 742release.
 743One can combine symbolic revisions and dates together to display changes in
 744a very generic way:
 745.DS
 746\f(CW	example% cvs patch -D 'December 1, 1988' \e
 747	                   -D 'January 1, 1989' sys\fP
 748.DE
 749This example displays the kernel changes made in the month of December,
 7501988.
 751To release a patch file, for example, to take the \fBcvs\fP distribution
 752from version 1.0 to version 1.4 might be done as follows:
 753.DS
 754\f(CW	example% cvs patch -rCVS_1_0 -rCVS_1_4 cvs\fP
 755.DE
 756.NH
 757CVS Experience
 758.NH 2
 759Statistics
 760.PP
 761A quick summary of the scale that \fBcvs\fP is addressing today
 762can be found in Table 1.
 763.KF
 764.TS
 765box center tab(:);
 766c s
 767c s
 768c | c
 769l | n .
 770\fB\s+2Revision Control Statistics at Prisma
 771as of 11/11/89\fP\s-2
 772_
 773How Many...:Total
 774=
 775Files:17243
 776Directories:1005
 777Lines of code:3927255
 778Removed files:131
 779Software developers:14
 780Software groups:6
 781Megabytes of source:128
 782.TE
 783.ce 100
 784.LG
 785\fBTable 1.\fP
 786.SM
 787\fBcvs\fP Statistics
 788.ce 0
 789.sp .3
 790.KE
 791Table 2 shows the history of files changed or added and the number
 792of source lines affected by the change at Prisma.
 793Only changes made to the kernel sources are included.
 794.KF
 795.TS
 796box center tab(:);
 797c s s s s
 798c s s s s
 799c || c | c || c | c
 800c || c | c || c | c
 801l || n | n || n | n.
 802\fB\s+2Prisma Kernel Source File Changes
 803By Month, 1988-1989\fP\s-2
 804_
 805Month:# Changed:# Lines:# Added:# Lines
 806\^:Files:Changed:Files:Added
 807=
 808Dec:87:3619:68:9266
 809Jan:39:4324:0:0
 810Feb:73:1578:5:3550
 811Mar:99:5301:18:11461
 812Apr:112:7333:11:5759
 813May:138:5371:17:13986
 814Jun:65:2261:27:12875
 815Jul:34:2000:1:58
 816Aug:65:6378:8:4724
 817Sep:266:23410:113:39965
 818Oct:22:621:1:155
 819Total:1000:62196:269:101799
 820.TE
 821.ce 100
 822.LG
 823\fBTable 2.\fP
 824.SM
 825\fBcvs\fP Usage History for the Kernel
 826.ce 0
 827.sp
 828.KE
 829The large number of source file changes made in September are the result of
 830merging the SunOS 4.0.3 sources into the kernel.
 831This merge process is described in section 3.3.
 832.NH 2
 833Performance
 834.PP
 835The performance of \fBcvs\fP is currently quite reasonable.
 836Little effort has been expended on tuning \fBcvs\fP, although performance
 837related decisions were made during the \fBcvs\fP design.
 838For example, \fBcvs\fP parses the
 839.SM
 840RCS
 841.LG
 842\*Q,v\*U files directly instead of running an
 843.SM
 844RCS
 845.LG
 846process.
 847This includes following branches as well as integrating with the vendor
 848source branches and the main trunk when checking out files based on a date.
 849.PP
 850Checking out the entire kernel source tree (1223 files/59 directories)
 851currently takes 16 wall clock minutes on a Sun-4/280.
 852However, bringing the tree up-to-date with the current kernel sources, once
 853it has been checked out, takes only 1.5 wall clock minutes.
 854Updating the \fIcomplete\fP 128 MByte source tree under \fBcvs\fP control
 855(17243 files/1005 directories) takes roughly 28 wall clock minutes and
 856utilizes one-third of the machine.
 857For now this is entirely acceptable; improvements on these numbers will
 858possibly be made in the future.
 859.NH 2
 860The SunOS 4.0.3 Merge
 861.PP
 862The true test of the \fBcvs\fP vendor branch support came with the arrival
 863of the SunOS 4.0.3 source upgrade tape.
 864As described above, the \fBcheckin\fP program was used to install the new
 865sources and the resulting output file listed the files that had been
 866locally modified, needing to be merged manually.
 867For the kernel, there were 94 files in conflict.
 868The \fBcvs\fP \*Qjoin\*U command was used on each of the 94 conflicting
 869files, and the remaining conflicts were resolved.
 870.PP
 871The \*Qjoin\*U command performs an \fBrcsmerge\fP operation.
