1 # Copyright (C) 2014-2020 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # Used throughout the project for reading configuration
6 # Note: I hate camelCase; but git-config(1) uses it, but it's better
7 # than alllowercasewithoutunderscores, so use lc('configKey') where
8 # applicable for readability
10 package PublicInbox::Config;
13 use PublicInbox::Inbox;
14 use PublicInbox::Spawn qw(popen_rd);
16 sub _array ($) { ref($_[0]) eq 'ARRAY' ? $_[0] : [ $_[0] ] }
18 # returns key-value pairs of config directives in a hash
19 # if keys may be multi-value, the value is an array ref containing all values
21 my ($class, $file) = @_;
22 $file //= default_file();
24 if (ref($file) eq 'SCALAR') { # used by some tests
25 open my $fh, '<', $file or die; # PerlIO::scalar
26 $self = config_fh_parse($fh, "\n", '=');
28 $self = git_config_dump($file);
32 $self->{-by_addr} = {};
33 $self->{-by_list_id} = {};
34 $self->{-by_name} = {};
35 $self->{-by_newsgroup} = {};
36 $self->{-by_eidx_key} = {};
37 $self->{-no_obfuscate} = {};
38 $self->{-limiters} = {};
39 $self->{-code_repos} = {}; # nick => PublicInbox::Git object
40 $self->{-cgitrc_unparsed} = $self->{'publicinbox.cgitrc'};
42 if (my $no = delete $self->{'publicinbox.noobfuscate'}) {
45 foreach my $n (@$no) {
46 my @n = split(/\s+/, $n);
48 if (/\S+@\S+/) { # full address
49 $self->{-no_obfuscate}->{lc $_} = 1;
51 # allow "example.com" or "@example.com"
53 push @domains, quotemeta($_);
57 my $nod = join('|', @domains);
58 $self->{-no_obfuscate_re} = qr/(?:$nod)\z/i;
60 if (my $css = delete $self->{'publicinbox.css'}) {
61 $self->{css} = _array($css);
68 sub fill_all ($) { each_inbox($_[0], \&noop) }
70 sub _lookup_fill ($$$) {
71 my ($self, $cache, $key) = @_;
72 $self->{$cache}->{$key} // do {
74 $self->{$cache}->{$key};
79 my ($self, $recipient) = @_;
80 _lookup_fill($self, '-by_addr', lc($recipient));
84 my ($self, $list_id) = @_;
85 _lookup_fill($self, '-by_list_id', lc($list_id));
88 sub lookup_name ($$) {
89 my ($self, $name) = @_;
90 $self->{-by_name}->{$name} // _fill($self, "publicinbox.$name");
94 my ($self, $name) = @_;
95 $self->{-ei_by_name}->{$name} //= _fill_ei($self, "extindex.$name");
98 # special case for [extindex "all"]
99 sub ALL { lookup_ei($_[0], 'all') }
102 my ($self, $cb, @arg) = @_;
103 # may auto-vivify if config file is non-existent:
104 foreach my $section (@{$self->{-section_order}}) {
105 next if $section !~ m!\Apublicinbox\.([^/]+)\z!;
106 my $ibx = lookup_name($self, $1) or next;
111 sub lookup_newsgroup {
112 my ($self, $ng) = @_;
113 _lookup_fill($self, '-by_newsgroup', lc($ng));
117 my ($self, $name) = @_;
118 $self->{-limiters}->{$name} //= do {
119 require PublicInbox::Qspawn;
120 my $max = $self->{"publicinboxlimiter.$name.max"} || 1;
121 my $limiter = PublicInbox::Qspawn::Limiter->new($max);
122 $limiter->setup_rlimit($name, $self);
127 sub config_dir { $ENV{PI_DIR} // "$ENV{HOME}/.public-inbox" }
130 $ENV{PI_CONFIG} // (config_dir() . '/config');
133 sub config_fh_parse ($$$) {
