Eric Wong [Fri, 9 Sep 2016 09:05:18 +0000 (09:05 +0000)]
TODO: updates for done items
The existing string -> number date range Xapian query is good
enough, and having too much flexibility is probably bad for
caching (as well as increasing our attack surface, because
parsing queries is tricky).
Tags-as-skiplists are probably not worth the effort given
Xapian, and we may have to import old messages after-the-fact,
anyways, and message delivery for mirrors is never orderly.
Other items are all done and need to be maintained (like the
search engine docs for the mairix-compatibility features that
just got pushed out)
Eric Wong [Fri, 9 Sep 2016 00:01:29 +0000 (00:01 +0000)]
search: avoid mindlessly calling body_set
It's not worth entering a complex codepath in Email::MIME to
save some (probably immeasurable amount of) memory, here. We've
already stopped doing this in our WWW code a while back, too.
If we really cared enough about it, we'd prioritize work on a
streaming replacement for Email::MIME.
Eric Wong [Fri, 9 Sep 2016 00:01:28 +0000 (00:01 +0000)]
search: fix compatibility with Debian wheezy
Specifying the "d:" field only worked for
NumberValueRangeProcessor in older versions of Xapian, such
as the one in Debian wheezy (libsearch-xapian-perl=1.2.10.0-1)
This slipped through since I rarely use wheezy, anymore, and
perhaps nobody else does, either. Perhaps wheezy support may be
dropped, soon.
Unfortunately, this requires a schema version bump.
Eric Wong [Fri, 9 Sep 2016 00:01:27 +0000 (00:01 +0000)]
search: increase term positions for each quoted hunk
We pay a storage cost for storing positional information
in Xapian, make good use of it by attempting to preserve
it for (hopefully) better search results.
Eric Wong [Fri, 9 Sep 2016 00:01:25 +0000 (00:01 +0000)]
search: fix space regressions from recent changes
As of Xapian 1.0.4 (from 2007) is possible to use
Search::Xapian::QueryParser::add_prefix multiple times with the
same user field name but different term prefixes.
This brings my current git@vger mirror from 6.5GB to 2.1GB
(both sizes are after xapian-compact).
Eric Wong [Fri, 9 Sep 2016 00:01:24 +0000 (00:01 +0000)]
search: more granular message body searching
"bs:" and "b:" are adapted from mairix(1)
We will also support searching explicitly for quoted vs
non-quoted text via "q:" and "nq:" prefixes since sometimes
readers will not care for quoted text.
In the future, we will support parsing diffs (perhaps when
repobrowse integration is complete).
Note: this roughly doubles the size of the Xapian database due
to the additional information; so this change may not be worth
it.
Eric Wong [Fri, 9 Sep 2016 00:01:23 +0000 (00:01 +0000)]
search: drop longer subject: prefix for search
We only document the "s:" anyways. While the long name is more
descriptive, the ambiguity makes agnostic caching (by Varnish or
similar) slightly harder and longer URLs are more likely to be
accidentally truncated when shared.
Eric Wong [Fri, 9 Sep 2016 00:01:22 +0000 (00:01 +0000)]
search: allow searching user fields (To/Cc/From)
Sometimes it can be useful to search based on who the
message was sent to, sent by, or Cc:-ed. Of course,
headers can be faked, but they usually are not...
Anyways this mostly matches the behavior of mairix(1).
Eric Wong [Thu, 8 Sep 2016 20:15:25 +0000 (20:15 +0000)]
doc: document PERL_INLINE_DIRECTORY usage
For now, we will document this since it allows better
performance without the burden of extensions. Perhaps one day
far in the future Perl can natively support vfork(2) AND that
version of Perl will be widely available, but I suspect that day
is at least a decade away, if not two:
Eric Wong [Thu, 8 Sep 2016 19:44:16 +0000 (19:44 +0000)]
view: handle missing Content-Type in message
Email::MIME internally assumes "text/plain" for messages
missing a Content-Type, but does not expose that in the
Email::MIME::content_type API method. We must assume it
ourselves to avoid uninitialized value warnings for the
rare (nowadays) MUAs which do not set it.
Eric Wong [Tue, 23 Aug 2016 21:23:53 +0000 (21:23 +0000)]
www: give tor2web some exposure, too
Not everybody can run Tor, hopefully more can use Tor2web
even if it compromises their privacy. This should help
make system more resilient for users unable to use Tor.
Eric Wong [Thu, 18 Aug 2016 04:44:07 +0000 (04:44 +0000)]
www: implement generic help text
Begin documenting some basic help functionality.
