From: Eric Wong <e@80x24.org>
Date: Tue, 21 May 2019 23:22:05 +0000 (+0000)
Subject: Merge remote-tracking branch 'origin/xap-optional' into master
X-Git-Tag: v1.2.0~269
X-Git-Url: http://www.git.stargrave.org/?p=public-inbox.git;a=commitdiff_plain;h=e220b8b2ee5cfd458167dc2c6c92726352c4c80e;hp=d0e8bfd866ed1e924e8d9f551939eecbea4920ef

Merge remote-tracking branch 'origin/xap-optional' into master

* origin/xap-optional:
  admin: improve warnings and errors for missing modules
  searchidx: do not create empty Xapian partitions for basic
  lazy load Xapian and make it optional for v2
  www: use Inbox->over where appropriate
  nntp: use Inbox->over directly
  inbox: add ->over method to ease access
---

diff --git a/Documentation/include.mk b/Documentation/include.mk
index 2a02611a..6415338b 100644
--- a/Documentation/include.mk
+++ b/Documentation/include.mk
@@ -10,6 +10,7 @@ RSYNC = rsync
 RSYNC_DEST = public-inbox.org:/srv/public-inbox/
 txt := INSTALL README COPYING TODO HACKING
 dtxt := design_notes.txt design_www.txt dc-dlvr-spam-flow.txt hosted.txt
+dtxt += marketing.txt
 dtxt += standards.txt
 dtxt := $(addprefix Documentation/, $(dtxt))
 docs := $(txt) $(dtxt)
diff --git a/Documentation/marketing.txt b/Documentation/marketing.txt
new file mode 100644
index 00000000..385e5172
--- /dev/null
+++ b/Documentation/marketing.txt
@@ -0,0 +1,27 @@
+marketing guide for public-inbox
+
+TL; DR: Don't market this.
+
+If you must: don't be pushy and annoying about it.  Slow down.
+Please no superlatives, hype or BS.
+
+It's online and public, so it already markets itself.
+Being informative is not a bad thing, being insistent is.
+
+Chances are, you're preaching to the choir; or the folks you're
+trying to convince are not ready for everything our project
+represents to the resistance against centralization.
+
+Baby steps...
+
+There's never a need for anybody to migrate to using our
+software, or to use any particular instance of it.  It's
+designed to coexist with other mail archives, especially
+other installations of public-inbox.
+
+Most importantly, we take victories even when our software
+doesn't get adopted.  Freedom from lock-in is more important
+than the adoption of any software.
+
+Every time somebody recognizes and rejects various forms of
+lock-in and centralization is already a victory for us.
diff --git a/INSTALL b/INSTALL
index 72e0763c..0246299b 100644
--- a/INSTALL
+++ b/INSTALL
@@ -36,11 +36,6 @@ Beyond that, there is a long list of Perl modules required, starting with:
                                    pkg: p5-TimeDate
                                    rpm: perl-TimeDate
 
-* Devel::Peek                      deb: libperl5.$MINOR (e.g. libperl5.24)
-                                   pkg: perl5
-                                   rpm: perl-Devel-Peek
-                                   (typically installed alongside Perl5)
-
 * Email::MIME                      deb: libemail-mime-perl
                                    pkg: p5-Email-MIME
                                    rpm: perl-Email-MIME
@@ -125,6 +120,12 @@ above, so there is no need to explicitly install them:
                                    rpm: perl-DBI
                                    (pulled in by DBD::SQLite)
 
+* Devel::Peek                      deb: libperl5.$MINOR (e.g. libperl5.24)
+                                   pkg: perl5
+                                   rpm: perl-Devel-Peek
+                                   (optional for stale FD cleanup in daemons,
+                                    typically installed alongside Perl5)
+
 - Filesys::Notify::Simple          deb: libfilesys-notify-simple-perl
                                    pkg: pkg-Filesys-Notify-Simple
                                    rpm: perl-Filesys-Notify-Simple
diff --git a/MANIFEST b/MANIFEST
index 1da40a90..2c356c62 100644
--- a/MANIFEST
+++ b/MANIFEST
@@ -7,6 +7,7 @@ Documentation/design_notes.txt
 Documentation/design_www.txt
 Documentation/hosted.txt
 Documentation/include.mk
+Documentation/marketing.txt
 Documentation/public-inbox-compact.pod
 Documentation/public-inbox-config.pod
 Documentation/public-inbox-convert.pod
diff --git a/Makefile.PL b/Makefile.PL
index 6be913b1..de0c49fd 100644
--- a/Makefile.PL
+++ b/Makefile.PL
@@ -38,10 +38,6 @@ WriteMakefile(
 		# `perl5' on FreeBSD
 		'Encode' => 0,
 
