CVS II: Parallelizing Software Development Brian Berliner Prisma, Inc. 5465 Mark Dabling Blvd. Colorado Springs, CO 80918 email@example.com ABSTRACT The program described in this paper fills a need in the UNIX community for a freely available tool to manage software revision and release con- trol in a multi-developer, multi-directory, multi-group environment. This tool also addresses the increasing need for tracking third-party ven- dor source distributions while trying to maintain local modifications to earlier releases. 1. Background In large software development projects, it is usually necessary for more than one software developer to be modify- ing (usually different) modules of the code at the same time. Some of these code modifications are done in an exper- imental sense, at least until the code functions correctly, and some testing of the entire program is usually necessary. Then, the modifications are returned to a master source repository so that others in the project can enjoy the new bug-fix or functionality. In order to manage such a project, some sort of revision control system is necessary. Specifically, UNIX kernel development is an excel- lent example of the problems that an adequate revision con- trol system must address. The SunOS kernel is composed of over a thousand files spread across a hierarchy of dozens of directories. Pieces of the kernel must be edited by many software developers within an organization. While _________________________  UNIX is a registered trademark of AT&T.  SunOS is a trademark of Sun Microsystems, Inc.  Yes, the SunOS 4.0 kernel is composed of over a thousand files! October 6, 2015 undesirable in theory, it is not uncommon to have two or more people making modifications to the same file within the kernel sources in order to facilitate a desired change. Existing revision control systems like RCS [Tichy] or SCCS [Bell] serialize file modifications by allowing only one developer to have a writable copy of a particular file at any one point in time. That developer is said to have "locked" the file for his exclusive use, and no other developer is allowed to check out a writable copy of the file until the locking developer has finished impeding oth- ers' productivity. Development pressures of productivity and deadlines often force organizations to require that multiple developers be able to simultaneously edit copies of the same revision controlled file. The necessity for multiple developers to modify the same file concurrently questions the value of serialization-based policies in traditional revision con- trol. This paper discusses the approach that Prisma took in adapting a standard revision control system, RCS, along with an existing public-domain collection of shell scripts that sits atop RCS and provides the basic conflict-resolution algorithms. The resulting program, cvs, addresses not only the issue of conflict-resolution in a multi-developer open- editing environment, but also the issues of software release control and vendor source support and integration. 2. The CVS Program cvs (Concurrent Versions System) is a front end to the RCS revision control system which extends the notion of revision control from a collection of files in a single directory to a hierarchical collection of directories each containing revision controlled files. Directories and files in the cvs system can be combined together in many ways to form a software release. cvs provides the functions neces- sary to manage these software releases and to control the concurrent editing of source files among multiple software developers. The six major features of cvs are listed below, and will be described in more detail in the following sections: 1. Concurrent access and conflict-resolution algo- rithms to guarantee that source changes are not "lost." 2. Support for tracking third-party vendor source distributions while maintaining the local modifi- cations made to those sources. 3. A flexible module database that provides a sym- bolic mapping of names to components of a larger October 6, 2015 software distribution. This symbolic mapping pro- vides for location independence within the software release and, for example, allows one to check out a copy of the "diff" program without ever knowing that the sources to "diff" actually reside in the "bin/diff" directory. 