Cleaning up the thread variable with the shared string stream is
something of a nightmare, for a rather marginal gain.
==9296== Invalid write of size 1
...
==9296== by 0x738C092: str (sstream:195)
==9296== by 0x738C092: std::__cxx11::basic_ostringstream<char, std::char_traits<char>, std::allocator<char> >::str(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) (sstream:649)
==9296== by 0x65383A: Log::beginLog[abi:cxx11](char const*) (Log.cpp:141)
==9296== by 0x551823: Admin::~Admin() (Admin.cpp:381)
==9296== by 0x7D9ECF7: __run_exit_handlers (exit.c:83)
==9296== by 0x7D9ED49: exit (exit.c:105)
==9296== by 0x7D86F50: (below main) (libc-start.c:342)
==9296== Address 0x8ba41c0 is 0 bytes inside a block of size 513 free'd
==9296== at 0x4C2FA1D: operator delete(void*) (vg_replace_malloc.c:576)
...
==9296== by 0x738784A: ~basic_stringbuf (sstream:65)
==9296== by 0x738784A: std::__cxx11::basic_ostringstream<char, std::char_traits<char>, std::allocator<char> >::~basic_ostringstream() (sstream:591)
==9296== by 0x7D9F27E: __call_tls_dtors (cxa_thread_atexit_impl.c:155)
==9296== by 0x7D9EC0A: __run_exit_handlers (exit.c:41)
==9296== by 0x7D9ED49: exit (exit.c:105)
==9296== by 0x7D86F50: (below main) (libc-start.c:342)
Good to log during shutdown / exit.
This reverts commit c315d219d5967f23fb1769e78021f61b8f9da6ec.
This reverts commit ce78fec310.
Change-Id: Ia4a15be336d89d8d883530943724d48e4b0ec9fe
Reviewed-on: https://gerrit.libreoffice.org/71444
Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
Tested-by: Michael Meeks <michael.meeks@collabora.com>
Avoids some N^2 log-line explosion; also make the method name
more findable.
Change-Id: I3ee8c521f1ac98a939cd4d758c720b577d3bfa57
Reviewed-on: https://gerrit.libreoffice.org/71443
Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
Tested-by: Michael Meeks <michael.meeks@collabora.com>
This is faster and reduces memory fragmentation.
Also, cleans up the logging macros and implementation.
Change-Id: I7fb00da041d1261c694c4b48b67a3c66ad0cbf8d
Reviewed-on: https://gerrit.libreoffice.org/71020
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
The former is the standard C++ approach
and is reportedly faster than __thread
(at least with gcc).
Change-Id: Ibdefd32172774a280637f73dd062282b7bf62025
Reviewed-on: https://gerrit.libreoffice.org/71019
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
The default deflate level of 6 is quite slow
and the benefits are hardly worth the high
latency that users experience.
Tested on a writer document with some small
images and a few pages of text:
Level 4 gives virtually identical compression
ratio to level 6, but is between 5-10% faster.
Level 3 runs almost twice as fast as level 6,
but the output is typically 2-3x larger.
Perhaps this should be exposed via config
so it would be possible to reduce latency
due to compression when CPU is scarce but
network bandwidth ample, and vice versa.
Change-Id: Iba88eea8f180d11458b33c68389e797234df1a60
Reviewed-on: https://gerrit.libreoffice.org/71018
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
Unix Domain Sockets are inaddressable remotely, and more efficient,
as well as allowing future SCM_CREDENTIALS / SCM_RIGHTS.
Change-Id: Ia2472260f75feb43e9022cdfa0fe005ccd489454
(What we cache is also the textual data: URLs even if we store them
using .png file names.)
This avoids the current back-and-forth-encoding: First we
base64-encode the complete binary "tile:" message (one text line
followed by a newline and the binary PNG) to pass to WebKit, then in
the JavaScript snippet passed to WebKit we decode the base64 and turn
it into an ArrayBuffer, and then we unpack the ArrayBuffer and encode
the PNG part to use as a data: URL.
Also fix unexpected concatenation error in Poco::URI::encode generating
eg. authorid=localhost0 xauthorid=localhost0localhost0 in the output.
Change-Id: I560e47e31884eeb1c662f468436ed7541cfb082d
(Note that when I say 'NUL' I mean the ASCII character called NUL,
i.e. a zero byte. Not to be confused with 'NULL'.)
Why FatalGdbString has to be a C style fixed size char array I don't
know. Or wait, I do know. Because SPEED!!! And using C strings safely
is trivial.
Change-Id: Id28b00a6e3219cf6f015c4209732f33216f83b22
A quite common logic that is best moved to a utility
and optimized for best performance.
