-# Copyright (C) 2014-2018 all contributors <meta@public-inbox.org>
+# Copyright (C) 2014-2020 all contributors <meta@public-inbox.org>
# License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
#
# Main web interface for mailing list archives
# We focus on the lowest common denominators here:
# - targeted at text-only console browsers (w3m, links, etc..)
# - Only basic HTML, CSS only for line-wrapping <pre> text content for GUIs
+# and diff/syntax-highlighting (optional)
# - No JavaScript, graphics or icons allowed.
# - Must not rely on static content
# - UTF-8 is only for user-content, 7-bit US-ASCII for us
package PublicInbox::WWW;
-use 5.008;
+use 5.010_001;
use strict;
use warnings;
+use bytes (); # only for bytes::length
use PublicInbox::Config;
use PublicInbox::Hval;
use URI::Escape qw(uri_unescape);
use PublicInbox::MID qw(mid_escape);
-require PublicInbox::Git;
use PublicInbox::GitHTTPBackend;
-our $INBOX_RE = qr!\A/([\w\.\-]+)!;
+use PublicInbox::UserContent;
+use PublicInbox::WwwStatic qw(r path_info_raw);
+
+# TODO: consider a routing tree now that we have more endpoints:
+our $INBOX_RE = qr!\A/([\w\-][\w\.\-]*)!;
our $MID_RE = qr!([^/]+)!;
our $END_RE = qr!(T/|t/|t\.mbox(?:\.gz)?|t\.atom|raw|)!;
-our $ATTACH_RE = qr!(\d[\.\d]*)-([[:alnum:]][\w\.-]+[[:alnum:]])!i;
+our $ATTACH_RE = qr!([0-9][0-9\.]*)-($PublicInbox::Hval::FN)!;
+our $OID_RE = qr![a-f0-9]{7,40}!;
sub new {
my ($class, $pi_config) = @_;
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;
- my $path_info = $env->{PATH_INFO};
+ my $path_info = path_info_raw($env);
my $method = $env->{REQUEST_METHOD};
if ($method eq 'POST') {
- if ($path_info =~ m!$INBOX_RE/(git-upload-pack)\z!) {
- my $path = $2;
+ if ($path_info =~ m!$INBOX_RE/(?:(?:git/)?([0-9]+)(?:\.git)?/)?
+ (git-upload-pack)\z!x) {
+ my ($epoch, $path) = ($2, $3);
return invalid_inbox($ctx, $1) ||
- serve_git($ctx, $path);
+ serve_git($ctx, $epoch, $path);
} elsif ($path_info =~ m!$INBOX_RE/!o) {
return invalid_inbox($ctx, $1) || mbox_results($ctx);
}
}
- elsif ($method !~ /\AGET|HEAD\z/) {
- return r(405, 'Method Not Allowed');
+ elsif ($method !~ /\A(?:GET|HEAD)\z/) {
+ return r(405);
}
# top-level indices and feeds
- if ($path_info eq '/') {
- r404();
+ if ($path_info eq '/' || $path_info eq '/manifest.js.gz') {
+ www_listing($self)->call($env);
} elsif ($path_info =~ m!$INBOX_RE\z!o) {
invalid_inbox($ctx, $1) || r301($ctx, $1);
} elsif ($path_info =~ m!$INBOX_RE(?:/|/index\.html)?\z!o) {
invalid_inbox($ctx, $1) || get_atom($ctx);
} elsif ($path_info =~ m!$INBOX_RE/new\.html\z!o) {
invalid_inbox($ctx, $1) || get_new($ctx);
- } elsif ($path_info =~ m!$INBOX_RE/
+ } elsif ($path_info =~ m!$INBOX_RE/description\z!o) {
+ get_description($ctx, $1);
+ } elsif ($path_info =~ m!$INBOX_RE/(?:(?:git/)?([0-9]+)(?:\.git)?/)?