4. Configurable logging support allows all "commit- ted" source file changes to be logged using an arbitrary program to save the log messages in a file, notesfile, or news database. 5. A software release can be symbolically tagged and checked out at any time based on that tag. An exact copy of a previous software release can be checked out at any time, regardless of whether files or directories have been added/removed from the "current" software release. As well, a "date" can be used to check out the exact version of the software release as of the specified date. 6. A "patch" format file [Wall] can be produced between two software releases, even if the releases span multiple directories. The sources maintained by cvs are kept within a single directory hierarchy known as the "source repository." This "source repository" holds the actual RCS ",v" files directly, as well as a special per-repository directory (CVSROOT.adm) which contains a small number of administra- tive files that describe the repository and how it can be accessed. See Figure 1 for a picture of the cvs tree. 2.1. Software Conflict Resolution cvs allows several software developers to edit personal copies of a revision controlled file concurrently. The revi- sion number of each checked out file is maintained indepen- dently for each user, and cvs forces the checked out file to be current with the "head" revision before it can be "com- mitted" as a permanent change. A checked out file is brought up-to-date with the "head" revision using the "update" com- mand of cvs. This command compares the "head" revision number with that of the user's file and performs an RCS _________________________  The basic conflict-resolution algorithms used in the cvs program find their roots in the original work done by Dick Grune at Vrije Universiteit in Amsterdam and posted to comp.sources.unix in the volume 6 release sometime in 1986. This original version of cvs was a collection of shell scripts that combined to form a front end to the RCS programs. October 6, 2015 __________________________________________________ line from 4.112,9.200 to 5.550,8.887 line from 5.447,8.884 to 5.550,8.887 to 5.458,8.933 line from 4.112,9.200 to 4.550,8.950 line from 4.451,8.978 to 4.550,8.950 to 4.476,9.021 line from 4.112,9.200 to 3.737,8.887 line from 3.798,8.971 to 3.737,8.887 to 3.830,8.932 line from 3.612,8.762 to 4.737,8.137 line from 4.638,8.164 to 4.737,8.137 to 4.662,8.208 line from 3.612,8.762 to 3.737,8.137 line from 3.693,8.231 to 3.737,8.137 to 3.742,8.240 line from 3.612,8.762 to 2.612,8.200 line from 2.687,8.271 to 2.612,8.200 to 2.712,8.227 line from 2.362,9.262 to 2.737,8.950 line from 2.645,8.995 to 2.737,8.950 to 2.677,9.033 line from 2.362,9.262 to 1.925,8.950 line from 1.992,9.028 to 1.925,8.950 to 2.021,8.988 line from 3.362,9.762 to 4.050,9.387 line from 3.950,9.413 to 4.050,9.387 to 3.974,9.457 line from 3.362,9.762 to 2.487,9.387 line from 2.570,9.450 to 2.487,9.387 to 2.589,9.404 "newfs.c,v" at 4.487,8.043 ljust "mkfs.c,v" at 3.487,8.043 ljust "Makefile,v" at 2.237,8.043 ljust "newfs" at 3.487,8.793 ljust "halt.c,v" at 5.487,8.793 ljust "Makefile,v" at 4.237,8.793 ljust "modules,v" at 2.487,8.793 ljust "loginfo,v" at 1.488,8.793 ljust "etc" at 3.987,9.293 ljust "CVSROOT.adm" at 1.988,9.293 ljust "/src/master" at 2.987,9.793 ljust __________________________________________________ Figure 1. cvs Source Repository merge operation if they are not the same. The result of the merge is a file that contains the user's modifications and those modifications that were "committed" after the user checked out his version of the file (as well as a backup copy of the user's original file). cvs points out any con- flicts during the merge. It is the user's responsibility to resolve these conflicts and to "commit" his/her changes when ready. Although the cvs conflict-resolution algorithm was defined in 1986, it is remarkably similar to the "Copy- Modify-Merge" scenario included with NSE and described in _________________________ October 6, 2015 [Honda] and [Courington]. The following explanation from [Honda] also applies to cvs: Simply stated, a developer copies an object without locking it, modifies the copy, and then merges the modified copy with the original. This paradigm allows developers to work in isolation from one another since changes are made to copies of objects. Because locks are not used, develop- ment is not serialized and can proceed in paral- lel. Developers, however, must