-		# libperl$PERL_VERSION on Debian, `perl5' on FreeBSD,
-		# but Fedora seems to need this separately
-		'Devel::Peek' => 0,
-
 		# TODO: these should really be made optional...
 		'Plack' => 0,
 		'URI::Escape' => 0,
diff --git a/lib/PublicInbox/Daemon.pm b/lib/PublicInbox/Daemon.pm
index 68ba9876..227ba5f9 100644
--- a/lib/PublicInbox/Daemon.pm
+++ b/lib/PublicInbox/Daemon.pm
@@ -13,6 +13,7 @@ use Time::HiRes qw(clock_gettime CLOCK_MONOTONIC);
 STDOUT->autoflush(1);
 STDERR->autoflush(1);
 require PublicInbox::DS;
+require PublicInbox::EvCleanup;
 require POSIX;
 require PublicInbox::Listener;
 require PublicInbox::ParentPipe;
@@ -463,6 +464,7 @@ sub master_loop {
 
 sub daemon_loop ($$) {
 	my ($refresh, $post_accept) = @_;
+	PublicInbox::EvCleanup::enable(); # early for $refresh
 	my $parent_pipe;
 	if ($worker_processes > 0) {
 		$refresh->(); # preload by default
@@ -485,7 +487,6 @@ sub daemon_loop ($$) {
 	@listeners = map {
 		PublicInbox::Listener->new($_, $post_accept)
 	} @listeners;
-	PublicInbox::EvCleanup::enable();
 	PublicInbox::DS->EventLoop;
 	$parent_pipe = undef;
 }
diff --git a/lib/PublicInbox/HTTP.pm b/lib/PublicInbox/HTTP.pm
index 11bd241e..10e6d6a4 100644
--- a/lib/PublicInbox/HTTP.pm
+++ b/lib/PublicInbox/HTTP.pm
@@ -18,6 +18,7 @@ use Plack::HTTPParser qw(parse_http_request); # XS or pure Perl
 use HTTP::Status qw(status_message);
 use HTTP::Date qw(time2str);
 use IO::Handle;
+require PublicInbox::EvCleanup;
 use constant {
 	CHUNK_START => -1,   # [a-f0-9]+\r\n
 	CHUNK_END => -2,     # \r\n
diff --git a/lib/PublicInbox/Import.pm b/lib/PublicInbox/Import.pm
index 12abf399..81a38fb6 100644
--- a/lib/PublicInbox/Import.pm
+++ b/lib/PublicInbox/Import.pm
@@ -367,10 +367,14 @@ sub add {
 	my @ct = msg_timestamp($hdr);
 	my $author_time_raw = git_timestamp(@at);
 	my $commit_time_raw = git_timestamp(@ct);
+
 	my $subject = $mime->header('Subject');
 	$subject = '(no subject)' unless defined $subject;
-	my $path_type = $self->{path_type};
+	# Mime decoding can create nulls replace them with spaces to protect git
+	$subject =~ tr/\0/ /;
+	utf8::encode($subject);
 
+	my $path_type = $self->{path_type};
 	my $path;
 	if ($path_type eq '2/38') {
 		$path = mid2path(v1_mid0($mime));
@@ -411,9 +415,6 @@ sub add {
 		print $w "reset $ref\n" or wfail;
 	}
 
-	# Mime decoding can create nulls replace them with spaces to protect git
-	$subject =~ tr/\0/ /;
-	utf8::encode($subject);
 	print $w "commit $ref\nmark :$commit\n",
 		"author $name <$email> $author_time_raw\n",
 		"committer $self->{ident} $commit_time_raw\n" or wfail;
diff --git a/lib/PublicInbox/Inbox.pm b/lib/PublicInbox/Inbox.pm
index 813ed997..0d86771f 100644
--- a/lib/PublicInbox/Inbox.pm
+++ b/lib/PublicInbox/Inbox.pm
@@ -7,26 +7,30 @@ use strict;
 use warnings;
 use PublicInbox::Git;
 use PublicInbox::MID qw(mid2path);
-use Devel::Peek qw(SvREFCNT);
 use PublicInbox::MIME;
-use POSIX qw(strftime);
 
+# Long-running "git-cat-file --batch" processes won't notice
+# unlinked packs, so we need to restart those processes occasionally.
+# Xapian and SQLite file handles are mostly stable, but sometimes an
+# admin will attempt to replace them atomically after compact/vacuum
+# and we need to be prepared for that.
 my $cleanup_timer;
-eval {
-	$cleanup_timer = 'disabled';
-	require PublicInbox::EvCleanup;
-	$cleanup_timer = undef; # OK if we get here
-};
-my $cleanup_broken = $@;
-
+my $cleanup_avail = -1; # 0, or 1
+my $have_devel_peek;
 my $CLEANUP = {}; # string(inbox) -> inbox
 sub cleanup_task () {
 	$cleanup_timer = undef;
 	my $next = {};
 	for my $ibx (values %$CLEANUP) {
 		my $again;
-		foreach my $f (qw(mm search over)) {
-			delete $ibx->{$f} if SvREFCNT($ibx->{$f}) == 1;
+		if ($have_devel_peek) {
+			foreach my $f (qw(mm search over)) {
+				# we bump refcnt by assigning tmp, here:
+				my $tmp = $ibx->{$f} or next;
+				next if Devel::Peek::SvREFCNT($tmp) > 2;
+				delete $ibx->{$f};
+				# refcnt is zero when tmp is out-of-scope
+			}
 		}
 		my $expire = time - 60;
 		if (my $git = $ibx->{git}) {
@@ -37,16 +41,31 @@ sub cleanup_task () {
 				$again = 1 if $git->cleanup($expire);
 			}
 		}
-		$again ||= !!($ibx->{over} || $ibx->{mm} || $ibx->{search});
+		if ($have_devel_peek) {
+			$again ||= !!($ibx->{over} || $ibx->{mm} ||
+			              $ibx->{search});
+		}
 		$next->{"$ibx"} = $ibx if $again;
 	}
 	$CLEANUP = $next;
 }
 