I may tweak the anchor names of the various HTML endpoints
to be more consistent with each other (old ones will be
supported for a short while), so I'm not documenting
those, for now.
This may become part of a builtin key-value store for
basic texts, but this probably shouldn't become a wiki
engine, either.
Eric Wong [Thu, 18 Aug 2016 02:02:50 +0000 (02:02 +0000)]
linkify: be stricter about matching RFC 3986
We're not to-the-letter about percent-encoding, but
we should allow all the characters. This is mainly
so we can effectively use the link to some Wikipedia
pages with parentheses in them:
Eric Wong [Thu, 18 Aug 2016 01:10:35 +0000 (01:10 +0000)]
view: try assuming UTF-8 for bogus charsets
For some reason, Alpine will set X-UNKNOWN for valid UTF-8.
Since we favor UTF-8 HTML anyways, try forcing Email::MIME to
handle text/plain as UTF-8 which might show up better.
Eric Wong [Thu, 18 Aug 2016 00:54:25 +0000 (00:54 +0000)]
view: try to display bogus charsets for text/plain
Alpine seems to set charset=X-UNKNOWN for valid UTF-8 text,
which causes Email::MIME::body_str to fail as X-UNKNOWN
is not a valid encoding. So, blindly display the body
as plain-text but warn users about possibly mangled text.
Reported-by: Thomas Ferris Nicolaisen <tfnico@gmail.com>
Eric Wong [Tue, 16 Aug 2016 08:49:26 +0000 (08:49 +0000)]
search: add YYYYMMDD search range via "d:" prefix
This is similar to mairix in that it uses a "d:" prefix; but
only takes YYYYMMDD, for now. Using custom date/time parsers
via Perl will be much more work:
Eric Wong [Tue, 16 Aug 2016 08:49:25 +0000 (08:49 +0000)]
search: drop pointless range processors for Unix timestamp
The Unix timestamp isn't meaningful for users searching,
we will start indexing the YYYYMMDD date stamp which may
use StringValueRangeProcessor, instead.
Eric Wong [Sun, 14 Aug 2016 10:21:10 +0000 (10:21 +0000)]
www: do not double-clean Message-IDs from internal DBs
Ensure we usually strip one level of '<>' from Message-IDs,
since our internal SQLite, Xapian, and SHA-1 storage all
assume that.
Realistically, we screw up if somebody has '<<' or '>>',
but those are screwed up mail clients and we can deal with
it another time. Currently, this means some messages with
'>>' in References or Message-Id are not handled correctly,
yet, but we match the behavior of Mail::Thread in keeping
the extra '>'.
Eric Wong [Sun, 14 Aug 2016 10:21:09 +0000 (10:21 +0000)]
www: do not unecessarily escape some chars in paths
Based on reading RFC 3986, it seems '@', ':', '!', '$', '&',
"'", '; '(', ')', '*', '+', ',', ';', '=' are all allowed
in path-absolute where we have the Message-ID.
In any case, it seems '@' is fairly common in path components
nowadays and too common in Message-IDs.
Eric Wong [Sun, 14 Aug 2016 10:21:17 +0000 (10:21 +0000)]
www: ensure XML validity for some odd ASCII chars
I've seen 0x1b (\e) in at least one message and some other
possibly non-printable chars. In any case, make sure they're
valid XML with us-ascii encoding as far as xmlstarlet(1) thinks
so.
Eric Wong [Sat, 13 Aug 2016 00:22:01 +0000 (00:22 +0000)]
extmsg: reorder and add a more Message-ID lookup services
gmane is down at the moment, so lower that in priority
(hopefully it will be brought back up, again). Wikipedia also
lists a few more project-specific list providers, so include
those as well: https://en.wikipedia.org/wiki/Message-ID
Eric Wong [Fri, 12 Aug 2016 19:52:35 +0000 (19:52 +0000)]
www: allow including links to NNTP sites in HTML footer
Improve the discoverability of NNTP endpoints for users
who still know what NNTP is.
==> ~/.public-inbox/config <==
; aliases for the locally-run nntpd can be specified in
; the "publicinbox" section:
[publicinbox]
nntpserver = nntp://ou63pmih66umazou.onion/
nntpserver = news.public-inbox.org
; NNTPS is not supported natively, yet,
; but one can use haproxy or similar
; nntpserver = nntps://news.public-inbox.invalid/
; mirrors for specific inboxes may be specified either as full
; NNTP (or NNTPS) URLs, or with the server name only if the
; newsgroup name is specfied for a local NNTP server
[publicinbox "git"]
...
newsgroup = inbox.a.b.c
nntpmirror = nntp://czquwvybam4bgbro.onion/
nntpmirror = hjrcffqmbrq6wope.onion
; there may be a mirror on a different server with a
; different name:
nntpmirror = nntp://news.example.com/differently.named.group
; (And I really need to write manpages for all this...)