merge objects after the changes have been made. In particular, a developer must resolve conflicts when the same object has been modified by someone else. In practice, Prisma has found that conflicts that occur when the same object has been modified by someone else are quite rare. When they do happen, the changes made by the other developer are usually easily resolved. This practical use has shown that the "Copy-Modify-Merge" paradigm is a correct and useful one. 2.2. Tracking Third-Party Source Distributions Currently, a large amount of software is based on source distributions from a third-party distributor. It is often the case that local modifications are to be made to this distribution, and that the vendor's future releases should be tracked. Rolling your local modifications forward into the new vendor release is a time-consuming task, but cvs can ease this burden somewhat. The checkin program of cvs initially sets up a source repository by integrating the source modules directly from the vendor's release, preserv- ing the directory hierarchy of the vendor's distribution. The branch support of RCS is used to build this vendor release as a branch of the main RCS trunk. Figure 2 shows how the "head" tracks a sample vendor branch when no local modifications have been made to the file. Once this is done, developers can check out files and make local changes to the vendor's source distribution. These local changes form a new branch to the tree which is then used as the source for future check outs. Figure 3 shows how the "head" moves to the main RCS trunk when a local modification is made. When a new version of the vendor's source distribution arrives, the checkin program adds the new and changed vendor's files to the already existing source repository. For files that have not been changed locally, the new file from the vendor becomes the current "head" revision. For _________________________  NSE is the Network Software Environment, a pro- duct of Sun Microsystems, Inc. October 6, 2015 __________________________________________________ ellipse at 3.237,6.763 wid 1.000 ht 0.500 dashwid = 0.050i line dashed from 3.237,7.513 to 3.737,7.513 to 3.737,9.762 to 4.237,9.762 line from 4.138,9.737 to 4.237,9.762 to 4.138,9.787 line dashed from 2.237,8.262 to 3.237,8.262 to 3.237,7.013 line from 3.212,7.112 to 3.237,7.013 to 3.262,7.112 line from 3.737,6.763 to 4.237,6.763 line from 4.138,6.737 to 4.237,6.763 to 4.138,6.788 line from 2.237,6.763 to 2.737,6.763 line from 2.637,6.737 to 2.737,6.763 to 2.637,6.788 line from 1.738,6.013 to 1.738,6.513 line from 1.762,6.413 to 1.738,6.513 to 1.713,6.413 line 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 line 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 line 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 line 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 line from 4.737,7.013 to 4.737,7.513 line from 4.763,7.413 to 4.737,7.513 to 4.712,7.413 line from 4.737,8.012 to 4.737,8.512 line from 4.763,8.412 to 4.737,8.512 to 4.712,8.412 line 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 line from 4.737,9.012 to 4.737,9.512 line from 4.763,9.412 to 4.737,9.512 to 4.712,9.412 line 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 " "'SunOS'" at 2.987,6.293 ljust "1.1.1" at 3.050,6.793 ljust "1.1" at 1.613,6.793 ljust "188.8.131.52" at 4.487,6.793 ljust "184.108.40.206" at 4.487,7.793 ljust "220.127.116.11" at 4.487,8.793 ljust "18.104.22.168" at 4.487,9.793 ljust "'SunOS_4_0'" at 5.487,6.793 ljust "'SunOS_4_0_1'" at 5.487,7.793 ljust "'YAPT_5_5C'" at 5.487,8.793 ljust "'SunOS_4_0_3'" at 5.487,9.793 ljust "rcsfile.c,v" at 2.987,5.543 ljust __________________________________________________ Figure 2. cvs Vendor Branch Example files that have been modified locally, checkin warns that the file must be merged with the new vendor release. The cvs "join" command is a useful tool that aids this process by performing the necessary RCS merge, as is done above when performing an "update." There is also limited support for "dual" derivations for source files. See Figure 4 for a sample dual-derived October 6, 2015 __________________________________________________ ellipse at 3.237,6.763 wid 1.000 ht 0.500 dashwid = 0.050i line dashed from 2.800,9.075 to 1.738,9.075 to 1.738,8.012 line from 1.713,8.112 to 1.738,8.012 to 1.762,8.112 line from 1.738,7.013 to 1.738,7.513 line from 1.762,7.413 to 