[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
CVS bug with -D and vendor branches
From: |
Unruh, Erwin |
Subject: |
CVS bug with -D and vendor branches |
Date: |
Mon, 27 May 2002 14:42:23 +0200 |
Hello,
I found an error in CVS, version 1.10.7. It is related to the feature of vendor
branches. If using these to track the progress of an external source, CVS does
not correctly handle date information. In specific, when a file was only changed
by the vendor for some time, and then a private change was made, CVS is not able
to track down the state of a certain date.
To reproduce the problem, see below the commands I did to reproduce the problem.
I assume that the problem has to do with the way, CVS handles these vendor
branches. When only the vendor changes the file, the internal version is 1.1.1.1
and 1.1.1.2 and follows. When doing a 'cvs update' the last from this branch is
taken.
When a private commit is done, it is given version 1.2.
Now there no longer is a history that version 1.1.1.2 was considered head and
used.
Is this a known problem?
Please also respond directly, as I do not read this mailing list.
Erwin
Version 1.10.7 on a SunOS 5.8 sparc machine
**********************************
# create a new archive
export CVSROOT=`pwd`/archive
mkdir archive
mkdir vendor
cvs -d $CVSROOT init
cd vendor
echo Version1 > file
cvs import -mcomment test MAIN START
# Date 1
date
sleep 120
cd ..
cvs checkout test
cd vendor
echo Version2 > file
cvs import -mcomment test MAIN VERSION
# Date 2
date
sleep 120
# now merge in the difference
cd ../test
cvs update
#cvs update -jSTART -jVERSION test
#cvs commit -mcomment
# Date 3
date
sleep 120
# change the file now
echo Version3 > file
cvs commit -mcomment file
# Date 4
date
# now try to get Date 3 version
cvs update -D "date3"
Erwin Unruh, Fujitsu Siemens Computers, C/C++ compiler group
- CVS bug with -D and vendor branches,
Unruh, Erwin <=