Previously, we were parsing the Status Line
and the header, both together. This of course
doesn't work when we had removed the Status
Line already. This wasn't an issue in parsing
responses from servers, but doesn't work when
parsing requests from clients (i.e. in a server).
Also, it's simpler.
Tests extended accordingly.
Change-Id: Id1c9a6385080c86b6072130fa8a169a699c42462
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
Here we are more tolerant with regards to the
minimum Status Line length, since a Reason
Phrase is informative and might be omitted.
While technically it should be provided, per
the RFC, we shouldn't break when it's missing.
Change-Id: Ic702db61e5bf4272e18f09d127405033277e5943
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
This adds support to parse client request
on the server side.
Unit-tests included.
Change-Id: I90e9ad3007a3215682991fd2383362e1c48589a8
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
There are scenarios where startThread is
called either when the thread is already
running or when it has long since finished.
Logs are added to catch them, and better
document the expected behavior in those
cases (they shouldn't happen).
Change-Id: I219a59b92c943445ec4520667a8ed6d9bd1c328d
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
And guard http data dumping with debug directives.
Change-Id: I22a725ba49bfb0399a27889ce9732dfe061e2563
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
The lifetime management of static objects
is extremely unpredictable and depends on
many variables outside of our control or
even reliable reproducibility.
Complex static objects that own threads
and other objects are doubly problematic
because of their dependency and/or
interaction with other objects.
Here we replace the static DelayPoll
instance with one we control its lifetime
in the LOOLWSD main body, such that it
is destroyed properly.
Specifically, DelayPoll's dtor was
accessing Poco's Logging subsystem out of
order. That is, after Poco had been
destroyed.
Another advantage to this approach is that
we don't even create the DelayPoll at all
if we don't need it. The onus now is on
the user of DelayPoll to make sure they
create a Delay object with a long-enough
lifetime to encompase it use.
For completeness, here is the stacktrace:
terminate called after throwing an instance of 'std::bad_alloc'
what(): std::bad_alloc
Program received signal SIGABRT, Aborted.
__GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51
51 ../sysdeps/unix/sysv/linux/raise.c: No such file or directory.
(gdb) bt
#0 __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51
#1 0x00007ffff613f801 in __GI_abort () at abort.c:79
#2 0x00007ffff6d51957 in ?? () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#3 0x00007ffff6d57ae6 in ?? () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#4 0x00007ffff6d56b49 in ?? () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#5 0x00007ffff6d574b8 in __gxx_personality_v0 () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#6 0x00007ffff671f573 in ?? () from /lib/x86_64-linux-gnu/libgcc_s.so.1
#7 0x00007ffff671fad1 in _Unwind_RaiseException () from /lib/x86_64-linux-gnu/libgcc_s.so.1
#8 0x00007ffff6d57d47 in __cxa_throw () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#9 0x00007ffff6d582dc in operator new(unsigned long) () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#10 0x00005555556b5927 in void std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_construct<char*>(char*, char*, std::forward_iterator_tag) [clone .isra.41] ()
#11 0x0000555555982a14 in Poco::Message::Message(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, Poco::Message::Priority) ()
#12 0x000055555585a909 in SocketPoll::~SocketPoll (this=0x555555d10f60 <DelayPoll>, __in_chrg=<optimized out>) at net/Socket.cpp:145
#13 0x00007ffff6142041 in __run_exit_handlers (status=0, listp=0x7ffff64ea718 <__exit_funcs>, run_list_atexit=run_list_atexit@entry=true, run_dtors=run_dtors@entry=true) at exit.c:108
#14 0x00007ffff614213a in __GI_exit (status=<optimized out>) at exit.c:139
#15 0x0000555555752d78 in LOOLWSD::innerInitialize (this=<optimized out>, self=...) at wsd/LOOLWSD.cpp:1213
#16 0x0000555555788b07 in LOOLWSD::initialize (this=<optimized out>, self=...) at wsd/LOOLWSD.hpp:432
#17 0x00005555558dd7e4 in Poco::Util::Application::run() ()
#18 0x00005555556b6574 in main (argc=2, argv=0x7fffffffe528) at wsd/LOOLWSD.cpp:4276
Change-Id: Ifda55fe869fa6734b9c2490da4497d2551ac21c1
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
Makes it accessible to more code.