1.738,7.513 to 1.713,7.413 line 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 line from 3.737,6.763 to 4.237,6.763 line from 4.138,6.737 to 4.237,6.763 to 4.138,6.788 line from 2.237,6.763 to 2.737,6.763 line from 2.637,6.737 to 2.737,6.763 to 2.637,6.788 line from 1.738,6.013 to 1.738,6.513 line from 1.762,6.413 to 1.738,6.513 to 1.713,6.413 line 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 line 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 line 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 line 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 line from 4.737,7.013 to 4.737,7.513 line from 4.763,7.413 to 4.737,7.513 to 4.712,7.413 line from 4.737,8.012 to 4.737,8.512 line from 4.763,8.412 to 4.737,8.512 to 4.712,8.412 line 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 line from 4.737,9.012 to 4.737,9.512 line from 4.763,9.412 to 4.737,9.512 to 4.712,9.412 line 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 "1.2" at 1.613,7.793 ljust " "'SunOS'" at 2.987,6.293 ljust "1.1.1" at 3.050,6.793 ljust "1.1" at 1.613,6.793 ljust "22.214.171.124" at 4.487,6.793 ljust "126.96.36.199" at 4.487,7.793 ljust "188.8.131.52" at 4.487,8.793 ljust "184.108.40.206" at 4.487,9.793 ljust "'SunOS_4_0'" at 5.487,6.793 ljust "'SunOS_4_0_1'" at 5.487,7.793 ljust "'YAPT_5_5C'" at 5.487,8.793 ljust "'SunOS_4_0_3'" at 5.487,9.793 ljust "rcsfile.c,v" at 2.987,5.543 ljust __________________________________________________ Figure 3. cvs Local Modification to Vendor Branch file. This example tracks the SunOS distribution but includes major changes from Berkeley. These BSD files are saved directly in the RCS file off a new branch. October 6, 2015 __________________________________________________ ellipse at 2.337,8.575 wid 0.700 ht 0.375 ellipse at 2.312,9.137 wid 0.700 ht 0.375 line from 1.225,9.012 to 1.225,9.363 line from 1.250,9.263 to 1.225,9.363 to 1.200,9.263 line 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 line 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 line 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 line 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 line 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 line from 4.425,8.762 to 4.425,9.113 line from 4.450,9.013 to 4.425,9.113 to 4.400,9.013 line from 4.425,9.475 to 4.425,9.850 line from 4.450,9.750 to 4.425,9.850 to 4.400,9.750 line 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 line 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 line 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 line from 1.238,8.075 to 1.238,8.637 line from 1.262,8.537 to 1.238,8.637 to 1.213,8.537 line from 1.613,8.825 to 1.975,8.575 line from 1.878,8.611 to 1.975,8.575 to 1.907,8.652 line from 2.675,8.575 to 4.050,8.575 line from 3.950,8.550 to 4.050,8.575 to 3.950,8.600 line from 2.675,9.137 to 3.050,9.137 line from 2.950,9.112 to 3.050,9.137 to 2.950,9.162 line from 3.425,9.325 to 3.425,9.637 line from 3.450,9.537 to 3.425,9.637 to 3.400,9.537 line from 1.613,8.825 to 1.925,9.137 line from 1.872,9.049 to 1.925,9.137 to 1.837,9.084 "'BSD'" at 2.138,9.481 ljust "1.2" at 1.113,9.543 ljust "1.1" at 1.125,8.831 ljust "220.127.116.11" at 4.175,8.543 ljust "18.104.22.168" at 4.175,9.281 ljust "22.214.171.124" at 4.175,9.993 ljust "126.96.36.199" at 3.175,9.793 ljust "188.8.131.52" at 3.175,9.106 ljust "rcsfile.c,v" at 2.425,7.706 ljust "1.1.1" at 2.175,8.568 ljust "'SunOS'" at 2.125,8.243 ljust "1.1.2" at 2.163,9.131 ljust __________________________________________________ Figure 4. cvs Support For "Dual" Derivations 2.3. Location Independent Module Database cvs contains support for a simple, yet powerful, "module" database. For reasons of efficiency, this database October 6, 2015 is stored in ndbm(3) format. The module database is used to apply names to collections of directories and files as a matter of convenience for checking out pieces of a large software distribution. The database records the physical location of the sources as a form of information hiding, allowing one to check out whole directory hierarchies or individual files without regard for their actual location within the global source distribution. Consider the following small sample of a module data- base, which must be tailored manually to each specific source repository environment: #key [-option argument] directory [files...] diff bin/diff libc lib/libc sys -o sys/tools/make_links