+sub cleanup_possible () {
+	# no need to require EvCleanup, here, if it were enabled another
+	# module would've require'd it, already
+	eval { PublicInbox::EvCleanup::enabled() } or return 0;
+
+	eval {
+		require Devel::Peek; # needs separate package in Fedora
+		$have_devel_peek = 1;
+	};
+	1;
+}
+
 sub _cleanup_later ($) {
 	my ($self) = @_;
-	return if $cleanup_broken;
-	return unless PublicInbox::EvCleanup::enabled();
+	$cleanup_avail = cleanup_possible() if $cleanup_avail < 0;
+	return if $cleanup_avail != 1;
 	$cleanup_timer ||= PublicInbox::EvCleanup::later(*cleanup_task);
 	$CLEANUP->{"$self"} = $self;
 }
diff --git a/lib/PublicInbox/Search.pm b/lib/PublicInbox/Search.pm
index b1e62f4c..eae10d8e 100644
--- a/lib/PublicInbox/Search.pm
+++ b/lib/PublicInbox/Search.pm
@@ -24,6 +24,10 @@ sub load_xapian () {
 
 		# n.b. FLAG_PURE_NOT is expensive not suitable for a public
 		# website as it could become a denial-of-service vector
+		# FLAG_PHRASE also seems to cause performance problems
+		# sometimes.
+		# TODO: make this an option, maybe?
+		# or make indexlevel=medium as default
 		FLAG_PHRASE()|FLAG_BOOLEAN()|FLAG_LOVEHATE()|FLAG_WILDCARD();
 	};
 };
diff --git a/lib/PublicInbox/SearchIdx.pm b/lib/PublicInbox/SearchIdx.pm
index f96f0d03..114420e4 100644
--- a/lib/PublicInbox/SearchIdx.pm
+++ b/lib/PublicInbox/SearchIdx.pm
@@ -19,7 +19,6 @@ use POSIX qw(strftime);
 use PublicInbox::OverIdx;
 use PublicInbox::Spawn qw(spawn);
 use PublicInbox::Git qw(git_unquote);
-use Compress::Zlib qw(compress);
 
 use constant {
 	BATCH_BYTES => defined($ENV{XAPIAN_FLUSH_THRESHOLD}) ?
diff --git a/lib/PublicInbox/WWW.pm b/lib/PublicInbox/WWW.pm
index 8e1b1afe..b6f18f8d 100644
--- a/lib/PublicInbox/WWW.pm
+++ b/lib/PublicInbox/WWW.pm
@@ -59,14 +59,14 @@ sub call {
 	my $ctx = { env => $env, www => $self };
 
 	# we don't care about multi-value
-	my %qp = map {
+	%{$ctx->{qp}} = map {
 		utf8::decode($_);
-		my ($k, $v) = split('=', uri_unescape($_), 2);
-		$v = '' unless defined $v;
-		$v =~ tr/+/ /;
-		($k, $v)
+		tr/+/ /;
+		my ($k, $v) = split('=', $_, 2);
+		$v = uri_unescape($v // '');
+		# none of the keys we care about will need escaping
+		$k => $v;
 	} split(/[&;]+/, $env->{QUERY_STRING});
-	$ctx->{qp} = \%qp;
 
 	# avoiding $env->{PATH_INFO} here since that's already decoded
 	my ($path_info) = ($env->{REQUEST_URI} =~ path_re($env));
diff --git a/t/config_limiter.t b/t/config_limiter.t
index 85a71257..b18951a6 100644
--- a/t/config_limiter.t
+++ b/t/config_limiter.t
@@ -38,7 +38,6 @@ my $cfgpfx = "publicinbox.test";
 	ok($lim, 'Limiter exists');
 	is($lim->{max}, 3, 'limiter has expected slots');
 	$ibx->{git} = undef;
-	PublicInbox::Inbox::cleanup_task;
 	my $new = $ibx->git;
 	isnt($old, "$new", 'got new Git object');
 	is("$new->{-httpbackend_limiter}", "$lim", 'same limiter');