Change-Id: I73dd1895defe47ae40873b6155203768055206ec
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
Especially error reporting is now better.
Change-Id: I032f6e8403660c2ac24be2d80b53b63831ec1066
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
Normally writeOutgoingData flushes all the available
write buffer to the socket. However, when an error
occures during the write, the flushing loop is
broken and the error is expected to be handled
in the poll loop (or, at any rate, the poll loop
will sooner or later retry).
This doesn't work when the socket is being shut
down. Details and stack-trace is added in the
code to document this scenario.
The proper fix to flushing before shutting
down is to do it asynchronously (i.e. flag for
shutdown, let poll flush the buffers, and then
close the socket), but currently this isn't
easily possible (again, details in the added
comment). For now, we do a second attempt to
write, before we give up and warn of lost data.
Worth noting that this was caught thanks to the
simulated socket error logic. Every so often the
Close Frame of the WebSocket was getting lost.
This frame is asserted in the 'crash test' and
the explanation is that it happened when the
simulated socket error happened to hit during
the writing of the Close Frame (as one statistic
this happened ~15% of the time, when running the
crash test in a loop). With the retrying in this
patch I was able to run the crash test successfully
a few 100 times without failures.
Finally, we unify the flushing with the mobile-app case
to minimize the differences as much as possible and
improved the logging a bit.
Change-Id: I0c0559d6095cd2af8250a7e65a972ba672ecd4b1
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
SSL signals whether it needs more data to read or write
so that we perform the appropriate polling and call
the respective API. Since during error simulation
we skip calling the SSL API altogether, we shouldn't
change the last desired operation it communicated to us.
This becomes clear when we consider that we always read,
unconditionally, in StreamSocket::handlePoll (but write
only when poll signals POLLOUT event). This means that
if we decide to simulate a socket error during such
uncoditional read, and SSL wants to write, we would
reset its last wantsTo flag from write to read!
Logically, since we aren't even invoking any SSL API
during a simulated error, we shouldn't change the
SSL state one way or the other.
Change-Id: Ie4ddbc959f8424ea4ac3bc52f30edbc7399b2c7a
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
The API is the same for both SSL and non-SSL sockets
(they are virtual functions), and the API relies on
errno to detect errors and decide to retry, log, or
break. As such, we must set the same EAGAIN when
simulating socket errors for both SSL and non-SSL
cases, so that the behavior would also be the same.
Change-Id: Ib1695cc94d93f5ecc53d7b22872f459ac8b11bbd
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
This converts Poco to http::Request in sync mode,
thereby not changing functionality. In a follow-up,
this will be converted to async.
Change-Id: Ifbecd44ff599394799c1131470d77f803ed8cc45
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
This guarantees that any early failure results
in the status to also be invalid. That is,
only when we do get 200 OK from the host do
we set it in the StatusLine, never by default.
Change-Id: I4e31cb48cce0ed8962a737c4784fd6453b9a48cf
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
Limit its length, we know the version length anyway.
Signed-off-by: Miklos Vajna <vmiklos@collabora.com>
Change-Id: I0db2b9227baf3e10055082ad394c0f555b9898e1
This implements HTTP/1.1 per RFC 7230, partially.
Unit-tests are provided with documentation on usage.
This is desgined to serve as the http implementation
throughout loolwsd, for both synchronous and
asynchronous requests.
Change-Id: Iaf1b8c5fcb8cec032445e27c9f70d2fb807aa4dc
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
Move the connect function into the NetUtil
translation unit to aid using it for the
upcoming async socket logic.
The NetUtil should also come in handy for
the miscellaneous network helpers we have.
Change-Id: I2ee0c6e3e1769fd87572d7407d3b4979b59ffe6a
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
When we simulate socket error it's hard to understand
what is going on, since there is no real errno to
explain the -1 return value. This is at least true
in the SSL case where we don't set errno (perhaps
we should?). At any rate, it's good form to log
that we are messing up with the return value and
that the real socket API hasn't been even invoked.
Change-Id: Ib76f0259dbdcfe0cfae97343c11ca45461079fa1
Signed-off-by: Ashod Nakashian <ashod.nakashian@collabora.co.uk>