($PublicInbox::GitHTTPBackend::ANY)\z!ox) {
- my $path = $2;
- invalid_inbox($ctx, $1) || serve_git($ctx, $path);
- } elsif ($path_info =~ m!$INBOX_RE/([\w-]+).mbox\.gz\z!o) {
+ my ($epoch, $path) = ($2, $3);
+ invalid_inbox($ctx, $1) || serve_git($ctx, $epoch, $path);
+ } elsif ($path_info =~ m!$INBOX_RE/([a-zA-Z0-9_\-]+).mbox\.gz\z!o) {
serve_mbox_range($ctx, $1, $2);
} elsif ($path_info =~ m!$INBOX_RE/$MID_RE/$END_RE\z!o) {
msg_page($ctx, $1, $2, $3);
invalid_inbox_mid($ctx, $1, $2) || get_attach($ctx, $idx, $fn);
# in case people leave off the trailing slash:
} elsif ($path_info =~ m!$INBOX_RE/$MID_RE/(T|t)\z!o) {
- my ($inbox, $mid, $suffix) = ($1, $2, $3);
+ my ($inbox, $mid_ue, $suffix) = ($1, $2, $3);
$suffix .= $suffix =~ /\A[tT]\z/ ? '/#u' : '/';
- r301($ctx, $inbox, $mid, $suffix);
+ r301($ctx, $inbox, $mid_ue, $suffix);
} elsif ($path_info =~ m!$INBOX_RE/$MID_RE/R/?\z!o) {
- my ($inbox, $mid) = ($1, $2);
- r301($ctx, $inbox, $mid, '#R');
+ my ($inbox, $mid_ue) = ($1, $2);
+ r301($ctx, $inbox, $mid_ue, '#R');
} elsif ($path_info =~ m!$INBOX_RE/$MID_RE/f/?\z!o) {
r301($ctx, $1, $2);
} elsif ($path_info =~ m!$INBOX_RE/_/text(?:/(.*))?\z!o) {
get_text($ctx, $1, $2);
-
+ } elsif ($path_info =~ m!$INBOX_RE/([a-zA-Z0-9_\-\.]+)\.css\z!o) {
+ get_css($ctx, $1, $2);
+ } elsif ($path_info =~ m!$INBOX_RE/manifest\.js\.gz\z!o) {
+ get_inbox_manifest($ctx, $1, $2);
+ } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s/\z!o) {
+ get_vcs_object($ctx, $1, $2);
+ } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s/
+ ($PublicInbox::Hval::FN)\z!ox) {
+ get_vcs_object($ctx, $1, $2, $3);
+ } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s\z!o) {
+ r301($ctx, $1, $2, 's/');
# convenience redirects order matters
} elsif ($path_info =~ m!$INBOX_RE/([^/]{2,})\z!o) {
r301($ctx, $1, $2);
}
}
-# for CoW-friendliness, MOOOOO!
+# for CoW-friendliness, MOOOOO! Even for single-process setups,
+# we want to get all immortal allocations done early to avoid heap
+# fragmentation since common allocators favor a large contiguous heap.
