Collected bugfixes, improvements, and cleanups.

 - Do not leak file descriptors when merging new include/exclude rules.
 - Improved error message when unable to open a directory.
 - Delete obsolete clean-segments.pl script (long since replaced by
   functionality in cumulus-util).

diff --git a/contrib/clean-segments.pl b/contrib/clean-segments.pl
deleted file mode 100755 (executable)
index 8f9aa80..0000000
--- a/contrib/clean-segments.pl
+++ /dev/null
@@ -1,79 +0,0 @@
-#!/usr/bin/perl -w
-#
-# Garbage collect segments in LBS snapshot directories.
-#
-# Find all segments which are not referenced by any current snapshots and print
-# out a listing of them so that they can be deleted.
-#
-# Takes no command-line arguments, and expects to be invoked from the directory
-# containing the snapshots.
-#
-# Copyright (C) 2007  Michael Vrable
-
-use strict;
-
-my $SEGMENT_PATTERN
-    = '[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}';
-
-# Set of all segments which are used by current snapshots.  Value is ignored.
-my %segments_used = ();
-
-# Iterate through all snapshots.  Snapshot descriptors should end with a ".lbs"
-# extension.  Find all segments which are used.
-foreach (glob "*.lbs") {
-    open DESCRIPTOR, "<", $_
-        or die "Cannot open backup descriptor file $_: $!";
-
-    # Parse the backup descriptor file.  We might not need the full parser, but
-    # it shouldn't hurt.
-    my %descriptor = ();
-    my ($line, $last_key);
-    while (defined($line = <DESCRIPTOR>)) {
-        chomp $line;
-
-        if ($line =~ m/^([-\w]+):\s*(.*)$/) {
-            $descriptor{$1} = $2;
-            $last_key = $1;
-        } elsif ($line =~/^\s/ && defined $last_key) {
-            $descriptor{$last_key} .= $line;
-        } else {
-            undef $last_key;
-            print STDERR "Ignoring line in backup descriptor: $line\n";
-        }
-    }
-
-    # Extract the list of segments from the parsed descriptor file.
-    foreach (split /\s+/, $descriptor{Segments}) {
-        next unless $_;
-        if (m/^$SEGMENT_PATTERN$/) {
-            $segments_used{$_} = 1;
-        } else {
-            warn "Invalid segment name: '$_'\n";
-        }
-    }
-
-    close DESCRIPTOR;
-}
-
-# Look for all segments in this directory, and match them against the list
-# generated above of segments which are used.  Pring out any segments which are
-# not used.
-my %segments_found = ();
-
-foreach (glob "*") {
-    if (m/^($SEGMENT_PATTERN)(\.\S+)?$/) {
-        $segments_found{$1} = 1;
-        if (!exists $segments_used{$1}) {
-            print $_, "\n";
-        }
-    }
-}
-
-# Perform a consistency check: were any segments referenced by snapshot but not
-# found in the directory?
-foreach (sort keys %segments_used) {
-    if (!exists $segments_found{$_}) {
-        print STDERR "Warning: Segment $_ not found\n";
-    }
-}
-

diff --git a/main.cc b/main.cc
index 6b9f457..a8e345f 100644 (file)
--- a/main.cc
+++ b/main.cc
@@ -541,6 +541,7 @@ void try_merge_filter(const string& path, const string& basedir)
      * one block (1 MB) worth of data.  If the file doesn't seems like it might
      * be larger than that, don't parse the rules in it. */
     ssize_t bytes = file_read(fd, block_buf, LBS_BLOCK_SIZE);
      * one block (1 MB) worth of data.  If the file doesn't seems like it might
      * be larger than that, don't parse the rules in it. */
     ssize_t bytes = file_read(fd, block_buf, LBS_BLOCK_SIZE);

+    close(fd);

     if (bytes < 0 || bytes >= static_cast<ssize_t>(LBS_BLOCK_SIZE - 1)) {
         /* TODO: Add more strict resource limits on merge files? */
         fprintf(stderr,
     if (bytes < 0 || bytes >= static_cast<ssize_t>(LBS_BLOCK_SIZE - 1)) {
         /* TODO: Add more strict resource limits on merge files? */
         fprintf(stderr,


@@ -585,7 +586,8 @@ void scanfile(const string& path)
         DIR *dir = opendir(path.c_str());
 
         if (dir == NULL) {
         DIR *dir = opendir(path.c_str());
 
         if (dir == NULL) {

-            fprintf(stderr, "Error: %m\n");
+            fprintf(stderr, "Error reading directory %s: %m\n",
+                    path.c_str());

             return;
         }
 
             return;
         }