Before this patch MP_BINARY_OP_IN had two meanings: coming from bytecode it
meant that the args needed to be swapped, but coming from within the
runtime meant that the args were already in the correct order. This lead
to some confusion in the code and comments stating how args were reversed.
It also lead to 2 bugs: 1) containment for a subclass of a native type
didn't work; 2) the expression "{True} in True" would illegally succeed and
return True. In both of these cases it was because the args to
MP_BINARY_OP_IN ended up being reversed twice.
To fix these things this patch introduces MP_BINARY_OP_CONTAINS which
corresponds exactly to the __contains__ special method, and this is the
operator that built-in types should implement. MP_BINARY_OP_IN is now only
emitted by the compiler and is converted to MP_BINARY_OP_CONTAINS by
swapping the arguments.
It's possible to use the methods (eg ilistdir) of a VFS FatFS object
without it being mounted in the VFS itself. This previously worked but
only because FatFS was "mounting" the filesystem automatically when any
function (eg f_opendir) was called. But it didn't work for ports that used
synchronisation objects (_FS_REENTRANT) because they are only initialised
via a call to f_mount. So, call f_mount explicitly when creating a new
FatFS object so that everything is set up correctly. Then also provide a
finaliser to do the f_umount call, but only if synchronisation objects are
enabled (since otherwise the f_umount call does nothing).
This patch simplifies the str creation API to favour the common case of
creating a str object that is not forced to be interned. To force
interning of a new str the new mp_obj_new_str_via_qstr function is added,
and should only be used if warranted.
Apart from simplifying the mp_obj_new_str function (and making it have the
same signature as mp_obj_new_bytes), this patch also reduces code size by a
bit (-16 bytes for bare-arm and roughly -40 bytes on the bare-metal archs).
The SHA1 hashing functionality is provided via the "axtls" library's
implementation, and hence is unavailable when the "axtls" library is not being
used. This change provides the same SHA1 hashing functionality when using the
"mbedtls" library by using its implementation instead.
If SSL_EAGAIN is returned (which is a feature of MicroPython's axTLS fork),
return EAGAIN.
Original axTLS returns SSL_OK both when there's no data to return to user
yet and when the underlying stream returns EAGAIN. That's not distinctive
enough, for example, original module code works well for blocking stream,
but will infinite-loop for non-blocking socket with EAGAIN. But if we fix
non-blocking case, blocking calls to .read() will return few None's initially
(while axTLS progresses thru handshake).
Using SSL_EAGAIN allows to fix non-blocking case without regressing the
blocking one.
Note that this only handles case of non-blocking reads of application data.
Initial handshake and writes still don't support non-blocking mode and must
be done in the blocking way.
Per the comment found here
https://github.com/micropython/micropython-esp32/issues/209#issuecomment-339855157,
this patch adds finaliser code to prevent memory leaks from ussl objects,
which is especially useful when memory for a ussl context is allocated
outside the uPy heap. This patch is in-line with the finaliser code found
in many modsocket implementations for various ports.
This feature is configured via MICROPY_PY_USSL_FINALISER and is disabled by
default because there may be issues using it when the ussl state *is*
allocated on the uPy heap, rather than externally.
VLAs can be expensive on stack usage due to stack alignment requirements,
and also the fact that extra local variables are needed to track the
dynamic size of the stack. So using fixed-size arrays when possible can
help to reduce code size and stack usage.
In this particular case, the maximum value of n_args in the VLA is 2 and so
it's more efficient to just allocate this array with a fixed size. This
reduces code size by around 30 bytes on Thumb2 and Xtensa archs. It also
reduces total stack usage of the function: on Thumb2 the usage with VLA is
between 40 and 48 bytes, which is reduced to 32; on Xtensa, VLA usage is
between 64 and 80 bytes, reduced to 32; on x86-64 it's at least 88 bytes
reduced to 80.