sys modules -i mkmodules CVSROOT.adm modules kernel -a sys lang/adb ps bin Makefile ps.c The "diff" and "libc" modules refer to whole directory hierarchies that are extracted on check out. The "sys" module extracts the "sys" hierarchy, and runs the "make_links" program at the end of the check out process (the -o option specifies a program to run on checkout). The "modules" module allows one to edit the module database file and runs the "mkmodules" program on checkin to regenerate the ndbm database that cvs uses. The "kernel" module is an alias (as the -a option specifies) which causes the remain- ing arguments after the -a to be interpreted exactly as if they had been specified on the command line. This is useful for objects that require shared pieces of code from far away places to be compiled (as is the case with the kernel debugger, kadb, which shares code with the standard adb debugger). The "ps" module shows that the source for "ps" lives in the "bin" directory, but only Makefile and ps.c are required to build the object. The module database at Prisma is now populated for the entire UNIX distribution and thereby allows us to issue the following convenient commands to check out components of the UNIX distribution without regard for their actual location within the master source repository: example% cvs checkout diff example% cvs checkout libc ps example% cd diff; make In building the module database file, it is quite pos- sible to have name conflicts within a global software October 6, 2015 distribution. For example, SunOS provides two cat programs: one for the standard environment, /bin/cat, and one for the System V environment, /usr/5bin/cat. We resolved this con- flict by naming the standard cat module "cat", and the Sys- tem V cat module "5cat". Similar name modifications must be applied to other conflicting names, as might be found between a utility program and a library function, though Prisma chose not to include individual library functions within the module database at this time. 2.4. Configurable Logging Support The cvs "commit" command is used to make a permanent change to the master source repository (where the RCS ",v" files live). Whenever a "commit" is done, the log message for the change is carefully logged by an arbitrary program (in a file, notesfile, news database, or mail). For example, a collection of these updates can be used to produce release notices. cvs can be configured to send log updates through one or more filter programs, based on a regular expression match on the directory that is being changed. This allows multiple related or unrelated projects to exist within a single cvs source repository tree, with each different pro- ject sending its "commit" reports to a unique log device. A sample logging configuration file might look as fol- lows: #regex filter-program DEFAULT /usr/local/bin/nfpipe -t %s utils.updates ^diag /usr/local/bin/nfpipe -t %s diag.updates ^local /usr/local/bin/nfpipe -t %s local.updates ^perf /usr/local/bin/nfpipe -t %s perf.updates ^sys /usr/local/bin/nfpipe -t %s kernel.updates This sample allows the diagnostics and performance groups to share the same source repository with the kernel and utilities groups. Changes that they make are sent directly to their own notesfile [Essick] through the "nfpipe" program. A sufficiently simple title is substituted for the "%s" argument before the filter program is executed. This logging configuration file is tailored manually to each specific source repository environment. 2.5. Tagged Releases and Dates Any release can be given a symbolic tag name that is stored directly in the RCS files. This tag can be used at any time to get an exact copy of any previous release. With equal ease, one can also extract an exact copy of the source files as of any arbitrary date in the past as well. Thus, all that's required to tag the current kernel, and to tag October 6, 2015 the kernel as of the Fourth of July is: example% cvs tag TEST_KERNEL kernel example% cvs tag -D 'July 4' PATRIOTIC_KERNEL kernel The following command would retrieve an exact copy of the test kernel at some later date: example% cvs checkout -fp -rTEST_KERNEL kernel The -f option causes only files that match the specified tag to be extracted, while the -p option automatically prunes empty directories. Consequently, directories added to the kernel after the test kernel was tagged are not included in the newly extracted copy of the test kernel. The cvs date support has exactly the same interface as that provided with RCS, however cvs must process the ",v" files directly due to the special handling required by the vendor branch support. The standard RCS date handling only processes one branch (or the main trunk) when checking out based on a date specification. cvs must instead process the current "head" branch and, if a match is not found, proceed to look for a match on the vendor branch. This, combined with reasons of performance, is why cvs processes revision (symbolic and numeric) and date specifications directly from the ",v" files. 