Includes unit-tests.
Change-Id: Id63a388690c17355eb2fac529070c38e9b082fd0
This is needed so we can use this inside ChildSession.
Change-Id: I88f2cc767412fd52dbb242938f0f9897d4277639
Reviewed-on: https://gerrit.libreoffice.org/63836
Reviewed-by: Tomaž Vajngerl <quikee@gmail.com>
Tested-by: Tomaž Vajngerl <quikee@gmail.com>
Add an entry to discovery.xml with the urlsrc where capabilities end
point can be found. Use json format to send back the feature list.
Change-Id: I390a53d956d53ca79e5a8090aead7f4131ec4ca0
Also support anonymization of downloadas documents
and renaming of documents.
Reviewed-on: https://gerrit.libreoffice.org/57541
Reviewed-by: Jan Holesovsky <kendy@collabora.com>
Tested-by: Jan Holesovsky <kendy@collabora.com>
(cherry picked from commit 78248a542c9ca31bf9ad4cad9b55d78690384395)
Change-Id: I81a80e6290217659987d73f625e5f0fb81cb7ef2
This is important for when we abort with some explanation.
Often said explanation doesn't show up anywhere to be useful.
Also, issue fatal logs for abnormal exist and use SFL to log errno.
Reviewed-on: https://gerrit.libreoffice.org/57540
Reviewed-by: Jan Holesovsky <kendy@collabora.com>
Tested-by: Jan Holesovsky <kendy@collabora.com>
(cherry picked from commit ad7964393eadb68873b820e0a620fb40f1e1b06a)
Change-Id: Ic67064ef40ef6e93d26e5847ecd32bdd49c3cc8b
The async-signal-safe functions to get thread-id
and thread-name, which cache the results, are
faster, cleaner, and signal-safe. No reason why
we shouldn't always use them.
Especially since it appears the logic was
inverted in Log::prefix, such that the signal
un-safe calls were made during signal-handling,
and the safe ones were called otherwise!
Instead of passing the signal-safe flag to
Log::prefix, we pass the buffer size, for
improved security.
Furthermore, reduce header dependencies
and reduce clutter.
Change-Id: I697689b2f0a290b6d8cce4babc3ac1e576141da6
The idea is that it would work sufficiently identically, so that even
people without a Mac and without an iOS device could participate in
development of the non-iOS-specific bits, like the JavaScript, or the
online MOBILEAPP-specific plumbing. Which would be great.
No, this doesn't do anything sane yet. It does compile the same online
C++ files as the iOS app, though. (Some minor tweaks were needed in a
couple of them to silence gcc warnings.)
There is a plain Makefile, but I should change to using autofoo, too.
Eventually, this will need to be built in a separate tree from a
normal online, just like when using the --enable-iosapp configure
switch. (But for now, doesn't matter.)
Change-Id: I13e4d921acb99d802d2f9da4b0df4a237ca60ad6
We don't have any user-generated signals to handle by shutting down in
an app.
One less thing to worry about. Now it's just the global
TerminationFlag that is problematic when the code runs in just one
process.
We try to decrease the network usage with avoiding sending out
to much tiles to the client. When we already sent out two versions
of the same tile without having the tileprocessed message from the
client we delay sending out the next version to avoid spamming tiles
on the network.
Change-Id: Ia47cd7c0d3fb829f6777f0c3265970433591df19
Re-think the plumbing between the different parts of the C++ Online
code. Do try to have it work more like in real Online on all but the
lowest socket level. Except that we don't have multiple processes, but
threads inside the same process. And instead of using actual system
sockets for WebSocket traffic between the threads, we use our own
FakeSocket things, with no WebSocket framing of messages.
Reduce the amount of #ifdef MOBILEAPP a bit also by compiling in the
UnitFoo things. Hardcode that so that no unit testing is ever
attempted, though. We don't try to dlopen any library.
Corresponding changes in the app Objective-C code. Plus fixes and
functionality improvements.
Now it gets so far that the JavaScript code thinks it has the document
tiles presented, and doesn't crash. But it hangs occasionally. And all
tiles show up blank.
Anyway, progress.
Change-Id: I769497c9a46ddb74984bc7af36d132b7b43895d4
The app is unimaginatively called "Mobile" for now.
Runs but crashes pretty quickly after loading the document by the LO
core. Will need some heavy changes to get a ClientSession object
created in there, too, to handle the (emulated) WebSocket messages
from the JavaScript. It would then handle some of these messages
itself, and forwards some to the ChildSession, which in this case is
in the same process. Now the messsages from the JavaScript go to a
ChildSession, which is wrong. As the assertion says, "Tile traffic
should go through the DocumentBroker-LoKit WS"
Re-think Linux vs mobile ifdefs a bit. Use #ifdef __linux only to
surround code that actually is Linux-specific. Use #ifdef MOBILEAPP
for code that is for a mobile version (with no separste wsd, forkit,
and kit processes, and with no WebSocket protocol used).