Without this the board will crash when deactivating a stream that doesn't
have a close() method (eg UART) or that raises an exception within the
method (eg user-defined function).
The uos.dupterm() signature and behaviour is updated to reflect the latest
enhancements in the docs. It has minor backwards incompatibility in that
it no longer accepts zero arguments.
The dupterm_rx helper function is moved from esp8266 to extmod and
generalised to support multiple dupterm slots.
A port can specify multiple slots by defining the MICROPY_PY_OS_DUPTERM
config macro to an integer, being the number of slots it wants to have;
0 means to disable the dupterm feature altogether.
The unix and esp8266 ports are updated to work with the new interface and
are otherwise unchanged with respect to functionality.
Header files that are considered internal to the py core and should not
normally be included directly are:
py/nlr.h - internal nlr configuration and declarations
py/bc0.h - contains bytecode macro definitions
py/runtime0.h - contains basic runtime enums
Instead, the top-level header files to include are one of:
py/obj.h - includes runtime0.h and defines everything to use the
mp_obj_t type
py/runtime.h - includes mpstate.h and hence nlr.h, obj.h, runtime0.h,
and defines everything to use the general runtime support functions
Additional, specific headers (eg py/objlist.h) can be included if needed.
Allow literal minus in char classes to be in trailing position, e.g. [a-c-].
(Previously, minus was allowed only at the start.)
This increases ARM Thumb2 code size by 8 bytes.
This patch also removes the empty type "pinbase_type" (which crashes if
accessed) and uses "machine_pinbase_type" instead as the type of the
PinBase singleton.
With MBEDTLS_DEBUG_C disabled the function mbedtls_debug_set_threshold()
doesn't exist. There's also no need to call mbedtls_ssl_conf_dbg() so a
few bytes can be saved on disabling that and not needing the mbedtls_debug
callback.
The unary-op/binary-op enums are already defined, and there are no
arithmetic tricks used with these types, so it makes sense to use the
correct enum type for arguments that take these values. It also reduces
code size quite a bit for nan-boxing builds.
This implementation ignores invalid characters in the input. This allows
it to decode the output of b2a_base64, and also mimics the behavior of
CPython.
- Changed: ValueError, TypeError, NotImplementedError
- OSError invocations unchanged, because the corresponding utility
function takes ints, not strings like the long form invocation.
- OverflowError, IndexError and RuntimeError etc. not changed for now
until we decide whether to add new utility functions.
Since the stride is specified in pixels, in a 4-bit horizontal format it
has to always be even, otherwise the computation is wrong and we can
write outside of the buffer sometimes.
The code conventions suggest using header guards, but do not define how
those should look like and instead point to existing files. However, not
all existing files follow the same scheme, sometimes omitting header guards
altogether, sometimes using non-standard names, making it easy to
accidentally pick a "wrong" example.
This commit ensures that all header files of the MicroPython project (that
were not simply copied from somewhere else) follow the same pattern, that
was already present in the majority of files, especially in the py folder.
The rules are as follows.
Naming convention:
* start with the words MICROPY_INCLUDED
* contain the full path to the file
* replace special characters with _
In addition, there are no empty lines before #ifndef, between #ifndef and
one empty line before #endif. #endif is followed by a comment containing
the name of the guard macro.
py/grammar.h cannot use header guards by design, since it has to be
included multiple times in a single C file. Several other files also do not
need header guards as they are only used internally and guaranteed to be
included only once:
* MICROPY_MPHALPORT_H
* mpconfigboard.h
* mpconfigport.h
* mpthreadport.h
* pin_defs_*.h
* qstrdefs*.h
VfsFat no longer has the listdir() method. Rather, if listdir()
functionality is needed then one should use uos.listdir() which will call
VfsFat.ilistdir().
This patch allows mounting of VFS objects right at the root directory, eg
os.mount(vfs, '/'). It still allows VFS's to be mounted at a path within
the root, eg os.mount(vfs, '/flash'), and such mount points will override
any paths within a VFS that is mounted at the root.