134 my ($fh, $rs, $fs) = @_;
136 my (%section_seen, @section_order);
138 while (defined(my $line = <$fh>)) {
140 my ($k, $v) = split($fs, $line, 2);
141 my ($section) = ($k =~ /\A(\S+)\.[^\.]+\z/);
142 unless (defined $section_seen{$section}) {
143 $section_seen{$section} = 1;
144 push @section_order, $section;
149 if (ref($cur) eq "ARRAY") {
152 $rv{$k} = [ $cur, $v ];
158 $rv{-section_order} = \@section_order;
163 sub git_config_dump {
165 return {} unless -e $file;
166 my @cmd = (qw/git config -z -l --includes/, "--file=$file");
167 my $cmd = join(' ', @cmd);
168 my $fh = popen_rd(\@cmd);
169 my $rv = config_fh_parse($fh, "\0", "\n");
170 close $fh or die "failed to close ($cmd) pipe: $?";
174 sub valid_inbox_name ($) {
177 # Similar rules found in git.git/remote.c::valid_remote_nick
178 # and git.git/refs.c::check_refname_component
179 # We don't reject /\.lock\z/, however, since we don't lock refs
180 if ($name eq '' || $name =~ /\@\{/ ||
181 $name =~ /\.\./ || $name =~ m![/:\?\[\]\^~\s\f[:cntrl:]\*]! ||
182 $name =~ /\A\./ || $name =~ /\.\z/) {
186 # Note: we allow URL-unfriendly characters; users may configure
187 # non-HTTP-accessible inboxes
191 # XXX needs testing for cgit compatibility
192 # cf. cgit/scan-tree.c::add_repo
193 sub cgit_repo_merge ($$$) {
194 my ($self, $base, $repo) = @_;
195 my $path = $repo->{dir};
196 if (defined(my $se = $self->{-cgit_strict_export})) {
197 return unless -e "$path/$se";
199 return if -e "$path/noweb";
200 # this comes from the cgit config, and AFAIK cgit only allows
201 # repos to have one URL, but that's just the PATH_INFO component,
202 # not the Host: portion
203 # $repo = { url => 'foo.git', dir => '/path/to/foo.git' }
204 my $rel = $repo->{url};
205 unless (defined $rel) {
206 my $off = index($path, $base, 0);
210 $rel = substr($path, length($base) + 1);
213 $rel =~ s!/\.git\z!! or
216 $self->{-cgit_remove_suffix} and
217 $rel =~ s!/?\.git\z!!;
219 $self->{"coderepo.$rel.dir"} //= $path;
220 $self->{"coderepo.$rel.cgiturl"} //= _array($rel);
225 -d "$git_dir/objects" && -f "$git_dir/HEAD";
228 # XXX needs testing for cgit compatibility
229 sub scan_path_coderepo {
230 my ($self, $base, $path) = @_;
231 opendir(my $dh, $path) or do {
232 warn "error opening directory: $path\n";
236 if (is_git_dir($git_dir) || is_git_dir($git_dir .= '/.git')) {
237 my $repo = { dir => $git_dir };
238 cgit_repo_merge($self, $base, $repo);
241 while (defined(my $dn = readdir $dh)) {
242 next if $dn eq '.' || $dn eq '..';
243 if (index($dn, '.') == 0 && !$self->{-cgit_scan_hidden_path}) {
246 my $dir = "$path/$dn";
247 scan_path_coderepo($self, $base, $dir) if -d $dir;
251 sub scan_tree_coderepo ($$) {
252 my ($self, $path) = @_;
253 scan_path_coderepo($self, $path, $path);
256 sub scan_projects_coderepo ($$$) {
257 my ($self, $list, $path) = @_;
258 open my $fh, '<', $list or do {
259 warn "failed to open cgit projectlist=$list: $!\n";
264 scan_path_coderepo($self, $path, "$path/$_");
269 my ($self, $cgitrc, $nesting) = @_;
