Date | Commit message (Collapse) |
|
No backwards compatibility to worry about for us; and fadvise
is superior anyways.
|
|
IO::Poll::_poll returns -1, which is "true" to Perl.
cf. https://rt.cpan.org/Ticket/Display.html?id=129484
|
|
IO::Kqueue seems unmaintained, so workaround a long-standing
bug where it falls over on signals:
https://rt.cpan.org/Ticket/Display.html?id=116615
|
|
There's other ways to profile and we don't need to add runtime
branches to do this.
|
|
More will likely be dropped in the future, but drop the obvious
ones we aren't using, for now; especially since some of them are
set at ->new time and unavoidable.
This saves 579 bytes per-client on my 64-bit Debian stable
system as measured by Devel::Size::total_size from
PublicInbox::HTTP::event_read. This adds up in C10K or C100K
situations.
Things we drop are:
* corked - MSG_MORE requires fewer syscalls
* read_push_back - tried to use it, ate CPU with slow clients
* IP/port fields - accept() already returns what we care about
|
|
Since our listen sockets are non-blocking and we may run
multiple httpd|nntpd processes; we need a way to avoid
thundering herds when there are multiple httpd|nntpd worker
processes.
EPOLLEXCLUSIVE was added just for that in Linux 4.5
|
|
These modules are unmaintained upstream at the moment, but I'll
be able to help with the intended maintainer once/if CPAN
ownership is transferred. OTOH, we've been waiting for that
transfer for several years, now...
Changes I intend to make:
* EPOLLEXCLUSIVE for Linux
* remove unused fields wasting memory
* kqueue bugfixes e.g. https://rt.cpan.org/Ticket/Display.html?id=116615
* accept4 support
And some lower priority experiments:
* switch to EV_ONESHOT / EPOLLONESHOT (incompatible changes)
* nginx-style buffering to tmpfile instead of string array
* sendfile off tmpfile buffers
* io_uring maybe?
|
|
It was a MITM hazard and been killed off by the DoS subsystem
last July: https://marc.info/?i=87d0vwwkbs.fsf@riseup.net
|
|
We already escape the user-provided Message-IDs (so there's no
security problem AFAIK), but the URL templates which exist in
our source code were not escaped properly.
This quiets down tidy(1).
|
|
It seems a common case for mangled patches is editors or MUAs
dropping trailing whitespace, and lines matching /^ $/ gets
the space dropped to only match /^$/.
|
|
Document `publicinbox.cgitdata' config directive, but allow it
to be unspecified and/or missing for installations which do not
wish to serve static data at all.
For users installing cgit from source to their home directory,
we can usually infer the cgit data path based on the cgit.cgi
binary path, even.
|
|
Not all inputs are highlight-able, so reuse the original
input and just linkify it if it can't be highlighted.
|
|
Otherwise, there's no reason to use this API over highlight(1).
Maybe this can be an option in the future; but I'm struggling to
find a reason to not do it by default.
|
|
Try to appear gramatically correct and state:
"only message in thread" when there's only one known (to us)
message in the thread.
|
|
Dangling parentheses with trailing punctuation usually means the
parentheses is not intended as part of the URL.
|
|
Empty subjects ("") and undefined Subjects: are now both
displayed as "(no subject)" for now.
|
|
The URLs at the top of WwwStream.pm weren't getting linkified
correctly.
|
|
For inboxes with SQLite enabled (all v2, and probably most v1);
we can use the overview DB to get the timestamp of the latest
message. It's faster than scanning git branches for commit times,
but not always the same.
|
|
This will be used for generating an HTML listing for v1 inboxes,
at least. The logic for this follows that of grokmirror,
and we may dynamically generate manifest.js.gz natively...
|
|
Followup-to: 6e6f7999361925e4
("cleanup: use '$ibx' consistently when referring to Inbox refs")
|
|
This is for consistency with other fields which follow
this pattern w.r.t. field-naming when referring to internal
fields.
|
|
'$inbox' is more human-readable, so that is for the more
human-readable name in most cases. Making our variable naming
more consistent should make the code easier-to-review and
harder to screw up.
|
|
We access the Git object via the Inbox object nowadays, so
there's no point in having a shortcut to it, anymore.
|
|
We parse cgitrc for "repo.path", while we use "coderepo.dir" to
mean the same thing for non-cgit users. So I ended up confusing
myself, here.
