okular/kpathsea/maintain.texi

\input texinfo @c -*-texinfo-*-
@c %**start of header
@setfilename maintain.info
@settitle Information For Maintainers of GNU Software
@c For double-sided printing, uncomment:
@c @setchapternewpage odd
@c UPDATE THIS DATE WHENEVER YOU MAKE CHANGES!
@set lastupdate 08 November 1994
@c %**end of header

@ifinfo
@format
START-INFO-DIR-ENTRY
* Maintaining: (maintain).        Maintaining GNU software.
END-INFO-DIR-ENTRY
@end format
@end ifinfo

@setchapternewpage off

@ifinfo
Information for maintainers of GNU software.
Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc.

Permission is granted to make and distribute verbatim copies of
this manual provided the copyright notice and this permission notice
are preserved on all copies.

@ignore
Permission is granted to process this file through TeX and print the
results, provided the printed document carries copying permission
notice identical to this one except for the removal of this paragraph
(this paragraph not being relevant to the printed manual).
@end ignore

Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided that the entire
resulting derived work is distributed under the terms of a permission
notice identical to this one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions,
except that this permission notice may be stated in a translation approved
by the Free Software Foundation.
@end ifinfo

@titlepage
@title Information For Maintainers of GNU Software
@author Richard Stallman
@author last updated @value{lastupdate}
@page

@vskip 0pt plus 1filll
Copyright @copyright{} 1992, 1993, 1994 Free Software Foundation, Inc.

Permission is granted to make and distribute verbatim copies of
this manual provided the copyright notice and this permission notice
are preserved on all copies.

Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided that the entire
resulting derived work is distributed under the terms of a permission
notice identical to this one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions,
except that this permission notice may be stated in a translation approved
by the Free Software Foundation.
@end titlepage

@ifinfo
@node Top, Preface, (dir), (dir)
@top Version

Last updated @value{lastupdate}.
@end ifinfo

@menu
* Preface::
* Legal Matters::
* Clean Ups::
* Mail::
* Old Versions::
* Archives::
* Distributions::
@end menu

@node Preface
@chapter About This Document

This file contains guidelines and advice for someone who is the
maintainer of a GNU program on behalf of the GNU project.  Anyone
can change GNU software, but there's no need to pay attention to
these guidelines unless you are maintaining a version for
widespread distribution.

Corrections or suggestions regarding this document should be sent to
@code{gnu@@prep.ai.mit.edu}.  If you make a suggestion, please include a
suggested new wording for it; our time is limited.  We prefer a context
diff to the @file{maintain.texi} file, but if you don't have that file,
please mail your suggestion anyway.

This release of the GNU Maintenance Instructions was last updated
@value{lastupdate}.

@node Legal Matters
@chapter Legal Matters

When incorporating changes from other people, make sure to follow the
correct procedures.  Doing this ensures that the FSF has the legal right
to distribute and defend GNU software.

@menu
* Copyrights::
* Recording Changes::
@end menu

@node Copyrights
@section Copyrights

For the sake of registering the copyright on later versions of the
software, you need to keep track of each person who makes significant
changes.  A change of ten lines or so, or a few such changes, in a 
large program is not significant.

@strong{Before} incorporating significant changes, make sure that the
person has signed copyright papers and that the Foundation has
received and signed them.

You can tell the person what papers to sign by email.  For large
changes, ask for an assignment.  Send the person a copy of
@file{/gd/gnuorg/assign.changes}, but first, go to the second page
and insert the person's name and the name of the program involved in
place of @samp{NAME OF PERSON} and @samp{NAME OF PROGRAM}.  Do this
before sending, because otherwise the person might sign without
noticing them.  Then the papers would be useless.

For medium to small changes, ask for a disclaimer.  Use the file
@file{/gd/gnuorg/disclaim.changes}.

To check whether papers have been received, look in
@file{/gd/gnuorg/copyright.list}.  If you can't look there directly,
@code{fsf-records@@prep.ai.mit.edu} can check for you, and can also
check for papers that are waiting to be entered and inform you
when expected papers arrive.

