This fixes a small memory leak in our queue workers, where we don't
reset the accumulated times contained in our query logging data.
Longer-term, we may want to make something mergable for mainline where
we only store on the connection object the totals; that would be a
fixed amount of emmory per connection and thus not have this problem.
(imported from commit 914fa13acfb576f73c5f35e0f64c2f4d8a56b111)
Now we can nest fenced code/quote blocks inside of quote
blocks down to arbitrary depths. Code blocks are always leafs.
Fenced blocks start with at least three tildes or backticks,
and the clump of punctuation then becomes the terminator for
the block. If the user ends their message without terminators,
all blocks are automatically closed.
When inside a quote block, you can start another fenced block
with any header that doesn't match the end-string of the outer
block. (If you don't want to specify a language, then you
can change the number of backticks/tildes to avoid amiguity.)
Most of the heavy lifting happens in FencedBlockPreprocessor.run().
The parser works by pushing handlers on to a stack and popping
them off when the ends of blocks are encountered. Parents communicate
with their children by passing in a simple Python list of strings
for the child to append to. Handlers also maintain their own
lists for their own content, and when their done() method is called,
they render their data as needed.
The handlers are objects returned by functions, and the handler
functions close on variables push, pop, and processor. The closure
style here makes the handlers pretty tightly coupled to the outer
run() method. If we wanted to move to a class-based style, the
tradeoff would be that the class instances would have to marshall
push/pop/processor etc., but we could test the components more
easily in isolation.
Dealing with blank lines is very fiddly inside of bugdown.
The new functionality here is captured in the test
BugdownTest.test_complexly_nested_quote().
(imported from commit 53886c8de74bdf2bbd3cef8be9de25f05bddb93c)
This should address user reports of huge bankruptcy counts even when
they are relatively caught up. The root issue is that we sometimes
don't mark messages as read for some reason.
(imported from commit 8799305a8665f9ee239575e6e95f603f89c1d427)
This reverts commit 1147814b22fb9737a807057ddbdbe0e9554086e0.
This seems to with some probability screw up our Zephyr mirroring
script.
(imported from commit 4f82452f1b0ca98e6b895db020e071d2daa325e4)
This requires a puppet apply on each of staging and prod0 to update
the nginx configuration to support the new URL when it is deployed.
(imported from commit a35a71a563fd1daca0d3ea4ec6874c5719a8564f)
Client objects are immutable and there are very few of them, so caching them in
memory is cheap and saves a trip to memcached.
(imported from commit 300b9b402f4e509f86a7fd86b5f898dc3f43738f)
UserProfile.show_admin was intended to be a check for users that have
administrative rights in other realms, which we've harmlessly but
erroneously been using to check if they are an admin in their realm.
Use the more straightforward check instead, with a more intuitive
name.
(imported from commit d81050c7dbbb19e59c5e31750be303a4630e1456)
There will be browser errors on staging when this is deployed due to the socket
protocol changing.
(imported from commit f1eda5b5c2ec9c60c23b3ca96277a61debadf5bb)
I believe there may also be others. I'm still not sure why clients would be
sending open requests without session or csrf values in their cookies, though.
(imported from commit 7e9660c1c4d5c2abf55ff21b433ba0117180eb82)
Previously, we counted not just the time required to process a
particular request, but also the time required to import+find the view
function via urls.py. The latter is usually fast, but when a new
Django thread starts up, it can take significant time, resulting in us
flagging slow requests on each server restart and also when a new
Django thread starts up on prod to handle requests.
This change means that we no longer include that startup time as part
of request processing time -- but we still log it in the case that it
was more than 5ms, so that we can identify why a particular request
was slower than expected if high startup times become a problem. We
annotate the time in log lines as "+start" rather than just "start" to
make clear that it's not counted in the total.
(imported from commit c677682e23b26005060390d85d386234f4f463ad)
This is useful for the occasional case where we cannot figure out what
is causing a particular problem, but it can be easily reproduced on
staging.
(imported from commit 8b51184a8b686814f2c6ff103ba355538463ceb0)
Jessica McKellar
Tim Abbott
Steve Howell
acrefoot
Zev Benjamin
Kevin Mehall