272 my %s = map { $_ => 1 } qw(/cgit.css /cgit.png
273 /favicon.ico /robots.txt);
274 $self->{-cgit_static} = \%s;
277 # same limit as cgit/configfile.c::parse_configfile
278 return if $nesting > 8;
280 open my $fh, '<', $cgitrc or do {
281 warn "failed to open cgitrc=$cgitrc: $!\n";
285 # FIXME: this doesn't support macro expansion via $VARS, yet
289 if (m!\Arepo\.url=(.+?)/*\z!) {
291 cgit_repo_merge($self, $repo->{dir}, $repo) if $repo;
292 $repo = { url => $nick };
293 } elsif (m!\Arepo\.path=(.+)\z!) {
297 warn "$_ without repo.url\n";
299 } elsif (m!\Ainclude=(.+)\z!) {
300 parse_cgitrc($self, $1, $nesting + 1);
301 } elsif (m!\A(scan-hidden-path|remove-suffix)=([0-9]+)\z!) {
302 my ($k, $v) = ($1, $2);
304 $self->{"-cgit_$k"} = $v;
305 } elsif (m!\A(project-list|strict-export)=(.+)\z!) {
306 my ($k, $v) = ($1, $2);
308 $self->{"-cgit_$k"} = $v;
309 } elsif (m!\Ascan-path=(.+)\z!) {
310 if (defined(my $list = $self->{-cgit_project_list})) {
311 scan_projects_coderepo($self, $list, $1);
313 scan_tree_coderepo($self, $1);
315 } elsif (m!\A(?:css|favicon|logo|repo\.logo)=(/.+)\z!) {
316 # absolute paths for static files via PublicInbox::Cgit
317 $self->{-cgit_static}->{$1} = 1;
320 cgit_repo_merge($self, $repo->{dir}, $repo) if $repo;
324 # Only git is supported at the moment, but SVN and Hg are possibilities
325 sub _fill_code_repo {
326 my ($self, $nick) = @_;
327 my $pfx = "coderepo.$nick";
329 # TODO: support gitweb and other repository viewers?
330 if (defined(my $cgitrc = delete $self->{-cgitrc_unparsed})) {
331 parse_cgitrc($self, $cgitrc, 0);
333 my $dir = $self->{"$pfx.dir"}; # aka "GIT_DIR"
334 unless (defined $dir) {
335 warn "$pfx.dir unset\n";
339 my $git = PublicInbox::Git->new($dir);
340 foreach my $t (qw(blob commit tree tag)) {
341 $git->{$t.'_url_format'} =
342 _array($self->{lc("$pfx.${t}UrlFormat")});
345 if (defined(my $cgits = $self->{"$pfx.cgiturl"})) {
346 $git->{cgit_url} = $cgits = _array($cgits);
347 $self->{"$pfx.cgiturl"} = $cgits;
349 # cgit supports "/blob/?id=%s", but it's only a plain-text
350 # display and requires an unabbreviated id=
351 foreach my $t (qw(blob commit tag)) {
352 $git->{$t.'_url_format'} //= map {
362 my ($val) = $_[-1]; # $_[0] may be $self, or $val
363 if ($val =~ /\A(?:false|no|off|[\-\+]?(?:0x)?0+)\z/i) {
365 } elsif ($val =~ /\A(?:true|yes|on|[\-\+]?(?:0x)?[0-9]+)\z/i) {
373 my ($self, $pfx) = @_;
376 for my $k (qw(watch nntpserver)) {
377 my $v = $self->{"$pfx.$k"};
378 $ibx->{$k} = $v if defined $v;
380 for my $k (qw(filter inboxdir newsgroup replyto httpbackendmax feedmax
381 indexlevel indexsequentialshard)) {
382 if (defined(my $v = $self->{"$pfx.$k"})) {
383 if (ref($v) eq 'ARRAY') {
385 W: $pfx.$k has multiple values, only using `$v->[-1]'
387 $ibx->{$k} = $v->[-1];
394 # backwards compatibility:
395 $ibx->{inboxdir} //= $self->{"$pfx.mainrepo"};
396 if (($ibx->{inboxdir} // '') =~ /\n/s) {
397 warn "E: `$ibx->{inboxdir}' must not contain `\\n'\n";
400 foreach my $k (qw(obfuscate)) {