 872This in turn uses \fI/usr/lib/diff3\fP to produce a three-way diff file.
 873As it happens, the \fBdiff3\fP program has a hard-coded limit of 200
 874source-file changes maximum.
 875This proved to be too small for a few of the kernel files that needed
 876merging by hand, due to the large number of local changes that Prisma had
 877made.
 878The \fBdiff3\fP problem was solved by increasing the hard-coded limit by an
 879order of magnitude.
 880.PP
 881The SunOS 4.0.3 kernel source upgrade distribution contained
 882346 files, 233 of which were modifications to previously released files,
 883and 113 of which were newly added files.
 884\fBcheckin\fP added the 113 new files to the source repository
 885without intervention.
 886Of the 233 modified files, 139 dropped in cleanly by \fBcheckin\fP, since
 887Prisma had not made any local changes to them, and 94 required manual
 888merging due to local modifications.
 889The 233 modified files consisted of 20,766 lines of differences.
 890It took one developer two days to manually merge the 94 files using the
 891\*Qjoin\*U command and resolving conflicts manually.
 892An additional day was required for kernel debugging.
 893The entire process of merging over 20,000 lines of differences was
 894completed in less than a week.
 895This one time-savings alone was justification enough for the \fBcvs\fP
 896development effort; we expect to gain even more when tracking future SunOS
 897releases.
 898.NH
 899Future Enhancements and Current Bugs
 900.PP
 901Since \fBcvs\fP was designed to be incomplete, for reasons of design
 902simplicity, there are naturally a good
 903number of enhancements that can be made to make it more useful.
 904As well, some nuisances exist in the current implementation.
 905.RS
 906.IP \(bu 3
 907\fBcvs\fP does not currently \*Qremember\*U who has a checked out a copy of a
 908module.
 909As a result, it is impossible to know who might be working on the same
 910module that you are.
 911A simple-minded database that is updated nightly would likely suffice.
 912.IP \(bu 3
 913Signal processing, keyboard interrupt handling in particular, is currently
 914somewhat weak.
 915This is due to the heavy use of the \fBsystem\fP\|(3) library
 916function to execute
 917.SM
 918RCS
 919.LG
 920programs like \fBco\fP and \fBci\fP.
 921It sometimes takes multiple interrupts to make \fBcvs\fP quit.
 922This can be fixed by using a home-grown \fBsystem\fP\|() replacement.
 923.IP \(bu 3
 924Security of the source repository is currently not dealt with directly.
 925The usual UNIX approach of user-group-other security permissions through
 926the file system is utilized, but nothing else.
 927\fBcvs\fP could likely be a set-group-id executable that checks a
 928protected database to verify user access permissions for particular objects
 929before allowing any operations to affect those objects.
 930.IP \(bu 3
 931With every checked-out directory, \fBcvs\fP maintains some administrative
 932files that record the current revision numbers of the checked-out files as
 933well as the location of the respective source repository.
 934\fBcvs\fP does not recover nicely at all if these administrative files are
 935removed.
 936.IP \(bu 3
 937The source code for \fBcvs\fP has been tested extensively on Sun-3 and
 938Sun-4 systems, all running SunOS 4.0 or later versions of the operating
 939system.
 940Since the code has not yet been compiled under other platforms, the overall
 941portability of the code is still questionable.
 942.IP \(bu 3
 943As witnessed in the previous section, the \fBcvs\fP method for tracking
 944third party vendor source distributions can work quite nicely.
 945However, if the vendor changes the directory structure or the file names
 946within the source distribution, \fBcvs\fP has no way of matching the old
 947release with the new one.
 948It is currently unclear as to how to solve this, though it is certain to
 949happen in practice.
 950.RE
 951.NH
 952Availability
 953.PP
 954The \fBcvs\fP program sources can be found in a recent posting to the
 955\fBcomp.sources.unix\fP newsgroup.
 956It is also currently available via anonymous ftp from \*Qprisma.com\*U.
 957Copying rights for \fBcvs\fP will be covered by the GNU General Public
 958License.
 959.NH
 960Summary
 961.PP
 962Prisma has used \fBcvs\fP since December, 1988.
 963It has evolved to meet our specific needs of revision and release control.
 964We will make our code freely available so that others can
 965benefit from our work, and can enhance \fBcvs\fP to meet broader needs yet.
 966.PP
 967Many of the other software release and revision control systems, like the
 968one described in [Glew], appear to use a collection of tools that are
 969geared toward specific environments \(em one set of tools for the kernel,
 970one set for \*Qgeneric\*U software, one set for utilities, and one set for
 971kernel and utilities.
 972Each of these tool sets apparently handle some specific aspect of the
 973problem uniquely.