Peer-closed socket is both readable and writable: read will return EOF,
write - error. Without this poll will hang on such socket.
Note that we don't return POLLHUP, based on argumentation in
http://www.greenend.org.uk/rjk/tech/poll.html that it should apply to
deeper disconnects, for example for networking, that would be link layer
disconnect (e.g. WiFi went down).
A shorter name takes less code size, less room in scripts and is faster to
type at the REPL.
Tests and HW-API examples are updated to reflect the change.
This implements the orginal idea is that Signal is a subclass of Pin, and
thus can accept all the same argument as Pin, and additionally, "inverted"
param. On the practical side, it allows to avoid many enclosed parenses for
a typical declararion, e.g. for Zephyr:
Signal(Pin(("GPIO_0", 1))).
Of course, passing a Pin to Signal constructor is still supported and is the
most generic form (e.g. Unix port will only support such form, as it doesn't
have "builtin" Pins), what's introduces here is just practical readability
optimization.
"value" kwarg is treated as applying to a Signal (i.e. accounts for possible
inversion).
MONO_xxx is much easier to read if you're not familiar with the code.
MVLSB is deprecated but kept for backwards compatibility, for the time
being.
This patch also updates the associated docs and tests.
This follows the pattern of how all other headers are now included, and
makes it explicit where the header file comes from. This patch also
removes -I options from Makefile's that specify the mp-readline/timeutils/
netutils directories, which are no longer needed.
Allows to get event time for a head item in the queue. The usecase
if waiting for the next event *OR* I/O completion. I/O completion may
happen before event triggers, and then wait should continue for the
remaining event time (or I/O completion may schedule another earlier
event altogether).
The new function has a strongly provisional status - it may be converted
to e.g. peek() function returning all of the event fields, not just time.
GIL behaviour should be handled by the port. And ports probably want to
define sleep_us so that it doesn't release the GIL, to improve timing
accuracy.
This patch refactors the error handling in the lexer, to simplify it (ie
reduce code size).
A long time ago, when the lexer/parser/compiler were first written, the
lexer and parser were designed so they didn't use exceptions (ie nlr) to
report errors but rather returned an error code. Over time that has
gradually changed, the parser in particular has more and more ways of
raising exceptions. Also, the lexer never really handled all errors without
raising, eg there were some memory errors which could raise an exception
(and in these rare cases one would get a fatal nlr-not-handled fault).
This patch accepts the fact that the lexer can raise exceptions in some
cases and allows it to raise exceptions to handle all its errors, which are
for the most part just out-of-memory errors during construction of the
lexer. This makes the lexer a bit simpler, and also the persistent code
stuff is simplified.
What this means for users of the lexer is that calls to it must be wrapped
in a nlr handler. But all uses of the lexer already have such an nlr
handler for the parser (and compiler) so that doesn't put any extra burden
on the callers.
For example, if the current directory is the root dir then this patch
allows one to do uos.listdir('mnt'), where 'mnt' is a valid mount point.
Previous to this patch such a thing would not work, on needed to do
uos.listdir('/mnt') instead.
By adding back monotonically increasing field in addition to time field.
As heapsort is not stable, without this, among entried added and readded
at the same time instant, some might be always selected, and some might
never be selected, leading to scheduling starvation.
Allows to iterate over the following without allocating on the heap:
- tuple
- list
- string, bytes
- bytearray, array
- dict (not dict.keys, dict.values, dict.items)
- set, frozenset
Allows to call the following without heap memory:
- all, any, min, max, sum
TODO: still need to allocate stack memory in bytecode for iter_buf.
If the mounted object doesn't have a "mount" method then assume it's a
block device and try to detect the filesystem. Since we currently only
support FAT filesystems, the behaviour is to just try and create a VfsFat
object automatically, using the given block device.
Each method asserts and deasserts signal respectively. They are equivalent
to .value(1) and .value(0) but conceptually simpler (and may help to avoid
confusion with inverted signals, where "asserted" state means logical 0
output).