You can also send the person @file{/gd/gnuorg/conditions.text}, which
explains his options (assign vs. disclaim) and their consequences.

@node Recording Changes
@section Recording Changes

@strong{Keep records of which portions were written by whom.}

These records don't need to be as detailed as a change log.  They don't
need to distinguish work done at different times, only different people.

They should say which files or functions were written by each person,
and which files or functions were revised by each person.  They don't
need to say what the purpose of the change was.  The Register of
Copyrights doesn't care what the program does.

For example, this would describe an early version of GAS:

@display
Dean Elsner   first version of all files except gdb-lines.c and m68k.c.
Jay Fenlason  entire files gdb-lines.c and m68k.c, most of app.c,
  extensive changes in messages.c, input-file.c, write.c,
  revisions elsewhere.
@end display

Please keep these records in a file named @file{AUTHORS} in the source
directory for the program itself.

@node Clean Ups
@chapter Cleaning Up Changes

If someone sends you changes which are ugly and will make the program
harder to understand and maintain in the future, such as a port to
another operating system containing ad hoc conditionals, don't
hesitate to ask the person to clean up his changes before you merge
them.

Since the amount of work we do is constant in any case, the more work
we get other people to do, the faster GNU will advance.

If the person will not or can not make the changes clean enough, then
say that you can't afford to merge them.  Invite him to distribute
his changes himself, or to find other people who can make them clean
enough for us to maintain.

The only reason to do these cleanups yourself is if (1) it is easy
enough that it is less work than telling the author what to clean up,
or (2) users will greatly appreciate the improvement, and you would
almost write it yourself if you had time.

The GNU Coding Standards are a good thing to send people who have to
clean up C programs (@pxref{Top, , Contents, standards, GNU Coding
Standards}).  The Emacs Lisp manual contains an appendix that gives
coding standards for Emacs Lisp programs; it is good to urge authors to
read it (@pxref{Tips, , Tips and Standards, elisp, The GNU Emacs Lisp
Reference Manual}).

@node Mail
@chapter Dealing With Mail

Once a program is in use, you will start getting bug reports.  Some
GNU programs have their own special lists for sending bug reports.
For miscellaneous programs that don't have their own lists, we use a
catch-all list, @code{bug-gnu-utils@@prep.ai.mit.edu}.  Talk with
@code{gnu@@prep.ai.mit.edu} to arrange to be added to the proper list
for your program.

When you receive bug reports, keep in mind that the main purpose of
the bug reports is to enable you to improve the next version of the
program.  Helping individuals is only secondary.  So you don't
need to give a substantial response unless you have a reason to
(for example, to ask for more information, or to ask the user to
test a fix).  But it is good to respond to the rest of the bug
reports with just ``Thanks.''  That is quick and tells the user that
the bug report was useful.

As a practical matter, any time you spend helping individuals beyond
what is necessary for you, takes time away from maintaining the program.
While this may seem ``friendly'' and ``helpful,'' actually it is
counterproductive.  If you help one person when you could instead have
helped thousands, the world is worse off.

@node Old Versions
@chapter Recording Old Versions

It is very important to keep backup files of all source files of GNU.
You can do this using RCS if you like.  The easiest way to use RCS is
via the Version Control library in Emacs; @ref{Concepts of VC, ,
Concepts of Version Control, emacs, The GNU Emacs Manual}.

Alternatively, you can keep backup files.

@menu
* Backup Files::
* Deleting Backup Files::
@end menu

@node Backup Files
@section Backup Files

Emacs makes a backup file by renaming the old source file to a new
name.  This means that if the file is also pointed to by a
distribution directory, the distribution directory continues to point
to the same version it always did---the right thing.

We want to keep more than one backup for all GNU sources.  So, if you
are going to edit GNU sources, @emph{make certain} to put
@example
(setq version-control t)
@end example
@noindent
into your @file{.emacs} file, so that Emacs always creates numbered
backup files.

