This was mainly a matter of reducing timeouts and delays, paying
special attention to every test that took longer than a second to
finish.
We could do better here; IMO anything over .7 sec is probably too
long, but it's a big win as it is.
Remember, interactive computing is a big win over batch processing:
anything that makes you get up and walk away from the terminal might
as well be making you carry your punch cards over to the mainframe.
As originally written, the test would only pass if the accept()
callbacks for the evconnlistener were all invoked before the last of
the CONNECTED/ERROR callbacks for the connecting/resolving bufferevent
had its call to event_base_loopexit() complete. But this was only
accidentally true in 2.0, and might not be true at all in 2.1 where
we schedule event_base_once() callbacks more aggressively.
Found by Sebastian Hahn.
You could previously do this with evbuffer_peek() and some memcpys,
but it was a bit more work than most folks wanted to get into.
Closes sourceforge ticket 3108072
Some people use event_base_once(EV_TIMEOUT) to make a callback get
called "immediately". But this is pretty roundabout: it uses the
timeout heap to immediately put the event onto the active queue, when
it could just use event_active. Additionally, it can lead to
surprising re-ordering behavior.
This patch changes event_base_once so it bypasses event_add() and
called event_active() directly on a pure-timeout event with an empty
timeout.
Add a zero check to the function `event_mm_malloc_',
i.e. simply return NULL if the sz argument is zero.
On failure, set errno to ENOMEM and return NULL.
Add a zero check to the function `event_mm_calloc_',
i.e. simply return NULL if either argument is zero.
Also add an unsigned integer multiplication check, and if an integer
overflow would occur, set errno to ENOMEM and return NULL.
On failure, set errno to ENOMEM and return NULL.
Add a NULL check to the function `event_mm_strdup_',
i.e. set errno to EINVAL and return NULL.
Also add an unsigned integer addition check, and if an integer
overflow would occur, set errno to ENOMEM and return NULL.
If a memory allocation error occurs, again set errno to ENOMEM
and return NULL.
Add unit tests to `test/regress_util.c'.
Sometimes DNS reply has nothing but query section. It does not look like
error, so it should be treated as NODATA with TTL=0 as soon as there is
no SOA record to deduce negative TTL from.
Instead of having a file segment born as one type and stay that way
forever, let them start out unmapped, but map themselves as needed
if they need to get written out on a non-drains_to_fd evbuffer.
The sendfile() implementation for evbuffer_add_file is potentially more
efficient, but it has a problem: you can only use it to send bytes over
a socket using sendfile(). If you are writing bytes via SSL_send() or
via a filter, or if you need to be able to inspect your buffer, it
doesn't work.
As an easy fix, this patch disables the sendfile-based implementation of
evbuffer_add_file on an evbuffer unless the user sets a new
EVBUFFER_FLAG_DRAINS_TO_FD flag on that evbuffer, indicating that the
evbuffer will not be inspected, but only written out via
evbuffer_write(), evbuffer_write_atmost(), or drained with stuff like
evbuffer_drain() or evbuffer_add_buffer(). This flag is off by
default, except for evbuffers used for output on bufferevent_socket.
In the future, it could be interesting to make a best-effort file
segment implementation that tries to send via sendfile, but mmaps on
demand. That's too much complexity for a stable release series, though.
