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->{-no_obfuscate} = {};
37 $self->{-limiters} = {};
38 $self->{-code_repos} = {}; # nick => PublicInbox::Git object
39 $self->{-cgitrc_unparsed} = $self->{'publicinbox.cgitrc'};
41 if (my $no = delete $self->{'publicinbox.noobfuscate'}) {
44 foreach my $n (@$no) {
45 my @n = split(/\s+/, $n);
47 if (/\S+@\S+/) { # full address
48 $self->{-no_obfuscate}->{lc $_} = 1;
50 # allow "example.com" or "@example.com"
52 push @domains, quotemeta($_);
56 my $nod = join('|', @domains);
57 $self->{-no_obfuscate_re} = qr/(?:$nod)\z/i;
59 if (my $css = delete $self->{'publicinbox.css'}) {
60 $self->{css} = _array($css);
67 sub fill_all ($) { each_inbox($_[0], \&noop) }
69 sub _lookup_fill ($$$) {
70 my ($self, $cache, $key) = @_;
71 $self->{$cache}->{$key} // do {
73 $self->{$cache}->{$key};
78 my ($self, $recipient) = @_;
79 _lookup_fill($self, '-by_addr', lc($recipient));
83 my ($self, $list_id) = @_;
84 _lookup_fill($self, '-by_list_id', lc($list_id));
87 sub lookup_name ($$) {
88 my ($self, $name) = @_;
89 $self->{-by_name}->{$name} // _fill($self, "publicinbox.$name");
93 my ($self, $name) = @_;
94 $self->{-ei_by_name}->{$name} //= _fill_ei($self, "extindex.$name");
97 # special case for [extindex "all"]
98 sub ALL { lookup_ei($_[0], 'all') }
101 my ($self, $cb, @arg) = @_;
102 # may auto-vivify if config file is non-existent:
103 foreach my $section (@{$self->{-section_order}}) {
104 next if $section !~ m!\Apublicinbox\.([^/]+)\z!;
105 my $ibx = lookup_name($self, $1) or next;
110 sub lookup_newsgroup {
111 my ($self, $ng) = @_;
112 _lookup_fill($self, '-by_newsgroup', lc($ng));
116 my ($self, $name) = @_;
117 $self->{-limiters}->{$name} //= do {
118 require PublicInbox::Qspawn;
119 my $max = $self->{"publicinboxlimiter.$name.max"} || 1;
120 my $limiter = PublicInbox::Qspawn::Limiter->new($max);
121 $limiter->setup_rlimit($name, $self);
126 sub config_dir { $ENV{PI_DIR} // "$ENV{HOME}/.public-inbox" }
129 $ENV{PI_CONFIG} // (config_dir() . '/config');
132 sub config_fh_parse ($$$) {
133 my ($fh, $rs, $fs) = @_;
135 my (%section_seen, @section_order);
137 while (defined(my $line = <$fh>)) {
139 my ($k, $v) = split($fs, $line, 2);
140 my ($section) = ($k =~ /\A(\S+)\.[^\.]+\z/);
141 unless (defined $section_seen{$section}) {
142 $section_seen{$section} = 1;
143 push @section_order, $section;
148 if (ref($cur) eq "ARRAY") {
151 $rv{$k} = [ $cur, $v ];
157 $rv{-section_order} = \@section_order;
162 sub git_config_dump {
164 return {} unless -e $file;
165 my @cmd = (qw/git config -z -l --includes/, "--file=$file");
166 my $cmd = join(' ', @cmd);
167 my $fh = popen_rd(\@cmd);
168 my $rv = config_fh_parse($fh, "\0", "\n");
169 close $fh or die "failed to close ($cmd) pipe: $?";
173 sub valid_inbox_name ($) {
176 # Similar rules found in git.git/remote.c::valid_remote_nick
177 # and git.git/refs.c::check_refname_component
178 # We don't reject /\.lock\z/, however, since we don't lock refs
179 if ($name eq '' || $name =~ /\@\{/ ||
180 $name =~ /\.\./ || $name =~ m![/:\?\[\]\^~\s\f[:cntrl:]\*]! ||
181 $name =~ /\A\./ || $name =~ /\.\z/) {
185 # Note: we allow URL-unfriendly characters; users may configure
186 # non-HTTP-accessible inboxes
190 # XXX needs testing for cgit compatibility