sub preload {
+ my ($self) = @_;
+ require PublicInbox::ExtMsg;
require PublicInbox::Feed;
require PublicInbox::View;
require PublicInbox::SearchThread;
require PublicInbox::MIME;
- require Digest::SHA;
- require POSIX;
-
- foreach (qw(PublicInbox::Search PublicInbox::SearchView
- PublicInbox::Mbox IO::Compress::Gzip
- PublicInbox::NewsWWW)) {
+ require PublicInbox::Mbox;
+ require PublicInbox::ViewVCS;
+ require PublicInbox::WwwText;
+ require PublicInbox::WwwAttach;
+ eval {
+ require PublicInbox::Search;
+ PublicInbox::Search::load_xapian();
+ };
+ foreach (qw(PublicInbox::SearchView PublicInbox::MboxGz)) {
eval "require $_;";
}
+ if (ref($self)) {
+ my $pi_config = $self->{pi_config};
+ if (defined($pi_config->{'publicinbox.cgitrc'})) {
+ $pi_config->limiter('-cgit');
+ }
+ $self->cgit;
+ $self->stylesheets_prepare($_) for ('', '../', '../../');
+ $self->www_listing;
+ $self->news_www;
+ $pi_config->each_inbox(\&preload_inbox);
+ }
+}
+
+sub preload_inbox {
+ my $ibx = shift;
+ $ibx->cloneurl;
+ $ibx->description;
+ $ibx->base_url;
}
# private functions below
my ($ctx) = @_;
if ($ctx && $ctx->{mid}) {
require PublicInbox::ExtMsg;
- searcher($ctx);
return PublicInbox::ExtMsg::ext_msg($ctx);
}
- r(404, 'Not Found');
+ r(404);
}
-# simple response for errors
-sub r { [ $_[0], ['Content-Type' => 'text/plain'], [ join(' ', @_, "\n") ] ] }
+sub news_cgit_fallback ($) {
+ my ($ctx) = @_;
+ my $www = $ctx->{www};
+ my $env = $ctx->{env};
+ my $res = $www->news_www->call($env);
+ $res->[0] == 404 ? $www->cgit->call($env) : $res;
+}
# returns undef if valid, array ref response if invalid
sub invalid_inbox ($$) {
my ($ctx, $inbox) = @_;
- my $www = $ctx->{www};
- my $obj = $www->{pi_config}->lookup_name($inbox);
- if (defined $obj) {
- $ctx->{git} = $obj->git;
- $ctx->{-inbox} = $obj;
+ my $ibx = $ctx->{www}->{pi_config}->lookup_name($inbox);
+ if (defined $ibx) {
+ $ctx->{-inbox} = $ibx;
return;
}
# generation and link things intended for nntp:// to https?://,
# so try to infer links and redirect them to the appropriate
# list URL.
- $www->news_www->call($ctx->{env});
+ news_cgit_fallback($ctx);
}
# returns undef if valid, array ref response if invalid
sub invalid_inbox_mid {
- my ($ctx, $inbox, $mid) = @_;
+ my ($ctx, $inbox, $mid_ue) = @_;
my $ret = invalid_inbox($ctx, $inbox);
return $ret if $ret;
- $ctx->{mid} = $mid;
- if ($mid =~ /\A[a-f0-9]{40}\z/) {
- # this is horiffically wasteful for legacy URLs:
- if ($mid = mid2blob($ctx)) {
- require Email::Simple;
- use PublicInbox::MID qw/mid_clean/;
- my $s = Email::Simple->new($mid);
- $ctx->{mid} = mid_clean($s->header('Message-ID'));
- }
+ my $mid = $ctx->{mid} = uri_unescape($mid_ue);
+ my $ibx = $ctx->{-inbox};
+ if ($mid =~ m!\A([a-f0-9]{2})([a-f0-9]{38})\z!) {
+ my ($x2, $x38) = ($1, $2);
+ # this is horrifically wasteful for legacy URLs:
+ my $str = $ctx->{-inbox}->msg_by_path("$x2/$x38") or return;
+ require Email::Simple;
+ my $s = Email::Simple->new($str);
+ $mid = PublicInbox::MID::mid_clean($s->header('Message-ID'));
+ return r301($ctx, $inbox, mid_escape($mid));