Using Emacs backup files works as long as people always make changes
with Emacs.  If you change the file in some other way, and use
@code{cp}, @code{ftp}, or @code{tar} to install it, you will
@emph{overwrite} the old version and fail to make a backup.  Don't do
that!

If you want to make a change to a source file with something other
than Emacs, you can write the changed file to another name, and use
@kbd{C-x C-w} in Emacs to write it under the real name.  This makes the
backup file properly.

You can use GNU @code{cp} or @code{mv} to install changed files if you
give them the @samp{--backup} (or, equivalently, @samp{-b}) option; then
they make backup copies the same way that Emacs does.  You should also
use the @samp{--version-control=t} option, or, alternately, first
@example
export VERSION_CONTROL=t
@end example
@noindent
(or the @code{csh} equivalent); this makes GNU @code{cp} and @code{mv}
create numbered backup files instead of a single backup file with a
@samp{~} appended to the filename.  For installing many changed files,
you can use shell wildcards and also give GNU @code{cp} or @code{mv} the
@samp{--update} (@samp{-u}) option, which only copies (or moves) files
that have been modified more recently than the existing destination
files, and the @samp{--verbose} (@samp{-v}) option, which prints the
names of the files that are actually copied (or moved).

Before you use @code{mv} or @code{cp} in this way, @emph{make sure it is
the GNU version}.  Do @samp{which mv} (in @code{csh}) or @samp{type mv}
(in @code{bash}) to verify you are not getting @file{/bin/mv} (or
likewise for @code{cp}).  Or just type @samp{cp} or @samp{mv} and look
at the usage message.

@node Deleting Backup Files
@section Deleting Backup Files

Always answer no when Emacs offers to delete backup files automatically.
The way to delete them is by hand, using @kbd{M-x dired}.

When you decide which backup files to delete, it is good to keep one
every couple of weeks or so, going back about 2 months.

If there is a long gap between versions, because that file did not change
during a long period of time, then keep the version before the gap
even if it is 2 months old.  Pretend it is just 2 weeks older than the
next kept version, so delete it when the next version is >= 6 weeks old.

If the changes in a program have been simple, then you don't need to
keep as many backup files.  Just one a month for 2 months is enough.

If you have made big changes, keep the versions before and after the
big change, until they are old enough.

If you made several changes the same day, usually the last version written
that day is best to keep.

It is almost always right to keep the most recent backup version.

@node Archives
@chapter Archives

For each program, you should keep a special magtape or cartridge as an
archive.  Each time you release a new version, @code{dd} the tar file onto
the end of the tape.  Keep a list of versions on the tape's paper
label, and add to it each time you add to the tape.

For cartridges, you can type
@example
mt -f /dev/nrst8 eom
@end example
@noindent
to go straight to the end of the data on the tape.

For reel-to-reel tapes, there is no automated way to go to the end of
the data on the tape.  You have to count the number of files (based
on the written label), and space forward over them with @samp{mt fsf}.

To be safe, it is important to check your count.  If the count is @var{n},
then do:
@example
mt -f /dev/nrmt8 fsf @var{n}-1
@end example
@noindent
This puts you at the beginning of the last existing tar file.

Then do
@example
tar tf /dev/nrmt8
@end example
@noindent
to list that tar file.  If the version number appears in a directory
name, which is a good idea, you can use this to verify that you have
reached the tar file you wanted to reach.

@kbd{C-c} the @code{tar} before it finishes; then do
@example
mt -f /dev/nrmt8 fsf
@end example
@noindent
to skip past it and its end-of-file marker.

To copy the new distribution file onto cartridge tape, do:
@example
dd if=@var{tar-file-name} of=/dev/nrst8 bs=102400
@end example
@noindent
(This specifies a blocking factor of 200.)

For reel-to-reel tape, do:
@example
dd if=@var{tar-file-name} of=/dev/nrmt8 bs=10240
@end example
@noindent
(This specifies a blocking factor of 20.)

When the tape gets full, put it aside permanently and start writing
another.

