This is the last commit in the series of commits for completing the
project of cleaning up our html templates to have 4 space and
valid indentation.
Fixes: #1236.
In this commit we also fix a test which would fail as a result of
doing this cleanup since the test wasn't designed to take into
account the space chars which might occur in the beginning of a
html line.
This test randomly fails far too often in Travis -- I think more than
all our other tests combined. It needs to be fixed before we can ask
everyone to look at build failures it causes.
Now that we have `eval_rst` and can explicitly exclude pages from the
toctree completely, we no longer need to set `includehidden`, and we
can return to using upstream's template.
(Meanwhile, our feature request upstream was successful! See
rtfd/sphinx_rtd_theme#485, which upstream implemented just a week
after we requested it. So that would have been another option.)
This reverts commit 11b8b8f48 "docs: Add rtd layout template."
It runs in kind of a peculiar environment -- in particular with the
`tags` identifier injected into the namespace -- and it contains
very little code more complex than `foo = "bar"`, so there's not
much to check anyway.
Before this fix, the installer has an extremely annoying bug where
when run inside a container with `lxc-attach`, when the installer
finishes, the `lxc-attach` just hangs and doesn't respond even to
C-c or C-z. The only way to get the terminal back is to root around
from some other terminal to find the PID and kill it; then run
something like `stty sane` to fix the messed-up terminal settings
left behind.
After bisecting pieces of the install script to locate which step
was causing the issue, it comes down to the `service camo restart`.
The comment here indicates that we knew about an annoying bug here
years ago, and just swept it under the rug by skipping this step
when in Travis. >_<
The issue can be reproduced by running simply `service camo restart`
under `lxc-attach` instead of the installer; or `service camo start`,
following a `service camo stop`. If `lxc-attach` is used to get an
interactive shell, these commands appear to work fine; but then when
that shell exits, the same hang appears. So, when we start camo
we're evidently leaving some kind of mess that entangles the daemon
with our shell.
Looking at the camo initscript where it starts the daemon, there's
not much code, and one flag jumps out as suspicious:
start-stop-daemon --start --quiet --pidfile $PIDFILE -bm \
--exec $DAEMON --no-close -c nobody --test > /dev/null 2>&1 \
|| return 1
start-stop-daemon --start --quiet --pidfile $PIDFILE -bm \
--no-close -c nobody --exec $DAEMON -- \
$DAEMON_ARGS >> /var/log/camo/camo.log 2>&1 \
|| return 2
What does `--no-close` do?
-C, --no-close
Do not close any file descriptor when forcing the daemon
into the background (since version 1.16.5). Used for
debugging purposes to see the process output, or to
redirect file descriptors to log the process output.
And in fact, looking in /proc/PID/fd while a hang is happening finds
that fd 0 on the camo daemon process, aka stdin, is connected to our
terminal.
So, stop that by denying the initscript our stdin in the first place.
This fixes the problem.
The Debian maintainer turns out to be "Zulip Debian Packaging Team",
at debian@zulip.com; so this package and its bugs are basically ours.
This is a tool that throws away `fsync` calls and other requests for
the system to sync files to disk. It may make the install faster; for
example, if it has to install a number of system packages, `dpkg` is
known to make a lot of `fsync` calls which slow things down
significantly. Conversely, if there's a power failure in the middle
of running a test install, we really don't mind if the test install's
data becomes corrupt.
When the install script is successful, one of the final things it
wants to do is to move the tree that Zulip was installed from into the
deployments directory. It can't do that, at least not in a naive way
with `mv`, if the tree is actually a mount point. So, stick the tree
inside some other directory that we create just for the purpose of
being the mount point and containing the install tree.
This saves several minutes off the install time. Sadly pip still
clones Git repos for dependencies that point to them, but for many
others (not all? not sure) it just gets a wheel from the cache.
The third-party `install-yarn.sh` script uses `curl`, and we invoke it
in `install-node`. So we need to install it as a dependency.
We've mostly gotten away with this because it's common for `curl` to
already be installed; but it isn't always.
This greatly simplifies iterating on changes to the installer and
associated code: just edit in the shared directory (or edit in your
worktree and rsync to the directory), and rerun.
With this change, the form with a directory is now really the main
way to run the script; the form accepting a tarball is really just
a convenience feature, unpacking the tarball and then proceeding with
that directory.
This will facilitate testing interesting installer features
using its own CLI.
On my laptop, with a recent base image (updated a few days ago with
`prepare-base`), it takes just 7 or 8 seconds to get to the installer
running, as timed by passing `--help` so that the installer promptly
exits.
In order to do development on the installer itself in a sane way,
we need a reasonably fast and automatic way to get a fresh environment
to try to run it in.