 974\fBcvs\fP took a somewhat different approach.
 975File sharing through symbolic or hard links is not addressed; instead, the
 976disk space is simply burned since it is \*Qcheap.\*U
 977Support for producing objects for multiple architectures is not addressed;
 978instead, a parallel checked-out source tree must be used for each
 979architecture, again wasting disk space to simplify complexity and ease of
 980use \(em punting on this issue allowed \fIMakefile\fPs to remain
 981unchanged, unlike the approach taken in [Mahler], thereby maintaining closer
 982compatibility with the third-party vendor sources.
 983\fBcvs\fP is essentially a source-file server, making no assumptions or
 984special handling of the sources that it controls.
 985To \fBcvs\fP:
 986.QP
 987A source is a source, of course, of course, unless of course the source is
 988Mr. Ed.\**
 989.FS
 990\fBcvs\fP, of course, does not really discriminate against Mr. Ed.\**
 991.FE
 992.FS
 993Yet.
 994.FE
 995.LP
 996Sources are maintained, saved, and retrievable at any time based on
 997symbolic or numeric revision or date in the past.
 998It is entirely up to \fBcvs\fP wrapper programs to provide for release
 999environments and such.
1000.PP
1001The major advantage of \fBcvs\fP over the
1002many other similar systems that have already been designed is the
1003simplicity of \fBcvs\fP.
1004\fBcvs\fP contains only three programs that do all the work of release
1005and revision control, and two manually-maintained administrative
1006files for each source repository.
1007Of course, the deciding factor of any tool is whether people use it, and if
1008they even \fIlike\fP to use it.
1009At Prisma, \fBcvs\fP prevented members of the kernel
1010group from killing each other.
1011.NH
1012Acknowledgements
1013.PP
1014Many thanks to Dick Grune at Vrije Universiteit in Amsterdam for his work
1015on the original version of \fBcvs\fP and for making it available to the
1016world.
1017Thanks to Jeff Polk of Prisma for helping with the design of the module
1018database, vendor branch support, and for writing the \fBcheckin\fP shell
1019script.
1020Thanks also to the entire software group at Prisma for taking the
1021time to review the paper and correct my grammar.
1022.NH
1023References
1024.IP [Bell] 12
1025Bell Telephone Laboratories.
1026\*QSource Code Control System User's Guide.\*U
1027\fIUNIX System III Programmer's Manual\fP, October 1981.
1028.IP [Courington] 12
1029Courington, W.
1030\fIThe Network Software Environment\fP,
1031Sun Technical Report FE197-0, Sun Microsystems Inc, February 1989.
1032.IP [Essick] 12
1033Essick, Raymond B. and Robert Bruce Kolstad.
1034\fINotesfile Reference Manual\fP,
1035Department of Computer Science Technical Report #1081,
1036University of Illinois at Urbana-Champaign, Urbana, Illinois,
10371982, p. 26.
1038.IP [Glew] 12
1039Glew, Andy.
1040\*QBoxes, Links, and Parallel Trees:
1041Elements of a Configuration Management System.\*U
1042\fIWorkshop Proceedings of the Software Management Conference\fP, USENIX,
1043New Orleans, April 1989.
1044.IP [Grune] 12
1045Grune, Dick.
1046Distributed the original shell script version of \fBcvs\fP in the
1047\fBcomp.sources.unix\fP volume 6 release in 1986.
1048.IP [Honda] 12
1049Honda, Masahiro and Terrence Miller.
1050\*QSoftware Management Using a CASE Environment.\*U
1051\fIWorkshop Proceedings of the Software Management Conference\fP, USENIX,
1052New Orleans, April 1989.
1053.IP [Mahler] 12
1054Mahler, Alex and Andreas Lampen.
1055\*QAn Integrated Toolset for Engineering Software Configurations.\*U
1056\fIProceedings of the ACM SIGSOFT/SIGPLAN Software Engineering Symposium on
1057Practical Software Development Environments\fP, ACM, Boston, November 1988.
1058Described is the \fBshape\fP toolkit posted to the
1059\fBcomp.sources.unix\fP newsgroup in the volume 19 release.
1060.IP [Tichy] 12
1061Tichy, Walter F.
1062\*QDesign, Implementation, and Evaluation of a Revision Control System.\*U
1063\fIProceedings of the 6th International Conference on Software
1064Engineering\fP, IEEE, Tokyo, September 1982.
1065.IP [Wall] 12
1066Wall, Larry.
1067The \fBpatch\fP program is an indispensable tool for applying a diff file
1068to an original.
1069Can be found on uunet.uu.net in ~ftp/pub/patch.tar.