@node Distributions
@chapter Distributions

It is important to follow the GNU conventions when making GNU software
distributions.

@menu
* Distribution tar Files::
* Distribution Patches::
* Distribution on prep::
* Test Releases::
@end menu

@node Distribution tar Files
@section Distribution tar Files

The tar file for version @var{m}.@var{n} of program @code{foo} should be
named @file{foo-@var{m}.@var{n}.tar}.  It should unpack into a
subdirectory named @file{foo-@var{m}.@var{n}}.  Tar files should not
unpack into files in the current directory, because this is inconvenient
if the user happens to unpack into a directory with other files in it.

Here is how the @file{Makefile} for Bison creates the tar file.
This method is good for other programs.

@example
dist: bison.info
        echo bison-`sed -e '/version_string/!d' \
          -e 's/[^0-9.]*\([0-9.]*\).*/\1/' -e q version.c` > .fname
        -rm -rf `cat .fname`
        mkdir `cat .fname`
        dst=`cat .fname`; for f in $(DISTFILES); do \
           ln $(srcdir)/$$f $$dst/$$f || @{ echo copying $$f; \
             cp -p $(srcdir)/$$f $$dst/$$f ; @} \
        done
        tar --gzip -chf `cat .fname`.tar.gz `cat .fname`
        -rm -rf `cat .fname` .fname
@end example

Source files that are symbolic links to other file systems cannot be
installed in the temporary directory using @code{ln}, so use @code{cp}
if @code{ln} fails.

@node Distribution Patches
@section Distribution Patches

If the program is large, it is useful to make a set of diffs for each
release, against the previous important release.

At the front of the set of diffs, put a short explanation of which
version this is for and which previous version it is relative to.
Also explain what else people need to do to update the sources
properly (for example, delete or rename certain files before
installing the diffs).

The purpose of having diffs is that they are small.  To keep them
small, exclude files that the user can easily update.  For example,
exclude info files, DVI files, tags tables, output files of Bison or
Flex.  In Emacs diffs, we exclude compiled Lisp files, leaving it up
to the installer to recompile the patched sources.

When you make the diffs, each version should be in a directory suitably
named---for example, @file{gcc-2.3.2} and @file{gcc-2.3.3}.  This way,
it will be very clear from the diffs themselves which version is which.

If you use GNU @code{diff} to make the patch, use the options
@samp{-rc2P}.  That will put any new files into the output as ``entirely
different.''

If the distribution has subdirectories in it, then the diffs probably
include some files in the subdirectories.  To help users install such
patches reliably, give them precise directions for how to run patch.
For example, say this:

@display
To apply these patches, cd to the main directory of the program
and then use `patch -p1'.   `-p1' avoids guesswork in choosing
which subdirectory to find each file in.
@end display

It's wise to test your patch by applying it to a copy of the old
version, and checking that the result exactly matches the new version.

@node Distribution on prep
@section Distribution on @code{prep}

Only the latest version of any program needs to be on @code{prep}.  Being an
archive of old versions is not the function of @code{prep}.

Diffs are another matter.  Since they are much smaller than
distribution files, it is good to keep the diffs around for quite a
while.

@node Test Releases
@section Test Releases

When you release a greatly changed new major version of a program,
you might want to do so as a beta test release.

Once a program gets to be widely used and people expect it to work
solidly, it is a good idea to do pretest releases before each ``real''
release.  This means that you make a tar file, but send it only to a
group of volunteers that you have recruited.  (Use a suitable GNU
mailing list/newsgroup to recruit them.)

One thing that you should never do is to release a distribution which
is considered a pretest or beta test but which contains the version
number for the planned real release.  Many people will look only at
the version number (in the tar file name, in the directory name that
it unpacks into, or wherever they can find it) to determine whether a
tar file is the latest version.  People might look at the test release
in this way and mistake it for the real release.  Therefore, always
change the number when you make a new tar file.

If you are about to release version 4.6 but you want to do a pretest
or beta test first, call it 4.5.90.  If you need a second pretest,
call it 4.5.91, and so on.

@contents

@bye