[mirrorbrain-commits] r7788 - in /trunk/docs: changes.rst configuration_reference.rst

From: <poeml_at_mirrorbrain.org>
Date: Fri, 04 Sep 2009 16:32:12 -0000
Author: poeml
Date: Fri Sep  4 18:32:09 2009
New Revision: 7788

URL: http://svn.mirrorbrain.org/viewvc/mirrorbrain?rev=7788&view=rev
Log:
docs: remove empty configuration_reference.rst

Removed:
    trunk/docs/configuration_reference.rst
Modified:
    trunk/docs/changes.rst

Modified: trunk/docs/changes.rst
URL: http://svn.mirrorbrain.org/viewvc/mirrorbrain/trunk/docs/changes.rst?rev=7788&r1=7787&r2=7788&view=diff
==============================================================================
--- trunk/docs/changes.rst (original)
+++ trunk/docs/changes.rst Fri Sep  4 18:32:09 2009
@@ -2,6 +2,75 @@
 
 Release Notes/Change History
 ============================
+
+
+Release 2.10.0 (Sep 4, 2009)
+----------------------------
+
+* The cache of metalink hashes, as created by the :program:`metalink-hasher`,
+  was changed to more reliably detect changes in the origin files. So far, the
+  file modification time was the criterion to invalidate cached hashes. When
+  files were replaced with *older* versions (version with smaller mtime), this
+  wasn't detected, and a cached hash would not be correctly invalidated.
+  https://bugzilla.novell.com/536495 reports this of being an issue.
+  
+  To fix this, the cache now also uses the file inode as criterion.
+
+  :program:`mod_mirrorbrain` was updated to use the new inode-wise metalink
+  hashes is used. At the same time, it knows how to use the previous scheme as
+  fallback. 
+  
+  Thus, the transition should be seamless, and no special steps should be
+  required when upgrading. Note however that all hashes are regenerated, which
+  could take a while for large file trees, and which could lead to cron jobs
+  stacking up.
+
+
+* There were a number of enhancements, and small bug fixes, in the
+  :program:`mb` tool (and accompanying Python module):
+
+  - :program:`mb new`:
+  
+    - When adding new mirrors, the hostname part in the HTTP base URL might
+      contain a port number. This is now recognized correctly, so the DNS
+      lookup, GeoIP lookup and ASN lookup for the hostname string can work.
+    - The commandline options ``--region-only``, ``--country-only``,
+      ``--as-only``, ``--prefix-only`` were added, each setting the respective
+      flag.
+    - The commandline options ``--operator-name`` and ``--operator-url`` were
+      added.
+    - The ``--score`` option is depreciated, since it has been renamed it to
+      ``--prio``.
+  
+  - :program:`mb scan`:
+  
+    - The passing of arguments to the scanner script was fixed in the case
+      where the ``-j`` (``--jobs``) option was used together with mirror
+      identifier specified on the commandline.
+
+  - :program:`mb list`:
+
+    - Command line options to display the boolean flags were added:
+      ``--region-only``, ``--country-only``, ``--as-only`` and
+      ``--prefix-only``.
+
+  - :program:`mb scan` / :program:`mb file ls --probe`:
+
+    - the lookup whether the :mod:`multiprocessing` or :mod:`processing` module
+      exist was fixed: it could print a false warning that none of them was
+      installed.
+
+* The :program:`mirrorprobe` program no longer logs to the console (stderr).
+  This allows for running the script without redirection its output to
+  :file:`/dev/null` — which could mean swallowing important errors in the end.
+
+  A scenario was documented where the mirrorprobe could fail on machines with
+  little memories and many mirrors to check. The fix is to properly set ulimit
+  to allow a large enough stack size.
+
+  Error handling was cleaned up; more errors are handled (e.g. socket timeouts
+  during response reading) and logged properly; and for exceptions yet
+  unhandled, info about the mirror that caused them is printed.
 
 
 Release 2.9.2 (Aug 21, 2009)




_______________________________________________
mirrorbrain-commits mailing list
Archive: http://mirrorbrain.org/archive/mirrorbrain-commits/

Note: To remove yourself from this list, send a mail with the content
 	unsubscribe
to the address mirrorbrain-commits-request_at_mirrorbrain.org
Received on Fri Sep 04 2009 - 16:32:16 GMT

This archive was generated by hypermail 2.2.0 : Fri Sep 04 2009 - 16:45:09 GMT