401 my $v = $self->{"$pfx.$k"};
403 if (defined(my $bval = git_bool($v))) {
406 warn "Ignoring $pfx.$k=$v in config, not boolean\n";
409 # TODO: more arrays, we should support multi-value for
410 # more things to encourage decentralization
411 foreach my $k (qw(address altid nntpmirror coderepo hide listid url
412 infourl watchheader)) {
413 if (defined(my $v = $self->{"$pfx.$k"})) {
414 $ibx->{$k} = _array($v);
418 return unless defined($ibx->{inboxdir});
420 $name =~ s/\Apublicinbox\.//;
422 if (!valid_inbox_name($name)) {
423 warn "invalid inbox name: '$name'\n";
427 $ibx->{name} = $name;
428 $ibx->{-pi_config} = $self;
429 $ibx = PublicInbox::Inbox->new($ibx);
430 foreach (@{$ibx->{address}}) {
431 my $lc_addr = lc($_);
432 $self->{-by_addr}->{$lc_addr} = $ibx;
433 $self->{-no_obfuscate}->{$lc_addr} = 1;
435 if (my $listids = $ibx->{listid}) {
436 # RFC2919 section 6 stipulates "case insensitive equality"
437 foreach my $list_id (@$listids) {
438 $self->{-by_list_id}->{lc($list_id)} = $ibx;
441 if (my $ngname = $ibx->{newsgroup}) {
443 delete $ibx->{newsgroup};
444 warn 'multiple newsgroups not supported: '.
445 join(', ', @$ngname). "\n";
446 # Newsgroup name needs to be compatible with RFC 3977
447 # wildmat-exact and RFC 3501 (IMAP) ATOM-CHAR.
448 # Leave out a few chars likely to cause problems or conflicts:
449 # '|', '<', '>', ';', '#', '$', '&',
450 } elsif ($ngname =~ m![^A-Za-z0-9/_\.\-\~\@\+\=:]!) {
451 delete $ibx->{newsgroup};
452 warn "newsgroup name invalid: `$ngname'\n";
454 # PublicInbox::NNTPD does stricter ->nntp_usable
455 # checks, keep this lean for startup speed
456 $self->{-by_newsgroup}->{$ngname} = $ibx;
459 $self->{-by_name}->{$name} = $ibx;
460 if ($ibx->{obfuscate}) {
461 $ibx->{-no_obfuscate} = $self->{-no_obfuscate};
462 $ibx->{-no_obfuscate_re} = $self->{-no_obfuscate_re};
463 fill_all($self); # noop to populate -no_obfuscate
466 if (my $ibx_code_repos = $ibx->{coderepo}) {
467 my $code_repos = $self->{-code_repos};
468 my $repo_objs = $ibx->{-repo_objs} = [];
469 foreach my $nick (@$ibx_code_repos) {
470 my @parts = split(m!/!, $nick);
472 $valid += valid_inbox_name($_) foreach (@parts);
473 $valid == scalar(@parts) or next;
475 my $repo = $code_repos->{$nick} //=
476 _fill_code_repo($self, $nick);
477 push @$repo_objs, $repo if $repo;
480 $self->{-by_eidx_key}->{$ibx->eidx_key} = $ibx;
484 my ($self, $pfx) = @_;
485 require PublicInbox::ExtSearch;
486 my $d = $self->{"$pfx.topdir"};
487 defined($d) && -d $d ? PublicInbox::ExtSearch->new($d) : undef;
491 my ($self, $key, $url) = @_;
492 state $urlmatch_broken; # requires git 1.8.5
493 return if $urlmatch_broken;
494 my $file = default_file();
495 my $cmd = [qw/git config -z --includes --get-urlmatch/,
496 "--file=$file", $key, $url ];
497 my $fh = popen_rd($cmd);
504 $urlmatch_broken = 1 if (($? >> 8) != 1);
512 for my $mod (qw(Cpanel::JSON::XS JSON::MaybeXS JSON JSON::PP)) {
513 eval "require $mod" or next;
514 # ->ascii encodes non-ASCII to "\uXXXX"
515 $json = $mod->new->ascii(1) and last;