Bypass UnitFoo for mobile. Possibly we do want the UnitFoo stuff after
all on mobile, to run in some special testing mode? Hard to say, let's
skipt it for now.
A problem comes up when only a part of the tile is visible
on the client side and invalidation affects the invisible
part of this tile. To get back the old / right behavior we
need to request this kind of tiles too.
gcc 8 warns: cast between incompatible function types from ‘void
(*)(int, siginfo_t*, void*)’ to ‘__sighandler_t’ {aka ‘void (*)(int’}
[-Werror=cast-function-type].
The struct sigaction already has an appropriately typed field
sa_sigaction in a union with the sa_handler field, so use that. (If
that is not present in some older Linux and/or glibc that we still
need to support, this commit will have to be reverted then.)
Change-Id: I67667073c89b7b22e7de1933ccaaa60868685866
Also gets rid of a potentially problematic strncpy() use that causes a
gcc warning: specified bound 256 equals destination size
[-Werror=stringop-truncation].
(Whether that would have caused a problem or not depends on how
LogPrefix would have been used, and whether it ever would have been
filled completely, without any terminating nul character, by that
strncpy().)
Change-Id: I92dba3726e3f46777d9b4c8cf88f557c02788fe0
I think the general policy should be to always log errno using both
Util::symbolicErrno() and std::strerror(), never log a naked errno.
But only in cases where we know that it is highly likely that it is
the most recent system call that has failed, so errno makes sense.
Change-Id: I4a1fb31e375ea949e7da17687464361efe7c1761
For a developer, it is much nicer to see "EXDEV" in a log than "18",
for instance. (Sure, we often also log strerror(), but might that
perhaps be localised? And the symbolic names are what one uses when
coding anyway.)
Change-Id: I456a8c2589147dcad87f1b4c3a20b3bd5a35d097
Better logging during wopi info parsing,
especially upon failures.
Refactored the code from Storage.cpp into
JsonUtil.hpp.
Minor optimizations.
Add unit-tests for the parsing logic.
Change-Id: Ifebc3f6b7030a6c7b3b399786633f6b5e8737478
Reviewed-on: https://gerrit.libreoffice.org/49927
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
And improve the logging support in unit-tests to
help troubleshoot issues faster and more accurately.
Also makes the code more readable (hopefully).
Change-Id: I4f8aafb5245e2f774b03231591a74544f9ec84aa
Reviewed-on: https://gerrit.libreoffice.org/48645
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
eventually the log file descriptor will be closed and unhandled exception it will throw it
"terminate called after throwing an instance of 'Poco::WriteFileException'
what(): Cannot write file"
Change-Id: I1d6ae3a4d4d4910f2ed2cdc80b162c27f93d55d9
Reviewed-on: https://gerrit.libreoffice.org/49055
Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
Tested-by: Michael Meeks <michael.meeks@collabora.com>
* logs helpful messages for various error corner-cases.
* optimized file descriptor closing for large fd counts.
Change-Id: I8cba9ecb3d71ddc6e22e20d89368d8c6b9b5097f
Now all logging is done after checking if the
level in question is enabled or not (thanks to
the macros LOG_XXX), which saves unnecessary
conversions and stringification when said level
is disabled.
Change-Id: Icde31e067f60269563896f04f8b0d65643720766
Reviewed-on: https://gerrit.libreoffice.org/47885
Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
Tested-by: Michael Meeks <michael.meeks@collabora.com>
Unlimited settings were logged as huge numbers.
In two cases settings were logged via LOG_SYS (and as errors)
instead of LOG_INF.
Change-Id: I1da493c0126ecf9d2382956ac1e60e57988696ee
Reviewed-on: https://gerrit.libreoffice.org/44731
Reviewed-by: Jan Holesovsky <kendy@collabora.com>
Tested-by: Jan Holesovsky <kendy@collabora.com>
To be able to set the support key directly from the command line, and to show
the option, etc.
Change-Id: Iac93bc47a6f4b9d5a5ad0ac8b06bda978e01b760
Reviewed-on: https://gerrit.libreoffice.org/43098
Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
Tested-by: Michael Meeks <michael.meeks@collabora.com>
The routine for handling the configuration for the max file size
limit, was wrongly using NOFILE. Now we handle both limits correctly.