SPI needs to be fast, and calling the EVENT_POLL_HOOK every byte makes it
unusable for ports that need to do non-trivial work in the EVENT_POLL_HOOK
call. And individual SPI transfers should be short enough in time that
EVENT_POLL_HOOK doesn't need to be called.
If something like this proves to be needed in practice then we will need
to introduce separate event hook macros, one for "slow" loops (eg
select/poll) and one for "fast" loops (eg software I2C, SPI).
machine.time_pulse_us() is intended to provide very fine timing, including
while working with signal bursts, where each transition is tracked in row.
Throwing and handling an exception may take too much time and "signal loss".
So instead, in case of a timeout, just return negative value. Cases of
timeout while waiting for initial signal stabilization, and during actual
timing, are recognized.
The documentation is updated accordingly, and rewritten somewhat to clarify
the function behavior.
A signal is like a pin, but ca also be inverted (active low). As such, it
abstracts properties of various physical devices, like LEDs, buttons,
relays, buzzers, etc. To instantiate a Signal:
pin = machine.Pin(...)
signal = machine.Signal(pin, inverted=True)
signal has the same .value() and __call__() methods as a pin.
This provides mp_vfs_XXX functions (eg mount, open, listdir) which are
agnostic to the underlying filesystem type, and just require an object with
the relevant filesystem-like methods (eg .mount, .open, .listidr) which can
then be mounted.
These mp_vfs_XXX functions would typically be used by a port to implement
the "uos" module, and mp_vfs_open would be the builtin open function.
This feature is controlled by MICROPY_VFS, disabled by default.
If MICROPY_VFS_FAT is enabled by a port then the port must switch to using
MICROPY_FATFS_OO. Otherwise a port can continue to use the FatFs code
without any changes.
import utimeq, utime
# Max queue size, the queue allocated statically on creation
q = utimeq.utimeq(10)
q.push(utime.ticks_ms(), data1, data2)
res = [0, 0, 0]
# Items in res are filled up with results
q.pop(res)
So long as a port defines relevant mp_hal_pin_xxx functions (and delay) it
can make use of this software SPI class without the need for additional
code.
These are basic drawing primitives. They work in a generic way on all
framebuf formats by calling the underlying setpixel or fill_rect C-level
primitives.
If you have longish operations on the db (such as logging data) it may
be desirable to periodically sync the database to the disk. The added
btree.sync() method merely exposes the berkley __bt_sync function to the
user.
The constants MP_IOCTL_POLL_xxx, which were stmhal-specific, are moved
from stmhal/pybioctl.h (now deleted) to py/stream.h. And they are renamed
to MP_STREAM_POLL_xxx to be consistent with other such constants.
All uses of these constants have been updated.
If the destination of os.rename() exists then it will be overwritten if it
is a file. This is the POSIX behaviour, which is also the CPython
behaviour, and so we follow suit.
See issue #2598 for discussion.
Fill is a very common operation (eg to clear the screen) and it is worth
optimising it, by providing a specialised fill_rect function for each
framebuffer format.
This patch improved the speed of fill by 10 times for a 16-bit display
with 160*128 pixels.
Rename FrameBuffer1 into FrameBuffer and make it handle different bit
depths via a method table that has getpixel and setpixel. Currently
supported formats are MVLSB (monochrome, vertical, LSB) and RGB565.
Also add blit() and fill_rect() methods.
If a port defines MICROPY_READER_POSIX or MICROPY_READER_FATFS then
lexer.c now provides an implementation of mp_lexer_new_from_file using
the mp_reader_new_file function.
Implementations of persistent-code reader are provided for POSIX systems
and systems using FatFS. Macros to use these are MICROPY_READER_POSIX and
MICROPY_READER_FATFS respectively. If an alternative implementation is
needed then a port can define the function mp_reader_new_file.