}
undef;
}
sub get_index {
my ($ctx) = @_;
require PublicInbox::Feed;
- searcher($ctx);
if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
require PublicInbox::SearchView;
PublicInbox::SearchView::sres_top_html($ctx);
}
}
-# just returns a string ref for the blob in the current ctx
-sub mid2blob {
- my ($ctx) = @_;
- $ctx->{-inbox}->msg_by_mid($ctx->{mid});
-}
-
# /$INBOX/$MESSAGE_ID/raw -> raw mbox
sub get_mid_txt {
my ($ctx) = @_;
sub get_mid_html {
my ($ctx) = @_;
require PublicInbox::View;
- searcher($ctx);
PublicInbox::View::msg_page($ctx) || r404($ctx);
}
# /$INBOX/$MESSAGE_ID/t/
sub get_thread {
my ($ctx, $flat) = @_;
- searcher($ctx) or return need_search($ctx);
+ $ctx->{-inbox}->over or return need($ctx, 'Overview');
$ctx->{flat} = $flat;
require PublicInbox::View;
PublicInbox::View::thread_html($ctx);
PublicInbox::WwwText::get_text($ctx, $key);
}
-sub ctx_get {
- my ($ctx, $key) = @_;
- my $val = $ctx->{$key};
- (defined $val && $val ne '') or die "BUG: bad ctx, $key unusable";
- $val;
-}
-
-# search support is optional, returns undef if Xapian is not installed
-# or not configured for the given GIT_DIR
-sub searcher {
- my ($ctx) = @_;
- eval {
- require PublicInbox::Search;
- $ctx->{srch} = $ctx->{-inbox}->search;
- };
+# show git objects (blobs and commits)
+# /$INBOX/_/$OBJECT_ID/show
+# /$INBOX/_/${OBJECT_ID}_${FILENAME}
+# KEY may contain slashes
+sub get_vcs_object ($$$;$) {
+ my ($ctx, $inbox, $oid, $filename) = @_;
+ my $r404 = invalid_inbox($ctx, $inbox);
+ return $r404 if $r404;
+ require PublicInbox::ViewVCS;
+ PublicInbox::ViewVCS::show($ctx, $oid, $filename);
}
-sub need_search {
- my ($ctx) = @_;
+sub need {
+ my ($ctx, $extra) = @_;
my $msg = <<EOF;
-<html><head><title>Search not available for this
-public-inbox</title><body><pre>Search is not available for this public-inbox
+<html><head><title>$extra not available for this
+public-inbox</title><body><pre>$extra is not available for this public-inbox
<a href="../">Return to index</a></pre></body></html>
EOF
[ 501, [ 'Content-Type' => 'text/html; charset=UTF-8' ], [ $msg ] ];
# especially on older systems. Stick to zlib since that's what git uses.
sub get_thread_mbox {
my ($ctx, $sfx) = @_;
- my $srch = searcher($ctx) or return need_search($ctx);
+ my $over = $ctx->{-inbox}->over or return need($ctx, 'Overview');
require PublicInbox::Mbox;
- PublicInbox::Mbox::thread_mbox($ctx, $srch, $sfx);
+ PublicInbox::Mbox::thread_mbox($ctx, $over, $sfx);
}
# /$INBOX/$MESSAGE_ID/t.atom -> thread as Atom feed
sub get_thread_atom {
my ($ctx) = @_;
- searcher($ctx) or return need_search($ctx);
+ $ctx->{-inbox}->over or return need($ctx, 'Overview');
require PublicInbox::Feed;
PublicInbox::Feed::generate_thread_atom($ctx);
}
} elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/f\z!o) {
r301($ctx, $1, $2);
} else {
- $ctx->{www}->news_www->call($ctx->{env});
+ news_cgit_fallback($ctx);
}
}
sub r301 {
- my ($ctx, $inbox, $mid, $suffix) = @_;
- my $obj = $ctx->{-inbox};
- unless ($obj) {
+ my ($ctx, $inbox, $mid_ue, $suffix) = @_;
+ my $ibx = $ctx->{-inbox};