Eric Wong [Thu, 11 Aug 2016 00:23:48 +0000 (00:23 +0000)]
search: support alt-ID for mapping legacy serial numbers
For some existing mailing list archives, messages are identified
by serial number (such as NNTP article numbers in gmane). Those
links may become inaccessible (as is the current case for
gmane), so ensure users can still search based on old serial
numbers.
Now, I run the following periodically to get article numbers
from gmane (while news.gmane.org remains):
; relative pathnames expand to $mainrepo/public-inbox/$file
altid = serial:gmane:file=gmane.sqlite3
And run "public-inbox-index --reindex /path/to/git.vger.git"
periodically.
This ought to allow searching for "gmane:12345" to work for
Xapian-enabled instances.
Disclaimer: while public-inbox supports NNTP and stable article
serial numbers, use of those for public links is discouraged
since it encourages centralization.
Eric Wong [Tue, 9 Aug 2016 23:59:10 +0000 (23:59 +0000)]
searchidx: allow searching Message-IDs in free-form text
It is not unheard of for users to attempt finding messages by
entering Message-IDs into the "Search" box instead of using the
existing URL structure. So make it possible for them.
Fwiw, I've definitely encountered users who enter entire URLs
into generic search engines.
Eric Wong [Tue, 9 Aug 2016 00:41:37 +0000 (00:41 +0000)]
searchidx: avoid holding Xapian lock in cat-file
We must ensure cat-file process is launched before Xapian
grabs lock, too. Our use of "git cat-file --batch" has
the same problem as "git log" did, (which was fixed in
commit 3713c727cda431a0dc2865a7878c13ecf9f21851)
"searchidx: release Xapian FDs before spawning git log"
Eric Wong [Sat, 6 Aug 2016 01:58:47 +0000 (01:58 +0000)]
mbox: be fair to other HTTP clients
At least for public-inbox-httpd, this allows us to avoid having
a client monopolize one event loop tick of the server for too
long. It hurts throughput for the /all.mbox.gz endpoint, but I
doubt anybody cares and the latency improvement for other
clients would be appreciated.
We already do the same fairness thing for HTML pages.
Eric Wong [Fri, 5 Aug 2016 22:07:25 +0000 (22:07 +0000)]
search: disable batching in newer versions of Xapian, for now
This warrants further investigation, but it appears we cannot
release Xapian reliably after forking "git log" due to the
lack of a close-on-exec flag on the Xapian flintlock FD
Eric Wong [Fri, 5 Aug 2016 18:10:50 +0000 (18:10 +0000)]
view: use <hr> to delineate in /$MID/T/ view
The sacrifice in vertical space might be worth it to improve
ease-of-reading, as it's unreasonable to expect an entire
message thread to be able to fit into a single window.
Eric Wong [Thu, 4 Aug 2016 21:58:40 +0000 (21:58 +0000)]
view: do not fail on empty In-Reply-To
Sometimes messages have an empty In-Reply-To header which throws
threaders off. This actually causes public-inbox-httpd to die,
which is probably bad and will be fixed elsewhere.
Eric Wong [Tue, 2 Aug 2016 10:02:54 +0000 (10:02 +0000)]
searchmsg: add git object ID to doc_data
Doing git tree lookups based on the SHA-1 of the Message-ID
is expensive as trees get larger, instead, use the SHA-1
object ID directly. This drastically reduces the amount
of time spent in the "git cat-file --batch" process for
fetching the /$INBOX/all.mbox.gz endpoint on the ~800MB
git@vger.kernel.org mirror
This retains backwards compatibility and allows existing
indices to be transparently upgraded without performance
degradation.
Eric Wong [Tue, 2 Aug 2016 01:47:55 +0000 (01:47 +0000)]
wwwstream: prioritize search in top title bar
search is probably more useful so users should be able to select
it sooner. Put it on its own line so it won't get scrolled off
the edge for non-CSS users.
Fix a minor spacing bug in the input tag while we're at it, too
Eric Wong [Tue, 2 Aug 2016 01:47:54 +0000 (01:47 +0000)]
daemon: do not chdir unless daemonizing
As far as most process managers are concerned (e.g. systemd),
they should already start in '/'. So avoid making our daemon
more complex to run by requiring absolute paths during
development.