But then again, git uses "--git-dir" and "GIT_DIR", so I suspect
"dir" is the better choice than "path", here
|
|
We'll be building off of this for showing diffs in
the coderepo views.
|
|
Hopefully this gets us closer to matching cgit upstream behavior
(which also lacks tests). We'll still need to support macro
expansion at some point for compatibility...
|
|
We can reduce the configuration needed to run cgit by reusing
the static file handling logic of the dumb git HTTP protocol.
I hate logos and icons, so don't expect public-inbox.org or
80x24.org to ever have those to waste users' bandwidth with :P
But I expect other users to find this useful.
|
|
project_list support still needs to be done
And tests need to be written... :<
|
|
We need to instate our cgit handler everywhere we use NewsWWW
to catch wildcard requests which our normal endpoints do not
handle.
|
|
Our high-level config already treats single limits as a
soft==hard limit for limiters; so stop handling that redundant
in the low-level spawn() sub.
|
|
I mainly need this to enforce RLIMIT_CPU (and RLIMIT_CORE)
when requests come which generate giant, unrealistic diffs.
Per-coderepo limiters may be added in the future. But for now,
I need to prevent cgit from monopolizing resources on my dinky
server.
|
|
This allows users to configure RLIMIT_{CORE,CPU,DATA} using
our "limiter" config directive when spawning external processes.
|
|
Requests intended for cgit are unlikely to conflict with
requests to inboxes. So we can safely hand those requests
off to cgit.cgi.
|
|
We depend on git-http-backend for smart HTTP clone support,
however; since cgit does not support smart clones natively.
WWW.pm will be able to cascade down to this as a 404 handler in
the future.
|
|
cgit (and most other CGI executables) is not typically installed
for use via $PATH, so we'll need to support absolute paths to
run it.
|
|
Reads to git-http-backend(1) could fail or EOF prematurely,
so we must be ready for that case.
Furthermore, cgit (and possibly other CGI) uses LF instead
of CRLF, so support those programs, too.
|
|
This will be useful for other CGI wrappers we make.
This also fixes a bug with some PSGI servers which did not
present a real IO::Handle in the psgi.input env field.
|
|
We can save admins the trouble of declaring [coderepo "..."]
sections in the public-inbox config by parsing the cgitrc
directly.
Macro expansion (e.g. $HTTP_HOST) expansion is not supported,
yet; but may be in the future.
|
|
We'll be spawning cgit and git-diff, which can take gigantic
amounts of CPU time and/or heap given the right (ermm... wrong)
input. Limit the damage that large/expensive diffs can cause.
|
|
No need to scan the entire string, but prefer to match git
behavior. This might be faster if/when Perl can create
substrings efficiently using CoW.
Fix a 80-column violation while we're at it.
|
|
Eventually, we'll have special displays for various git objects
(commit, tree, tag). But for now, we'll just use git-show
to spew whatever comes from git.
|
|
This will be useful for extracting titles/subjects from
commit objects when displaying commits.
|
|
This will be useful for reproducibility when mirroring
coderepos and generating diffs.
|
|
Not entirely sure what is causing this, but it appears to
be causing infinite loops when attempting to display certain
blobs.
Fortunately, the fair scheduling of public-inbox-httpd prevented
this from becoming a real problem aside from increasing CPU
usage.
|
|
There's no reason for us to have git-config(1) warn users when a
config file is entirely missing.
|
|
Newly-cloned epochs need to be in alternates file of
all.git for the web and NNTP interfaces to work. So
allow invocations of "public-inbox-index" to idempotently
ensure the epoch is visible from the all.git repo.
|
|
We'll be using this sub to fill $GIT_DIR/objects/info/alternates
if somebody uses clone --mirror, too
|
|
All of our internal epoch rollover calculations are done using
the estimated unpacked (and uncompressed) size of the repo. The
importer instance needs to check that unpacked size before
selecting an epoch when an epoch already has packed data.
This bug did not impact the initial mass imports since we only
initialize the Import instance once-per-epoch and did not need
to take existing epochs into account.
Tested manually with -mda on a local clone of LKML
Reported-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
CSS specified by the BOFH must never take precedence over
what a user sets in userContent.css.
|
|
The sample userContent.css needs a higher priority than what
the BOFH specifies. In other words, user preference must
ALWAYS take precedence.
Reported-by: Dmitry Alexandrov <321942@gmail.com>
cf. https://public-inbox.org/meta/87mumn4kx8.fsf@gmail.com/
|