xref: /netbsd-src/external/gpl2/xcvs/dist/BUGS (revision a7c918477dd5f12c1da816ba05caf44eab2d06d6)

See the Cederqvist manual (cvs.texinfo) for information on how to
report bugs (and what will happen to your bug reports if you do).

The following is a list of some of the known bugs.  It may or may not
be comprehensive.  We would dearly love for people to volunteer to
help us keep it up to date (for starters, if you notice any
inaccuracies, please let bug-cvs know as described in the Cederqvist
manual).  There are some other reported bugs in MINOR-BUGS; the
difference, at least in theory, is that those bugs are less serious.


* For platform-specific information (in some cases including known
bugs), see README.VMS, windows-NT/README, or os2/README.  There is no
similar file for the unix-like operating systems (not yet, at least).
This file also might contain some platform-specific bugs.


* If your login name contains a space or various other characters
(particularly an issue on Windows), CVS will have trouble (it will
write invalid RCS files, probably).  The fix would be to have CVS
change such characters to underscores before writing them to the RCS
file.  Furthermore, the LOGNAME or USER environment variables usually
won't override the system login name, so this can be hard to work
around.


* If you specify the -w global option to client/server CVS, it only
overrides a CVSREAD environment variable set on the client, not a
CVSREAD variable which was set on the server (for example, in .bashrc
when the server was run via rsh).  The fix of course will be to
provide a "Option-read-write" request which sends -w, in addition to
"Global_option -r" which sends -r.


* Symbolic links to files will not work with or without LockDir.  In the
repository, you should avoid using symbolic links to files since this issue
can cause data loss.  Symlinks are only a problem when writing files.  If your
repository does not allow any write access, symlinks are not a problem.


* Symbolic links to directories will not work with LockDir.  In the
repository, you should avoid using symbolic links to directories if
you intend to use LockDir as the correct directory will NOT be locked
by CVS during write.  Directory symlinks are not recommended, but should work
as long as LockDir is not being used.  Symlinks are only a problem when
writing files.  If your repository does not allow any write access, symlinks
are never a problem, whether or not LockDir is in use.


* The -m option to "cvs add" does not work with client/server CVS.
CVS will accept the option, but it won't actually set the
file's description.


* cvs update walks into a user's work directory if there's a directory
  of the same name in the repository even if the user's directory
  doesn't yet have a CVS admin sub-directory.  This can greatly confuse
  users who try to add the same directory at nearly the same time.


* From: "Charles M. Hannum" <mycroft@ai.mit.edu>
  To: info-cvs@prep.ai.mit.edu
  Subject: Still one more bug
  Date: Sat, 25 Feb 1995 17:01:15 -0500

  mycroft@duality [1]; cd /usr/src/lib/libc
  mycroft@duality [1]; cvs diff -C2 '-D1 day ago' -Dnow
  cvs server: Diffing .
  cvs server: Diffing DB
  cvs [server aborted]: could not chdir to DB: No such file or directory
  mycroft@duality [1];

  `DB' is an old directory, which no longer has files in it, and is
  removed automatically when I use the `-P' option to checkout.

  This error doesn't occur when run locally.

  P.S.  Is anyone working on fixing these bugs?


* CVS does not always seem to be waiting to the next filesystem timestamp
quanta after commits.  So far this has only shown up in testing under the BSDI
OS.  The symptoms are that ocassionally CVS will not notice that modified files
are modified, though the file must be modified within a short time after the
commit, probably milliseconds or seconds, for this symptom to be noticed.  One
suspected cause is that one of the calls to sleep_past() is being called with
an incorrect value, though this does not explain why symptoms have only been
noticed under BSDI.


* The CVS server is leaving a temp directory (/tmp/cvs-serv*) on AIX 4.3 under
very rare circumstances (one out of c. 10,500 test cases).  This appears to be
dependent on some sort of race condition as it disappears with tracing enabled
and under the debugger.  Informative feedback is welcome.


* UNICOS 9.0 on Cray currently fails testing in both client/server and
  writeproxy modes.


* Status

  This experimental version of CVS contains new features which may not have
  been tested as thoroughly as the stable release.  It is classified as:

                          /*-------------.
                          | Experimental |
                          `-------------*/

                     /*-------------------------.
                     | Sane for full scale use. |
                     `-------------------------*/