Change-Id: Ie8b63617286f66af6d4eb1b35b9e4f4b28f3c2a6
Reviewed-on: https://gerrit.libreoffice.org/42803
Reviewed-by: Jan Holesovsky <kendy@collabora.com>
Tested-by: Jan Holesovsky <kendy@collabora.com>
Reviewed-on: https://gerrit.libreoffice.org/42811
Reviewed-by: Marco Cecchetti <mrcekets@gmail.com>
Tested-by: Marco Cecchetti <mrcekets@gmail.com>
Trying to combine the Poco's http server together with our polling loop leads
only to problem; so instead let's introduce a hook where we can do the WOPI
serving directly in the unit test.
Change-Id: Id3fec6ff93c3ad652aa4e0fc6309c5b7639728cb
Start killing documents when memory usage goes above threshold.
Also make it possible to close documents from admin instance.
In DocumentBroker::closeDocument, just set the _stop flag and wake
up the polling thread which will terminate the children, instead of
manually terminating the children.
Change-Id: Ie70e05b3fb6ea816a87b6dcfaed92cdddb94aa90
Changes protocol to use 'wid' instead of 'hash' everywhere. Wire-ids
are monotonically increasing integers that can be mapped to hash
values for all of the hash values and tiles we cache internally.
Change-Id: Ibcb25817bab0f453e93d52a6f99d3ff65059e47d
Seems to have no effect, so gone in
favor of RLIMIT_AS (virtual memeory).
Change-Id: I210879ec9285f420c9f9839cdabf45c42d865fb3
Reviewed-on: https://gerrit.libreoffice.org/38720
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
userextrainfo is a json array that contains
extra user-specific links.
Currently 'avatar' is assumed to hold the
image url for the user's avatar.
'mail' and other links can also be added.
Change-Id: I37c4c68bfa0b7ee659e017b4867dcb8cf5c2ca2f
Reviewed-on: https://gerrit.libreoffice.org/38120
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
The new password hash property is called secure_password in the config
file. `loolconfig` tool should be used to set the password hash in
appropriate format with desired salt length, password length, number of
iterations in PBKDF2.
To be backward compatible, plain-text password for admin-console in
config file is still accepted in case secure_password property is
missing from the config file.
Change-Id: If229999dac62856e368555c0242c4aa6f8061fba
Fixes the case when the client reconnects on idle
disconnection (because it never got the 'close: idle'
message).
Also, show informative message to users in this case
instead of grey screen.
Change-Id: Ia2e1f2ffefe6d35dd1552e7cc44e490aab86c600
Reviewed-on: https://gerrit.libreoffice.org/37891
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
Jail paths are now generate from a PRNG
instead of using the PID of the kit process.
The PRN is converted to base-64 and used
as the directory name where a given
kit is jailed.
Change-Id: I8e4bc35d9ccdfdae0e542ab707c417cd29ad52f3
Reviewed-on: https://gerrit.libreoffice.org/37372
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
Dung out overlapping return enumerations. Move more work into 'move'
callbacks at a safer time, etc.
Change-Id: I62ba5a35f12073b7b9c8de4674be9dae519a8aca
POST requests require the full request to be
left in the socket buffer to be parsed in full.
But GET requests, especially WS upgrade, must
have the request cleared from the socket, as
there is more data expected to be read after
the upgrade, which happens by the DocBroker
thread, so clearing the buffer must be done
before the upgrade.
This patch accomodates these two conflicting
cases and refactors the code slightly to
make it more structured and readable.
Change-Id: Ia7357a745a3900f986099ba14af2a0946023018b
Reviewed-on: https://gerrit.libreoffice.org/36873
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
This slows things down terribly, particularly the setting on the websocket
made tiles appearing one by one. Let's keep the possibility to zero the buffer
sizes for debugging, but hide that behind an env. variable (and in debug
builds only anyway).
Change-Id: Ie4d2cdb3c0ec3c50f1a2b4f9941a462ac4f2d196
This was a workaround to Poco's limitation
of requiring socket receiveFrame be given
preallocated buffer, which couldn't be
exceeded by a larger payload. This meant
the receiver had to know the maximum
payload in advance.
Since only the Kit uses Poco sockets,
and the Kit never receives large payloads,
this preamble is now obsolete.
100% (94/94) of old-style tests PASS.
Change-Id: I76776f89497409e5755e335a3e25553e91cf0876
Reviewed-on: https://gerrit.libreoffice.org/36037
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
With this every other client would be able to know about other client's
permission i.e whether they have opened the document as readonly. This
could be important eg: to hide the cursor overlay of readonly users in
the UI or to mark these users as readonly in the userlist.