2.6. Building "patch" Source Distributions cvs can produce a "patch" format [Wall] output file which can be used to bring a previously released software distribution current with the newest release. This patch file supports an entire directory hierarchy within a single patch, as well as being able to add whole new files to the previous release. One can combine symbolic revisions and dates together to display changes in a very generic way: example% cvs patch -D 'December 1, 1988' \ -D 'January 1, 1989' sys This example displays the kernel changes made in the month of December, 1988. To release a patch file, for example, to take the cvs distribution from version 1.0 to version 1.4 might be done as follows: example% cvs patch -rCVS_1_0 -rCVS_1_4 cvs 3. CVS Experience October 6, 2015 3.1. Statistics A quick summary of the scale that cvs is addressing today can be found in Table 1. _______________________________________ | Revision Control Statistics at Prisma| | as of 11/11/89 | |____________________|_________________| | How Many... | Total | |_____________________|__________________| | Files | 17243 | | Directories | 1005 | | Lines of code | 3927255 | | Removed files | 131 | | Software developers| 14 | | Software groups | 6 | | Megabytes of source| 128 | |____________________|_________________| Table 1. cvs Statistics Table 2 shows the history of files changed or added and the number of source lines affected by the change at Prisma. Only changes made to the kernel sources are included. _________________________________________________ | Prisma Kernel Source File Changes | | By Month, 1988-1989 | |______|___________|_________|_________|_________| | | # Changed| # Lines| # Added| # Lines| | Month| Files | Changed| Files | Added | |________|____________|___________|__________|__________| | Dec | 87 | 3619 | 68 | 9266 | | Jan | 39 | 4324 | 0 | 0 | | Feb | 73 | 1578 | 5 | 3550 | | Mar | 99 | 5301 | 18 | 11461 | | Apr | 112 | 7333 | 11 | 5759 | | May | 138 | 5371 | 17 | 13986 | | Jun | 65 | 2261 | 27 | 12875 | | Jul | 34 | 2000 | 1 | 58 | | Aug | 65 | 6378 | 8 | 4724 | | Sep | 266 | 23410 | 113 | 39965 | | Oct | 22 | 621 | 1 | 155 | | Total| 1000 | 62196 | 269 | 101799 | |______|___________|_________|_________|_________| Table 2. cvs Usage History for the Kernel The large number of source file changes made in September are the result of merging the SunOS 4.0.3 sources into the kernel. This merge process is described in section 3.3. October 6, 2015 3.2. Performance The performance of cvs is currently quite reasonable. Little effort has been expended on tuning cvs, although per- formance related decisions were made during the cvs design. For example, cvs parses the RCS ",v" files directly instead of running an RCS process. This includes following branches as well as integrating with the vendor source branches and the main trunk when checking out files based on a date. Checking out the entire kernel source tree (1223 files/59 directories) currently takes 16 wall clock minutes on a Sun-4/280. However, bringing the tree up-to-date with the current kernel sources, once it has been checked out, takes only 1.5 wall clock minutes. Updating the complete 128 MByte source tree under cvs control (17243 files/1005 direc- tories) takes roughly 28 wall clock minutes and utilizes one-third of the machine. For now this is entirely accept- able; improvements on these numbers will possibly be made in the future. 