+ unless ($ibx) {
my $r404 = invalid_inbox($ctx, $inbox);
return $r404 if $r404;
- $obj = $ctx->{-inbox};
+ $ibx = $ctx->{-inbox};
}
- my $url = $obj->base_url($ctx->{env});
+ my $url = $ibx->base_url($ctx->{env});
my $qs = $ctx->{env}->{QUERY_STRING};
- $url .= (mid_escape($mid) . '/') if (defined $mid);
+ if (defined $mid_ue) {
+ # common, and much nicer as '@' than '%40':
+ $mid_ue =~ s/%40/@/g;
+ $url .= $mid_ue . '/';
+ }
$url .= $suffix if (defined $suffix);
$url .= "?$qs" if $qs ne '';
}
sub msg_page {
- my ($ctx, $inbox, $mid, $e) = @_;
+ my ($ctx, $inbox, $mid_ue, $e) = @_;
my $ret;
- $ret = invalid_inbox_mid($ctx, $inbox, $mid) and return $ret;
+ $ret = invalid_inbox_mid($ctx, $inbox, $mid_ue) and return $ret;
'' eq $e and return get_mid_html($ctx);
'T/' eq $e and return get_thread($ctx, 1);
't/' eq $e and return get_thread($ctx);
}
sub serve_git {
- my ($ctx, $path) = @_;
- PublicInbox::GitHTTPBackend::serve($ctx->{env}, $ctx->{git}, $path);
+ my ($ctx, $epoch, $path) = @_;
+ my $env = $ctx->{env};
+ my $ibx = $ctx->{-inbox};
+ my $git = defined $epoch ? $ibx->git_epoch($epoch) : $ibx->git;
+ $git ? PublicInbox::GitHTTPBackend::serve($env, $git, $path) : r404();
}
sub mbox_results {
my ($ctx) = @_;
if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
- searcher($ctx) or return need_search($ctx);
+ $ctx->{-inbox}->search or return need($ctx, 'search');
require PublicInbox::SearchView;
return PublicInbox::SearchView::mbox_results($ctx);
}
my ($ctx, $inbox, $range) = @_;
invalid_inbox($ctx, $inbox) || eval {
require PublicInbox::Mbox;
- searcher($ctx);
PublicInbox::Mbox::emit_range($ctx, $range);
}
}
}
}
+sub cgit {
+ my ($self) = @_;
+ $self->{cgit} ||= do {
+ my $pi_config = $self->{pi_config};
+
+ if (defined($pi_config->{'publicinbox.cgitrc'})) {
+ require PublicInbox::Cgit;
+ PublicInbox::Cgit->new($pi_config);
+ } else {
+ require Plack::Util;
+ Plack::Util::inline_object(call => sub { r404() });
+ }
+ }
+}
+
+sub www_listing {
+ my ($self) = @_;
+ $self->{www_listing} ||= do {
+ require PublicInbox::WwwListing;
+ PublicInbox::WwwListing->new($self);
+ }
+}
+
+# GET $INBOX/manifest.js.gz
+sub get_inbox_manifest ($$$) {
+ my ($ctx, $inbox, $key) = @_;
+ my $r404 = invalid_inbox($ctx, $inbox);
+ return $r404 if $r404;
+ require PublicInbox::WwwListing;
+ PublicInbox::WwwListing::js($ctx->{env}, [$ctx->{-inbox}]);
+}
+
sub get_attach {
my ($ctx, $idx, $fn) = @_;
require PublicInbox::WwwAttach;
PublicInbox::WwwAttach::get_attach($ctx, $idx, $fn);
}
+# User-generated content (UGC) may have excessively long lines
+# and screw up rendering on some browsers, so we use pre-wrap.
+#
+# We also force everything to the same scaled font-size because GUI
+# browsers (tested both Firefox and surf (webkit)) uses a larger font
+# for the Search <form> element than the rest of the page. Font size
+# uniformity is important to people who rely on gigantic fonts.
+# Finally, we use monospace to ensure the Search field and button
+# has the same size and spacing as everything else which is
+# <pre>-formatted anyways.