This calls for some form of virtualization. Choices include
* A public cloud, like EC2 or Digital Ocean. These could work, if we
wrote some suitable scripts against their APIs, to manage
appropriate base images (as AMIs or snapshots respectively) and to
start fresh instances/droplets from a base image. There'd be some
latency on starting a new VM, and this would also require the user
to have an account on the relevant cloud with API access to create
images and VMs.
* A local whole-machine VM system (hypervisor) like VirtualBox or
VMware, perhaps managing the configuration through Vagrant. These
hypervisors can be unstable and painfully slow. They're often the
only way to get development work done on a Mac or Windows machine,
which is why we use them there for the normal Zulip development
environment; but I don't really want to find out how their
instability scales when constantly spawning fresh VMs from an image.
* Containers. The new hotness, the name on everyone's lips, is Docker.
But Docker is not designed for virtualizing a traditional Unix server,
complete with its own init system and a fleet of processes with a
shared filesystem -- in other words, the platform Zulip's installer
and deployment system are for. Docker brings its own quite
different model of deployment, and someday we may port Zulip from
the traditional Unix server to the Docker-style deployment model,
but for testing our traditional-Unix-server deployment we need a
(virtualized) traditional Unix server.
* Containers, with LXC. LXC provides containers that function as
traditional Unix servers; because of the magic of containers, the
overhead is quite low, and LXC offers handy snapshotting features
so that we can quickly start up a fresh environment from a base
image. Running LXC does require a Linux base system. For
contributors whose local development machine isn't already Linux,
the same solutions are available as for our normal development
environment: the base system for running LXC could be e.g. a
Vagrant-managed VirtualBox VM, or a machine in a public cloud.
This commit adds a first version of such a thing, using LXC to manage
a base image plus a fresh container for each test run. The test
containers function as VMs: once installed, all the Zulip services run
normally in them and can be managed in the normal production ways.
This initial version has a shortage of usage messages or docs, and
likely has some sharp edges. It also requires familiarity with the
basics of LXC commands in order to make good use of the resulting
containers: `lxc-ls -f`, `lxc-attach`, `lxc-stop`, and `lxc-start`,
in particular.
Stripe Checkout means using JS code provided by Stripe to handle
almost all of the UI, which is great for us.
There are more features we should add to this page and changes we
should make, but this gives us an MVP.
[greg: expanded commit message; fixed import ordering and some types.]
This is fairly often -- though not always! -- failing, with a nasty
failure mode where it takes like 6 minutes to time out. See
discussion on #7748 (search for "bad link").
Actually, after seeing it happen just now when running
test-documentation on my laptop, on some other link, it occurs to me
that I've seen this before -- it's fairly common in Travis, too. It's
just that it doesn't actually cause the build to fail :-/, and on
Travis we haven't been paying as close attention to slow builds as we
are on Circle right now.
Generally stderr is the conventional place for this sort of running
commentary, and it's better set up for it: by default stdout may have
a buffer inside the process so that things written to it don't reach
the outside until later, while stderr is always by default unbuffered,
so messages are printed immediately.
Here, until the previous commit, because our color-reset sequence was
being printed without a following newline (with `echo -n`), it was
getting buffered; and then error messages from `scrapy` to stderr were
being erroneously painted with the color intended for the message
"Testing links in documentation...".
The autoreload code of Django works by looping over the files associated
with all the loaded modules. This loop is run after every 1 second. If
the file is found for the first time by the loop, it is assumed that the
file is new and is not modified between the time it is loaded and is
checked by the loop. This assumption is the source of a race condition.
We can either implement a more sensitive version of the loop or we can
just allow enough time to the Django loop to touch every file at least
once.
For the time being, we are going with the second option.
Previously, there were following problems with the implmentation:
* Same file handle was being used to read and write. We used to do
`seek(0)` and then `read()`. This had a chance to overwrite
file data. Now we use different file handles to read and write data.
* We were using text streams. Text streams cannot be used with
`bufferring=0`. Now we use binary streams without buffering so that
data is available for reading without any delay.
This commit also updates the key(s) that we search in the logfile.
Previously, launch of all queues was announced in the log, now we only
anounce the number of threads that were launched.
This commit also makes sure that we always exit after gracefull shutting
down the development server.
Previously, there were following problems with the implmentation:
* Same file handle was being used to read and write. We used to do
`seek(0)` and then `read()`. This had a chance to overwrite file
data. Now we use different file handles to read and write data.
* We were using text streams. Text streams cannot be used with
`bufferring=0`. Now we use binary streams without buffering so that
data is available for reading without any delay.
This commit just copies all the code from MissedMessageSendingWorker
class to a new EmailSendingWorker class. All the logic to send an email
through a queue was already there. This commit only makes the logic
generic. It does so by creating a special purpose queue called
'email_senders' to send any type of email. To make
MissedMessageSendingWorker still work we derive it from
EmailSendingWorker. All the tests that were testing
MissedMessageSendingWorker now run against EmailSendingWorker.