191 # cf. cgit/scan-tree.c::add_repo
192 sub cgit_repo_merge ($$$) {
193 my ($self, $base, $repo) = @_;
194 my $path = $repo->{dir};
195 if (defined(my $se = $self->{-cgit_strict_export})) {
196 return unless -e "$path/$se";
198 return if -e "$path/noweb";
199 # this comes from the cgit config, and AFAIK cgit only allows
200 # repos to have one URL, but that's just the PATH_INFO component,
201 # not the Host: portion
202 # $repo = { url => 'foo.git', dir => '/path/to/foo.git' }
203 my $rel = $repo->{url};
204 unless (defined $rel) {
205 my $off = index($path, $base, 0);
209 $rel = substr($path, length($base) + 1);
212 $rel =~ s!/\.git\z!! or
215 $self->{-cgit_remove_suffix} and
216 $rel =~ s!/?\.git\z!!;
218 $self->{"coderepo.$rel.dir"} //= $path;
219 $self->{"coderepo.$rel.cgiturl"} //= _array($rel);
224 -d "$git_dir/objects" && -f "$git_dir/HEAD";
227 # XXX needs testing for cgit compatibility
228 sub scan_path_coderepo {
229 my ($self, $base, $path) = @_;
230 opendir(my $dh, $path) or do {
231 warn "error opening directory: $path\n";
235 if (is_git_dir($git_dir) || is_git_dir($git_dir .= '/.git')) {
236 my $repo = { dir => $git_dir };
237 cgit_repo_merge($self, $base, $repo);
240 while (defined(my $dn = readdir $dh)) {
241 next if $dn eq '.' || $dn eq '..';
242 if (index($dn, '.') == 0 && !$self->{-cgit_scan_hidden_path}) {
245 my $dir = "$path/$dn";
246 scan_path_coderepo($self, $base, $dir) if -d $dir;
250 sub scan_tree_coderepo ($$) {
251 my ($self, $path) = @_;
252 scan_path_coderepo($self, $path, $path);
255 sub scan_projects_coderepo ($$$) {
256 my ($self, $list, $path) = @_;
257 open my $fh, '<', $list or do {
258 warn "failed to open cgit projectlist=$list: $!\n";
263 scan_path_coderepo($self, $path, "$path/$_");
268 my ($self, $cgitrc, $nesting) = @_;
271 my %s = map { $_ => 1 } qw(/cgit.css /cgit.png
272 /favicon.ico /robots.txt);
273 $self->{-cgit_static} = \%s;
276 # same limit as cgit/configfile.c::parse_configfile
277 return if $nesting > 8;
279 open my $fh, '<', $cgitrc or do {
280 warn "failed to open cgitrc=$cgitrc: $!\n";
284 # FIXME: this doesn't support macro expansion via $VARS, yet
288 if (m!\Arepo\.url=(.+?)/*\z!) {
290 cgit_repo_merge($self, $repo->{dir}, $repo) if $repo;
291 $repo = { url => $nick };
292 } elsif (m!\Arepo\.path=(.+)\z!) {
296 warn "$_ without repo.url\n";
298 } elsif (m!\Ainclude=(.+)\z!) {
299 parse_cgitrc($self, $1, $nesting + 1);
300 } elsif (m!\A(scan-hidden-path|remove-suffix)=([0-9]+)\z!) {
301 my ($k, $v) = ($1, $2);
303 $self->{"-cgit_$k"} = $v;
304 } elsif (m!\A(project-list|strict-export)=(.+)\z!) {
305 my ($k, $v) = ($1, $2);
307 $self->{"-cgit_$k"} = $v;
308 } elsif (m!\Ascan-path=(.+)\z!) {
309 if (defined(my $list = $self->{-cgit_project_list})) {
310 scan_projects_coderepo($self, $list, $1);
312 scan_tree_coderepo($self, $1);
314 } elsif (m!\A(?:css|favicon|logo|repo\.logo)=(/.+)\z!) {
315 # absolute paths for static files via PublicInbox::Cgit
316 $self->{-cgit_static}->{$1} = 1;
319 cgit_repo_merge($self, $repo->{dir}, $repo) if $repo;
323 # Only git is supported at the moment, but SVN and Hg are possibilities
324 sub _fill_code_repo {
325 my ($self, $nick) = @_;
326 my $pfx = "coderepo.$nick";
328 # TODO: support gitweb and other repository viewers?