3.3. The SunOS 4.0.3 Merge The true test of the cvs vendor branch support came with the arrival of the SunOS 4.0.3 source upgrade tape. As described above, the checkin program was used to install the new sources and the resulting output file listed the files that had been locally modified, needing to be merged manu- ally. For the kernel, there were 94 files in conflict. The cvs "join" command was used on each of the 94 conflicting files, and the remaining conflicts were resolved. The "join" command performs an rcsmerge operation. This in turn uses /usr/lib/diff3 to produce a three-way diff file. As it happens, the diff3 program has a hard-coded limit of 200 source-file changes maximum. This proved to be too small for a few of the kernel files that needed merging by hand, due to the large number of local changes that Prisma had made. The diff3 problem was solved by increasing the hard-coded limit by an order of magnitude. The SunOS 4.0.3 kernel source upgrade distribution con- tained 346 files, 233 of which were modifications to previ- ously released files, and 113 of which were newly added files. checkin added the 113 new files to the source reposi- tory without intervention. Of the 233 modified files, 139 dropped in cleanly by checkin, since Prisma had not made any local changes to them, and 94 required manual merging due to local modifications. The 233 modified files consisted of 20,766 lines of differences. It took one developer two days to manually merge the 94 files using the "join" command and resolving conflicts manually. An additional day was required for kernel debugging. The entire process of merging over October 6, 2015 20,000 lines of differences was completed in less than a week. This one time-savings alone was justification enough for the cvs development effort; we expect to gain even more when tracking future SunOS releases. 4. Future Enhancements and Current Bugs Since cvs was designed to be incomplete, for reasons of design simplicity, there are naturally a good number of enhancements that can be made to make it more useful. As well, some nuisances exist in the current implementation. + cvs does not currently "remember" who has a checked out a copy of a module. As a result, it is impossi- ble to know who might be working on the same module that you are. A simple-minded database that is updated nightly would likely suffice. + Signal processing, keyboard interrupt handling in particular, is currently somewhat weak. This is due to the heavy use of the system(3) library function to execute RCS programs like co and ci. It sometimes takes multiple interrupts to make cvs quit. This can be fixed by using a home-grown system() replacement. + Security of the source repository is currently not dealt with directly. The usual UNIX approach of user-group-other security permissions through the file system is utilized, but nothing else. cvs could likely be a set-group-id executable that checks a protected database to verify user access permissions for particular objects before allowing any opera- tions to affect those objects. + With every checked-out directory, cvs maintains some administrative files that record the current revi- sion numbers of the checked-out files as well as the location of the respective source repository. cvs does not recover nicely at all if these administra- tive files are removed. + The source code for cvs has been tested extensively on Sun-3 and Sun-4 systems, all running SunOS 4.0 or later versions of the operating system. Since the code has not yet been compiled under other plat- forms, the overall portability of the code is still questionable. + As witnessed in the previous section, the cvs method for tracking third party vendor source distributions can work quite nicely. However, if the vendor changes the directory structure or the file names within the source distribution, cvs has no way of October 6, 2015 matching the old release with the new one. It is currently unclear as to how to solve this, though it is certain to happen in practice. 5. Availability The cvs program sources can be found in a recent post- ing to the comp.sources.unix newsgroup. It is also currently available via anonymous ftp from "prisma.com". Copying rights for cvs will be covered by the GNU General Public License. 