We get the following error (edited slightly):
Dec 19 06:13:27 commit_messages| An error occurred while executing
'/usr/bin/git rev-list --max-count=-1 upstream/master..HEAD':
fatal: ambiguous argument 'upstream/master..HEAD':
unknown revision or path not in the working tree.
Use '--' to separate paths from revisions, like this:
'git <command> [<revision>...] -- [<file>...]'
We'll need to adjust the remotes to make `upstream` mean what we expect.
This isn't really the right way to do this -- commit
dcd80e665 "travis/setup-backend: Remove the '--travis' flag"
took us in the wrong direction by introducing more magic, deeper in
the stack.
But it's the same way we do it for Travis. For now, just copy that.
[Thanks to hackerkid for cleaning up my original crude hack.]
At this point if we were accidentally using `/srv/zulip-venv` for
anything, we'd have run into it by now. So just drop the bit of
historical logic that we had to ensure that.
This reverts commit 66261f1cc. See parent commit for reason; here,
provision worked but `tools/run-dev.py` would give errors.
We need to figure out a test that reproduces these issues, then make a
version of these changes that keeps that test working, before we
re-merge them.
We should omit these for mypy. For most class definitions,
mypy doesn't need `Any`, and it provides no real useful info.
For clever monkeypatches, you should provide a more specific
type than `Any`.
These commands are super boring standard Docker commands,
so this probably isn't helpful for anyone who is familiar
with building Docker images... but I had to consult docs
to work out the right commands again today, so they'd help me.
Provision was failing at tools/setup/postgres-init-dev-db
with this in the log:
Sep 20 02:27:01 + sudo -i -u postgres psql ''
[sudo] password for circleci:
The issue is that the old version of this line (from Circle upstream)
only lets the `circleci` user sudo to root -- not to other users, or
not directly anyway -- because sudoers syntax is complicated. Fix it,
after studying `man sudoers`.
This is only an optimization -- if this list is missing anything,
we'll get to it in `provision` in the actual build. That's important,
because we want an existing image to work fine for testing new
versions of our codebase, including changes that may install more
packages in `provision`.
What this does accomplish is keeping provision's `apt-get install`
fast, by leaving it very little work to do.
The list comes from looking at the APT output during provision in an
actual run without this step, and leaving out two packages which
aren't available at this stage, because we get them from PPAs:
postgresql-9.3-pgroonga and postgresql-9.3-tsearch-extras.
Install `jq` with APT -- that's a lot simpler to read than this
explicit download.
And coalesce several commands, following Docker upstream's
recommendation and avoiding unnecessary overhead.
This is nearly the same as Circle's version, linked in the comment.
I've
* changed the FROM line to get Ubuntu,
* added a couple of distro packages to compensate, and
* revised the comments.
This seems to have been causing the travis production suite to fail.
It's a direct consequence of removing travis' giant library of apt
sources.list files; now that those are gone, there aren't copies of
all these extra packages available anyway.
This method was new in Tornado 4.0. It saves us from having to get
the time ourselves and do the arithmetic -- which not only makes the
code a bit shorter, but also easier to get right. Tornado docs (see
http://www.tornadoweb.org/en/stable/ioloop.html) say we should have
been getting the time from `ioloop.time()` rather than hardcoding
`time.time()`, because the loop could e.g. be running on the
`time.monotonic()` clock.
tools/setup/generate_zulip_bots_static_files now starts off by
deleting static/generated/bots/ (if it hasn't been removed already)
so that outdated static files from older versions of the zulip_bots
package don't supress errors in the main repo that would otherwise
break.
For more info, see #7542Fixes: #7542.
The `re.match` function in the Python stdlib is a trap for the unwary,
with surprising and asymmetrical semantics; we should probably add a
lint rule to ban it entirely. The docstring says:
> Try to apply the pattern at the start of the string, [...]
In other words, it effectively adds a `^` at the start (or `\A`, where
the distinction matters.) It's bad enough that this differs from what
grep, sed, perl, less, and every other tool I can think of do when
looking for matches to a regex; on top of that, it treats the
beginning of the string differently from the end, for no obvious
reason. The function that does what the rest of the world understands
by "match against this regex" is `re.search`.
In this case, it's unlikely that anyone intended for comments with
URLs, or `api_url` references, to miss out on their respective
exceptions to the long-line rule if they happen to start after the
first column. So fix those rules by just switching to `re.search`
with the same pattern.
I think Markdown URL references may have to start at the beginning of
the line, so I've left a `^` there to preserve -- but now make
explicit -- the `re.match` behavior.
Tweaked by tabbott to move changes from the next commit that are
required for this to pass tests into this commit.
Note that this exports a few items that were not previously exported.
Aditya Bansal
Greg Price
Tim Abbott
Rhea Parekh
Vishnu Ks
Steve Howell
Rohitt Vashishtha
Cynthia Lin
Umair Khan
rht
cPhost
Tommy Ip
Symt
picapi_
Eeshan Garg