Change-Id: I5dcb1b4e5a22c9b546d16b69b9216cc7653cff04
Pull the notification pieces out of SigUtil.cpp - not signal safe,
and invoked only from LOOLWSD anyway.
In a non-blocking world, the socket close notification sends are
instant - so more work required to count-down / timeout remaining
clients.
These are really GET requests that aren't
WebSocket upgrade. Should rename to something
less misleading.
Re-enabled testSlideShow which depended on this.
Change-Id: I52b7f67b650fcdcbae7c2bff020b756099263141
When the socket is closed the last WS frame
will not have any payload, just a frame.
In this case the socket should still fire
handleMessage so this frame could trigger
application logic, however in this case
ClientSession has nothing to do, so we skip it.
Change-Id: Ia2b13026e31460ffceb8f9d9cfa39d36fbc57146
There is no need to use `ps` here as reading
directly is trivial and has far less overhead.
Change-Id: I27d0432c1f3a9d35763d67fc445d8bd828f1b27e
Reviewed-on: https://gerrit.libreoffice.org/34052
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
We no longer tell the clinet "This is embarrassing..."
when we disconnect and unload an idle document. Instead,
the client UI remains greyed out so the user can resume
as if it was inactive (and reload the document in this case).
Also, we now always send the "close: " message prior
to shutting down a client websocket. This is more
reasonable and consistent when we intentionally disconnect,
so clients can rely on it to signal intent and give reason.
Otherwise, a disconnection without this application-level
message should be unexpected and is therefore reasonable
to show the "This is embarrassing..." message.
Change-Id: Ic7439bcc9267be155586ccd5d122e9fe60225516
A race condition between the client socket thread
and the idle-document cleanup caused segfault
on the websocket.
Now the ChildProcess object doesn't reset
the websocket on closing, rather on destruction.
Change-Id: I10d0dfb1ba677a65479df85b7a53de8c5f1b44c3
Each Kit process now reports its own PSS,
which is much more accurate as they share
a significant ratio of their pages with
one another.
Admin tracks the PSS values of the Kits
and reports to the console.
Change-Id: Ifa66d17749c224f0dc211db80c44f7c913f2d6c4
Reviewed-on: https://gerrit.libreoffice.org/33864
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
Around 1.5x faster than Poco,
which first enumerates files into
a container, then iterates over
them and stats before unlinking.
Here we enumerate and unlink in
a single pass.
Change-Id: I79d1c0f1b5ec557ccc4f0e2ec7a0609051d8d212
Reviewed-on: https://gerrit.libreoffice.org/33680
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
Now we don't get a situation where there would be a tremendous amount of
invalidates & tile render requests piled in the queue, so we can do it
deterministic again.
The only thing that could potentially pile in the queue are the keypresses
events sent from the clients, but that is a different problem anyway.
This reverts commit c326228774.
Change-Id: I98e199eab0187bf5f47ce322ac1b1b2e3b976b85
User input is batched together to reduce
overheads. This initial implementation
will batch all input of the same type
together.
Change-Id: Ia0069de9cf5acecf637941543267f86518c04640
Reviewed-on: https://gerrit.libreoffice.org/33422
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
Throw when empty payload is enqueued
and return empty payload on get timeout
(instead of throwing).
Change-Id: Iab5df775caa46d5c212d0850645cda6cca16f20b
Reviewed-on: https://gerrit.libreoffice.org/33421
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
Tested-by: Ashod Nakashian <ashnakash@gmail.com>
Otherwise we are getting completely confused times - various processes start
at various times, so for one process the epoch start can be eg. 20 minutes
later than for the other.
Change-Id: I6d87e98682a5fcd0348a584cf66f7ffa5813ca66
The server tells the client the hash of each tile it sends (calculated
from the contents of the tile, not its PNG encoding). When the client
asks for a tile to be refreshed, it tells the server what the hash of
the existing tile is. If the server notices that the tile contents
hasn't actually changed, it doesn't PNG encode it and doesn't send it
to the client.
The intent is that this will reduce load on the server and also avoid
unnecessary tile traffic.
Change-Id: Ia06ca68655ea984ed4319f24f4470afda322eccf
If we are logging a message, we want to see the first line of it in
its entirety if possible. Especially now with more parameters being
added to tile messages, 120 was not enough to see the added
interesting ones.
Bin the silly test that used knowledge of what the limit is. We should
not test a coindidental arbitrary number that is not a documented part
of an API. If we want to test the default abbreviation functionality,
we need to at least make that default limit (now 500) public in
Protocol.hpp.
Change-Id: Iea59ba46e8331e2a839c792146f123fed9df2b82