6. Summary Prisma has used cvs since December, 1988. It has evolved to meet our specific needs of revision and release control. We will make our code freely available so that oth- ers can benefit from our work, and can enhance cvs to meet broader needs yet. Many of the other software release and revision control systems, like the one described in [Glew], appear to use a collection of tools that are geared toward specific environ- ments - one set of tools for the kernel, one set for "gen- eric" software, one set for utilities, and one set for ker- nel and utilities. Each of these tool sets apparently handle some specific aspect of the problem uniquely. cvs took a somewhat different approach. File sharing through symbolic or hard links is not addressed; instead, the disk space is simply burned since it is "cheap." Support for producing objects for multiple architectures is not addressed; instead, a parallel checked-out source tree must be used for each architecture, again wasting disk space to simplify com- plexity and ease of use - punting on this issue allowed Makefiles to remain unchanged, unlike the approach taken in [Mahler], thereby maintaining closer compatibility with the third-party vendor sources. cvs is essentially a source-file server, making no assumptions or special handling of the sources that it controls. To cvs: A source is a source, of course, of course, unless of course the source is Mr. Ed. Sources are maintained, saved, and retrievable at any time based on symbolic or numeric revision or date in the past. It is entirely up to cvs wrapper programs to provide for release environments and such. _________________________  cvs, of course, does not really discriminate against Mr. Ed.  Yet. October 6, 2015 The major advantage of cvs over the many other similar systems that have already been designed is the simplicity of cvs. cvs contains only three programs that do all the work of release and revision control, and two manually-maintained administrative files for each source repository. Of course, the deciding factor of any tool is whether people use it, and if they even like to use it. At Prisma, cvs prevented members of the kernel group from killing each other. 7. Acknowledgements Many thanks to Dick Grune at Vrije Universiteit in Amsterdam for his work on the original version of cvs and for making it available to the world. Thanks to Jeff Polk of Prisma for helping with the design of the module database, vendor branch support, and for writing the checkin shell script. Thanks also to the entire software group at Prisma for taking the time to review the paper and correct my gram- mar. 8. References [Bell] Bell Telephone Laboratories. "Source Code Con- trol System User's Guide." UNIX System III Programmer's Manual, October 1981. [Courington]Courington, W. The Network Software Environment, Sun Technical Report FE197-0, Sun Microsystems Inc, February 1989. [Essick] Essick, Raymond B. and Robert Bruce Kolstad. Notesfile Reference Manual, Department of Com- puter Science Technical Report #1081, University of Illinois at Urbana-Champaign, Urbana, Illi- nois, 1982, p. 26. [Glew] Glew, Andy. "Boxes, Links, and Parallel Trees: Elements of a Configuration Management System." Workshop Proceedings of the Software Management Conference, USENIX, New Orleans, April 1989. [Grune] Grune, Dick. Distributed the original shell script version of cvs in the comp.sources.unix volume 6 release in 1986. [Honda] Honda, Masahiro and Terrence Miller. "Software Management Using a CASE Environment." Workshop Proceedings of the Software Management Confer- ence, USENIX, New Orleans, April 1989. [Mahler] Mahler, Alex and Andreas Lampen. "An Integrated Toolset for Engineering Software Configura- tions." Proceedings of the ACM SIGSOFT/SIGPLAN October 6, 2015 Software Engineering Symposium on Practical Software Development Environments, ACM, Boston, November 1988. Described is the shape toolkit posted to the comp.sources.unix newsgroup in the volume 19 release. [Tichy] Tichy, Walter F. "Design, Implementation, and Evaluation of a Revision Control System." Proceedings of the 6th International Conference on Software Engineering, IEEE, Tokyo, September 1982. [Wall] Wall, Larry. The patch program is an indispens- able tool for applying a diff file to an origi- nal. Can be found on uunet.uu.net in ~ftp/pub/patch.tar. October 6, 2015
Generated on 2015-10-06 19:36:22 by $MirOS: src/scripts/roff2htm,v 1.80 2015/01/02 13:54:19 tg Exp $
These manual pages and other documentation are copyrighted by their respective writers;
their source is available at our CVSweb,
AnonCVS, and other mirrors. The rest is Copyright © 2002–2015 The MirOS Project, Germany.
This product includes material provided by Thorsten Glaser.
This manual page’s HTML representation is supposed to be valid XHTML/1.1; if not, please send a bug report – diffs preferred.