+our $STYLE = 'pre{white-space:pre-wrap}*{font-size:100%;font-family:monospace}';
+
+sub stylesheets_prepare ($$) {
+ my ($self, $upfx) = @_;
+ my $mini = eval {
+ require CSS::Minifier;
+ sub { CSS::Minifier::minify(input => $_[0]) };
+ } || eval {
+ require CSS::Minifier::XS;
+ sub { CSS::Minifier::XS::minify($_[0]) };
+ } || sub { $_[0] };
+
+ my $css_map = {};
+ my $stylesheets = $self->{pi_config}->{css} || [];
+ my $links = [];
+ my $inline_ok = 1;
+
+ foreach my $s (@$stylesheets) {
+ my $attr = {};
+ local $_ = $s;
+ foreach my $k (qw(media title href)) {
+ if (s/\s*$k='([^']+)'// || s/\s*$k=(\S+)//) {
+ $attr->{$k} = $1;
+ }
+ }
+
+ if (defined $attr->{href}) {
+ $inline_ok = 0;
+ } else {
+ my $fn = $_;
+ my ($key) = (m!([^/]+?)(?:\.css)?\z!i);
+ if ($key !~ /\A[a-zA-Z0-9_\-\.]+\z/) {
+ warn "ignoring $fn, non-ASCII word character\n";
+ next;
+ }
+ open(my $fh, '<', $fn) or do {
+ warn "failed to open $fn: $!\n";
+ next;
+ };
+ my $ctime = 0;
+ my $local = do { local $/; <$fh> };
+ if ($local =~ /\S/) {
+ $ctime = sprintf('%x',(stat($fh))[10]);
+ $local = $mini->($local);
+ }
+
+ # do not let BOFHs override userContent.css:
+ if ($local =~ /!\s*important\b/i) {
+ warn "ignoring $fn since it uses `!important'\n";
+ next;
+ }
+
+ $css_map->{$key} = $local;
+ $attr->{href} = "$upfx$key.css?$ctime";
+ if (defined($attr->{title})) {
+ $inline_ok = 0;
+ } elsif (($attr->{media}||'screen') eq 'screen') {
+ $attr->{-inline} = $local;
+ }
+ }
+ push @$links, $attr;
+ }
+
+ my $buf = "<style>$STYLE";
+ if ($inline_ok) {
+ my @ext; # for media=print and whatnot
+ foreach my $attr (@$links) {
+ if (defined(my $str = delete $attr->{-inline})) {
+ $buf .= $str;
+ } else {
+ push @ext, $attr;
+ }
+ }
+ $links = \@ext;
+ }
+ $buf .= '</style>';
+
+ if (@$links) {
+ foreach my $attr (@$links) {
+ delete $attr->{-inline};
+ $buf .= "<link\ntype=text/css\nrel=stylesheet";
+ while (my ($k, $v) = each %$attr) {
+ $v = qq{"$v"} if $v =~ /[\s=]/;
+ $buf .= qq{\n$k=$v};
+ }
+ $buf .= ' />';
+ }
+ $self->{"-style-$upfx"} = $buf;
+ } else {
+ $self->{-style_inline} = $buf;
+ }
+ $self->{-css_map} = $css_map;
+}
+
+# returns an HTML fragment with <style> or <link> tags in them
+# Called by WwwStream by nearly every HTML page
+sub style {
+ my ($self, $upfx) = @_;
+ $self->{-style_inline} || $self->{"-style-$upfx"} || do {
+ stylesheets_prepare($self, $upfx);
+ $self->{-style_inline} || $self->{"-style-$upfx"}
+ };
+}
+
+# /$INBOX/$KEY.css endpoint
+# CSS is configured globally for all inboxes, but we access them on
+# a per-inbox basis. This allows administrators to setup per-inbox
+# static routes to intercept the request before it hits PSGI
+sub get_css ($$$) {
+ my ($ctx, $inbox, $key) = @_;
+ my $r404 = invalid_inbox($ctx, $inbox);
+ return $r404 if $r404;
+ my $self = $ctx->{www};
+ my $css_map = $self->{-css_map} || stylesheets_prepare($self, '');
+ my $css = $css_map->{$key};
+ if (!defined($css) && $key eq 'userContent') {
+ my $env = $ctx->{env};
+ $css = PublicInbox::UserContent::sample($ctx->{-inbox}, $env);
+ }
+ defined $css or return r404();
+ my $h = [ 'Content-Length', bytes::length($css),
+ 'Content-Type', 'text/css' ];
+ PublicInbox::GitHTTPBackend::cache_one_year($h);
+ [ 200, $h, [ $css ] ];
+}
+
+sub get_description {
+ my ($ctx, $inbox) = @_;
+ invalid_inbox($ctx, $inbox) || do {
+ my $d = $ctx->{-inbox}->description . "\n";
+ [ 200, [ 'Content-Length', bytes::length($d),
+ 'Content-Type', 'text/plain' ], [ $d ] ];
+ };
+}
+
1;