Eric Wong [Sun, 31 Jul 2016 00:02:06 +0000 (00:02 +0000)]
search: support reindexing existing search indices
This should make tweaking the way we search more efficiet
by allowing us to avoid doubling destroying the index every
time we want to change something.
We also give priority to incremental indexing via
public-inbox-{watch,mda} and have manual invocations of
public-inbox-index perform batch updates while releasing
ssoma.lock.
Eric Wong [Sun, 31 Jul 2016 00:02:05 +0000 (00:02 +0000)]
msgmap: fix use of transactions
We want transactions to be the responsibility of the
caller when possible; this fixes the potential for
the msgmap to internally become inconsistent when
using it from inside searchidx.
Eric Wong [Sat, 30 Jul 2016 23:33:11 +0000 (23:33 +0000)]
t/config_limiter: fix check for identical Git object
If we completely undef an object, it is likely possible
to have the same scalar address as the original object
even if they are different. So keep the same object
around and only force creation of the same reference.
Tested on Perl 5.14.2 on Debian 7.x wheezy.
Eric Wong [Fri, 29 Jul 2016 18:58:51 +0000 (18:58 +0000)]
daemon: re-enable SIGWINCH without setsid
This allows systemd users to use SIGWINCH to temporarily
(and gracefully) stop an instance of a service without
doing a code reload to bring it back up:
# start temporary new service code
systemctl start public-inbox-nntpd@2.service
# momentarily paralyze original service
systemctl kill -s WINCH public-inbox-nntpd@1.service
if new_code_at_2_sucks
then
# restart original workers
systemctl kill -s HUP public-inbox-nntpd@1.service
else # new is better than old, replace original instance
systemctl restart public-inbox-nntpd@1.service
fi
# cleanup the temporary service
systemctl stop public-inbox-nntpd@2.service
Eric Wong [Thu, 21 Jul 2016 01:23:03 +0000 (01:23 +0000)]
view: split up --cc args for git-send-email
Having long Cc: lines is inevitable for large threads
with many participants, and git-send-email only gained
the ability to recognize ',' in the "--cc" arg recently
with the release of git v2.6.0 in September 2015.
Eric Wong [Thu, 21 Jul 2016 01:23:02 +0000 (01:23 +0000)]
www: label sections and hopefully improve navigation
Clearly label "Thread overview" and "Reply instructions"
so users can quickly skip stuff they're not interested in.
Additionally, note the fact the thread view allows quick
navigation within the thread to avoid extra network requests
and improve the display for single-message threads.
Finally, use <hr> to better-delineate sections of each page.
Eric Wong [Sun, 17 Jul 2016 23:27:02 +0000 (23:27 +0000)]
extmsg: favor user-provided URL on partial matches
While an inbox may have multiple URLs, we will favor
the existing URL for the current inbox on partial matches
to avoid confusing users or slowing them down by requiring
a new TCP connection.
Eric Wong [Sat, 9 Jul 2016 07:53:17 +0000 (07:53 +0000)]
view: improve grouping for topic view
This reduces the amount of mbox/Atom links while keeping
better track of overall thread count. We no longer loop
to fill up slots to simplify the code a bit and hopefully
get better grouping.
Eric Wong [Sat, 9 Jul 2016 04:51:37 +0000 (04:51 +0000)]
httpd/async: reinstate D::S timer usage for cleanup
EvCleanup::asap events are not guaranteed to run after
Danga::Socket closes sockets at the event loop. Thus we
must use slower Danga::Socket timers which are guaranteed
to run at the end of the event loop.
Eric Wong [Sat, 9 Jul 2016 04:51:36 +0000 (04:51 +0000)]
httpd/async: do not attempt future writes on closed sockets
Danga::Socket::close does not clear the write_buf_size field,
so it's conceivable we could attempt to queue up data and
callbacks we can never flush out.
Eric Wong [Sat, 9 Jul 2016 03:18:35 +0000 (03:18 +0000)]
www: add configurable limiters
Currently only for git-http-backend use, this allows limiting
the number of spawned processes per-inbox or by group, if there
are multiple large inboxes amidst a sea of small ones.
For example, a "big" repo limiter could be used for big inboxes:
which would be shared between multiple repos:
[limiter "big"]
max = 4
[publicinbox "git"]
address = git@vger.kernel.org
mainrepo = /path/to/git.git
; shared limiter with giant:
httpbackendmax = big
[publicinbox "giant"]
address = giant@project.org
mainrepo = /path/to/giant.git
; shared limiter with git:
httpbackendmax = big
; This is a tiny inbox, use the default limiter with 32 slots:
[publicinbox "meta"]
address = meta@public-inbox.org
mainrepo = /path/to/meta.git