[wpkg-announce] Announcement: WPKG 1.1.1 release
Rainer Meier
r.meier at wpkg.org
Thu Jun 25 21:09:18 CEST 2009
Hi all,
After fixing up many smaller issues and adding some nice features I produly
announce the release of WPKG 1.1.1.
Again WPKG 1.1.1 is fully backwards compatible to version 1.1.0 (if not, then
please report it).
As long was WPKG 1.1.1 is not available on the main WPKG home page you can fetch
it from Bugzilla:
<http://bugzilla.wpkg.org/attachment.cgi?id=151>
Here's a list of changes:
Change notes
============
WPKG Version: 1.1.1
Author: : Rainer Meier <skybeam (at) users.sourceforge.net>
Date : 2009-06-25
Status : released
Changes/fixes visible to the user:
Summary:
- Added new /noUpgradeBeforeRemove switch.
- Reverted STDOUT/STDERR handling to WPKG 1.0 algorithm.
- Updated README.TXT in tools/ folder.
- Fixed typo in DB inconsistency message.
- Changed wording in database-inconsistency message.
- Changed install success message wording.
- Enhanced variable expansion within variable definitions.
- Enhanced human-style version comparison.
- Improved human-style version comparison algorithm.
- Extended volatile release marker list.
- Added possibility to define volatile release markres in config.xml.
- Updated profiles.xsd
- Fixed wrong default for /noUpgradeBeforeRemove switch
- Fixed typo in install-success message. Fixes Bug 163.
- Fixed variable overwriting. Fixes Bug 156.
NEW: Added a new switch: /noUpgradeBeforeRemove.
This switch allows to disable the upgrade-before-remove feature.
Usually WPKG upgrades a package to the latest available version before it
removes the package. This allows administrators to fix bugs in the package
and assure proper removal.
However this feature can lead to (sometimes unexpected) re-execution of
packages. For example if execution=always is used the package will run
once again before it is finally removed.
It's advised to leave this option on default unless you really know what
you're doing.
Thanks to Simon.
NEW: Unfortunately the %COMSPEC%-wrapping introduced in WPKG 1.1.0 to allow
STDOUT/STDERR flushing (see Bug 155) does not work in any case as
expected. Therefore I decided to revert the change and fall back to the
default execution behavior of WPKG 1.0. This should solve execution and
return code problems for the moment.
However the issue that commands which print > 4k to STDOUT/STDERR will
block and not finish until the timeout has been reached. So make sure
your installers are (quite-) silent on STDOUT/STDERR or to redirect the
output to a file or NUL by appending ">NUL 2>&1" to commands which are
using verbose printing.
MOD: Updated README.TXT in tools/ folder to reflect the truth that the tools
put into the tools folder are not automatically added to the PATH.
Thanks to Chris Allen.
MOD: Fixed typo in in database-inconsistency message. Thanks to Gerd Ott.
MOD: Changed Wording after successful installation of a package.
Thanks to Simon.
MOD: Enhanced variable expansion within <variable /> nodes. If the value
contains shell variables they are expanded immediately.
Thanks to Chris Allen for reporting and Stefan Pendl for fix suggestion.
MOD: The comparison of of version strings the same way how a human would read
it is quite difficult. Up to now WPKG supported only versions containing a
sequence of colon-separated digits. The algorithm has been re-worked to
support much more complicated version strings. Some examples include:
A B Result
"1" "2" B is newer
"1" "15" B is newer
"1.0" "1.2.b" B is newer
"1.35" "1.35-2" B is newer
"1.35-2" "1.36" B is newer
"1.35R3" "1.36" B is newer
"1" "1.0.00.0000" Versions are equal
"1" "1.0" Versions are equal
"1.35" "1.35-2" B is newer
"1.35-2" "1.35" A is newer
"1.35R3" "1.36R4" B is newer
"1.35-2" "1.35-2.0" Versions are equal
"1.35.1" "1.35.1.0" Versions are equal
Thanks to Jens Geile for asking an improved algorithm.
MOD: Updated version comparison algorithm to treat some version appendixes as
"volatile". Volatile means that WPKG would claim that a revision witout
that suffix is newer than the one which carries this suffix. Usually a
version is assumed to be higher if it has an additional suffix but in case
of RC (Release Candidate), M (Milestone) and I (Integration) releases it's
usually the opposite way around. For example:
A B Result
"1.0RC2" "1.0" B us newer (A is an "RC" version)
"1.5" "1.5I3656" A is newer (B is an "I"/intgration version)
"1.5" "1.5M3656" A is newer (B is an "M"/milestone version)
but (still)
"1.5" "1.5u3656" B is newer (B is an update version)
MOD: Updated the list of volatile release markers (see previous change).
NEW: Added the possibility to define "volatileReleaseMarker" parameters within
config.xml. This allows users to extend the list of volatile release
markers on-the-fly.
MOD: Updated profiles.xsd. However I could not find a problem with the current
profile definition but Brian reported on the mailing list that his editor
throws a validation error. So I changed the definition according to his
proposal.
FIX: The /noUpgradeBeforeRemove remove switch was "on" by default. Somehow I
left that in the code by accident from my testing sessions.
Thanks to Stefan Pendl for reporting.
FIX: Fixed typo in install-success notification:
"Package and all chained packages ininstalled successfully."
changed to
"Package and all chained packages installed successfully."
Thanks to Stefan Pendl for reporting.
Fixes Bug 163.
FIX: Fixed WPKG quit if a variable definition appears twice (re-defined). For
example if a variable is defined in profiles.xml and overwritten in
hosts.xml.
Thanks to Jindrich Vorlicek for reporting.
Fixes Bug 156.
Thanks to all contributors who helped to evolve WPKG to the current state!
br,
Rainer
More information about the wpkg-announce
mailing list