Give the changelog some more love

Add references to bug numbers and be more verbose.

1 files changed, 24 insertions, 14 deletions

diff --git a/CHANGELOG b/CHANGELOG
index 7df1d3d..1aca43f 100644
--- a/CHANGELOG
+++ b/CHANGELOG
@@ -4,37 +4,47 @@ Version 0.8.0   - UNRELEASED
     when writing From_ line timestamps, don't use UTC but local time, without
     timezone information.
   * To determine the delivery date of a message, archivemail now looks for the
-    timestamp of the latest 'Received' header before resorting to 'Date' or
-    'Resent-Date'.  This should give much better results when there is no
-    'Delivery-date' header.  (Thanks Andrew Ferrier & Christian Brabandt)
+    timestamp of the latest 'Received' header before resorting to
+    'Resent-Date' or 'Date'.  This should give much better results when there
+    is no 'Delivery-date' header, which is still checked first.
+    (Thanks Andrew Ferrier & Christian Brabandt)
     Closes: #1481316, #1764855, Debian bug #272666.
   * If present, the 'Resent-date' header now takes precedence over 'Date'.
   * IMAP: recognize when a server advertises LOGINDISABLED.
   * New option --debug-imap; this just sets imaplib.Debug, thereby dumping
     archivemail's chat with the server to stdout.
-  * Fixed crash with Python 2.5 when archiving an empty maildir. (Thanks
-    "Aidant") Closes: #1878940.
+  * Fixed crash with Python 2.5 when archiving an empty maildir.
+    (Thanks "Aidant")  Closes: #1878940.
   * New option --all to archive all messages in a mailbox.  Closes: #1764846.
   * Fixed a crash when archiving maildirs with --days=0. (Thanks John Goerzen)
   * IMAP: automatically add NAMESPACE prefix to a mailbox path if necessary.
   * Removed the feature to setuid to the mailbox owners when run as root.
     This was a bad idea; it's really hard to do safely, if at all possible.
+    Obsoletes: patch #2783134.
   * Replaced some simple minded file operation security checks with more
-    decent ones.
+    decent ones.  This means we can safely operate in /tmp, for example.  The
+    price is that we no longer accept symlinked files.  Obsoletes: patch
+    #1874868.
   * Don't use rename() to update mbox files and the archive, but write the
     files directly.  This is more fragile, but required for correct mbox
     locking, and also for mboxes in mail spool directories where we don't have
-    permission to create files.
-  * Fixed the test suite to deal with nanosecond file timestamps.
+    permission to create files.  It also means that if selinux is enabled,
+    archivemail now preserves the selinux security context of an mbox.
+    Closes: #2210732.
+  * Fixed the test suite to deal with nanosecond file timestamps.  These are
+    provided by ext4 and XFS, for example.  Closes: #2043900.
   * Cleaned up the test suite, replacing a lot of duplicated code and avoiding
     a lot of redundand testing.  This speeds up the test suite by a factor of
-    20 or so.
-  * mbox locking: completely rewritten.  Switched from flock to fcntl locking,
-    which is NFS-safe, and we now lock with fcntl first, then with a dotlock,
-    instead of the other way around.  (This is makes archivemail compatible
-    with Debian systems. ;)
+    15 or so.
+  * mbox locking got completely rewritten.  Switched from flock to lockf
+    locking, which is NFS-safe and portable, and we now lock with lockf first,
+    then with a dotlock, instead of the other way around.  (This is makes
+    archivemail compatible with Debian systems. ;)
   * We now omit the dotlock if we don't have sufficient permissions to create
-    it in the mbox directory. Closes: #855269.
+    it in the mbox directory. (The file is still locked with lockf.)
+    Since we also no longer use rename() to commit changes to an mbox, (see
+    above) this means archivemail can now operate on mbox files in the system
+    mail spool.  Closes: #855269.
   * Refactoring of the mbox classes; much of the code got rewritten.
   * The archive now also gets locked while archivemail updates it.
   * Various Python language fixes (for example don't use "0" and "1" as