329 if (defined(my $cgitrc = delete $self->{-cgitrc_unparsed})) {
330 parse_cgitrc($self, $cgitrc, 0);
332 my $dir = $self->{"$pfx.dir"}; # aka "GIT_DIR"
333 unless (defined $dir) {
334 warn "$pfx.dir unset\n";
338 my $git = PublicInbox::Git->new($dir);
339 foreach my $t (qw(blob commit tree tag)) {
340 $git->{$t.'_url_format'} =
341 _array($self->{lc("$pfx.${t}UrlFormat")});
344 if (defined(my $cgits = $self->{"$pfx.cgiturl"})) {
345 $git->{cgit_url} = $cgits = _array($cgits);
346 $self->{"$pfx.cgiturl"} = $cgits;
348 # cgit supports "/blob/?id=%s", but it's only a plain-text
349 # display and requires an unabbreviated id=
350 foreach my $t (qw(blob commit tag)) {
351 $git->{$t.'_url_format'} //= map {
361 my ($val) = $_[-1]; # $_[0] may be $self, or $val
362 if ($val =~ /\A(?:false|no|off|[\-\+]?(?:0x)?0+)\z/i) {
364 } elsif ($val =~ /\A(?:true|yes|on|[\-\+]?(?:0x)?[0-9]+)\z/i) {
372 my ($self, $pfx) = @_;
375 for my $k (qw(watch nntpserver)) {
376 my $v = $self->{"$pfx.$k"};
377 $ibx->{$k} = $v if defined $v;
379 for my $k (qw(filter inboxdir newsgroup replyto httpbackendmax feedmax
380 indexlevel indexsequentialshard)) {
381 if (defined(my $v = $self->{"$pfx.$k"})) {
382 if (ref($v) eq 'ARRAY') {
384 W: $pfx.$k has multiple values, only using `$v->[-1]'
386 $ibx->{$k} = $v->[-1];
393 # backwards compatibility:
394 $ibx->{inboxdir} //= $self->{"$pfx.mainrepo"};
395 if (($ibx->{inboxdir} // '') =~ /\n/s) {
396 warn "E: `$ibx->{inboxdir}' must not contain `\\n'\n";
399 foreach my $k (qw(obfuscate)) {
400 my $v = $self->{"$pfx.$k"};
402 if (defined(my $bval = git_bool($v))) {
405 warn "Ignoring $pfx.$k=$v in config, not boolean\n";
408 # TODO: more arrays, we should support multi-value for
409 # more things to encourage decentralization
410 foreach my $k (qw(address altid nntpmirror coderepo hide listid url
411 infourl watchheader)) {
412 if (defined(my $v = $self->{"$pfx.$k"})) {
413 $ibx->{$k} = _array($v);
417 return unless defined($ibx->{inboxdir});
419 $name =~ s/\Apublicinbox\.//;
421 if (!valid_inbox_name($name)) {
422 warn "invalid inbox name: '$name'\n";
426 $ibx->{name} = $name;
427 $ibx->{-pi_config} = $self;
428 $ibx = PublicInbox::Inbox->new($ibx);
429 foreach (@{$ibx->{address}}) {
430 my $lc_addr = lc($_);
431 $self->{-by_addr}->{$lc_addr} = $ibx;
432 $self->{-no_obfuscate}->{$lc_addr} = 1;
434 if (my $listids = $ibx->{listid}) {
435 # RFC2919 section 6 stipulates "case insensitive equality"
436 foreach my $list_id (@$listids) {
437 $self->{-by_list_id}->{lc($list_id)} = $ibx;
440 if (my $ngname = $ibx->{newsgroup}) {
442 delete $ibx->{newsgroup};
443 warn 'multiple newsgroups not supported: '.
444 join(', ', @$ngname). "\n";
445 # Newsgroup name needs to be compatible with RFC 3977
446 # wildmat-exact and RFC 3501 (IMAP) ATOM-CHAR.
447 # Leave out a few chars likely to cause problems or conflicts:
448 # '|', '<', '>', ';', '#', '$', '&',
449 } elsif ($ngname =~ m![^A-Za-z0-9/_\.\-\~\@\+\=:]!) {
450 delete $ibx->{newsgroup};
451 warn "newsgroup name invalid: `$ngname'\n";
453 # PublicInbox::NNTPD does stricter ->nntp_usable
454 # checks, keep this lean for startup speed
455 $self->{-by_newsgroup}->{$ngname} = $ibx;
458 $self->{-by_name}->{$name} = $ibx;
459 if ($ibx->{obfuscate}) {
460 $ibx->{-no_obfuscate} = $self->{-no_obfuscate};
461 $ibx->{-no_obfuscate_re} = $self->{-no_obfuscate_re};
462 fill_all($self); # noop to populate -no_obfuscate
465 if (my $ibx_code_repos = $ibx->{coderepo}) {
466 my $code_repos = $self->{-code_repos};
467 my $repo_objs = $ibx->{-repo_objs} = [];
468 foreach my $nick (@$ibx_code_repos) {
469 my @parts = split(m!/!, $nick);
471 $valid += valid_inbox_name($_) foreach (@parts);
472 $valid == scalar(@parts) or next;
474 my $repo = $code_repos->{$nick} //=
475 _fill_code_repo($self, $nick);
476 push @$repo_objs, $repo if $repo;
484 my ($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;