Its addition was due to an early exploration on how to add CPython-like
stream interface. It's clear that it's not needed and just takes up
bytes in all ports.
As required for further elaboration of uasyncio, like supporting baremetal
systems with wraparound timesources. This is not intended to be public
interface, and likely will be further refactored in the future.
Now the function properly uses ring arithmetic to return signed value
in range (inclusive):
[-MICROPY_PY_UTIME_TICKS_PERIOD/2, MICROPY_PY_UTIME_TICKS_PERIOD/2-1].
That means that function can properly process 2 time values away from
each other within MICROPY_PY_UTIME_TICKS_PERIOD/2 ticks, but away in
both directions. For example, if tick value 'a' predates tick value 'b',
ticks_diff(a, b) will return negative value, and positive value otherwise.
But at positive value of MICROPY_PY_UTIME_TICKS_PERIOD/2-1, the result
of the function will wrap around to negative -MICROPY_PY_UTIME_TICKS_PERIOD/2,
in other words, if a follows b in more than MICROPY_PY_UTIME_TICKS_PERIOD/2 - 1
ticks, the function will "consider" a to actually predate b.
Based on the earlier discussed RFC. Practice showed that the most natural
order for arguments corresponds to mathematical subtraction:
ticks_diff(x, y) <=> x - y
Also, practice showed that in real life, it's hard to order events by time
of occurance a priori, events tend to miss deadlines, etc. and the expected
order breaks. And then there's a need to detect such cases. And ticks_diff
can be used exactly for this purpose, if it returns a signed, instead of
unsigned, value. E.g. if x is scheduled time for event, and y is the current
time, then if ticks_diff(x, y) < 0 then event has missed a deadline (and e.g.
needs to executed ASAP or skipped). Returning in this case a large unsigned
number (like ticks_diff behaved previously) doesn't make sense, and such
"large unsigned number" can't be reliably detected per our definition of
ticks_* function (we don't expose to user level maximum value, it can be
anything, relatively small or relatively large).
In order to have more fine-grained control over how builtin functions are
constructed, the MP_DECLARE_CONST_FUN_OBJ macros are made more specific,
with suffix of _0, _1, _2, _3, _VAR, _VAR_BETEEN or _KW. These names now
match the MP_DEFINE_CONST_FUN_OBJ macros.
As long as a port implement mp_hal_sleep_ms(), mp_hal_ticks_ms(), etc.
functions, it can just use standard implementations of utime.sleel_ms(),
utime.ticks_ms(), etc. Python-level functions.
This refactors ujson.loads(s) to behave as ujson.load(StringIO(s)).
Increase in code size is: 366 bytes for unix x86-64, 180 bytes for
stmhal, 84 bytes for esp8266.
As per discussion in #2449, using write requests instead of read requests
for I2C.scan() seems to support a larger number of devices, especially
ones that are write-only. Even a read-only I2C device has to implement
writes in order to be able to receive the address of the register to read.
Adds check that LZ offsets fall into the sliding dictionary used. This
catches a case when uzlib.DecompIO with a smaller dictionary is used
to decompress data which was compressed with a larger dictionary.
Previously, this would lead to producing invalid data or crash, now
an exception will be thrown.
The delay_half parameter must be specified by the port to set up the
timing of the software SPI. This allows the port to adjust the timing
value to better suit its timing characteristics, as well as provide a
more accurate printing of the baudrate.
There is no need to take src_len and dest_len arguments. The case of
reading-only with a single output byte (originally src_len=1, dest_len>1)
is now handled by using the output buffer as the input buffer, and using
memset to fill the output byte into this buffer. This simplifies the
implementations of the spi_transfer protocol function.
The memory read/write I2C functions now take an optional keyword-only
parameter that specifies the number of bits in the memory address.
Only mem-addrs that are a multiple of 8-bits are supported (otherwise
the behaviour is undefined).
Due to the integer type used for the address, for values larger than 32
bits, only 32 bits of address will be sent, and the rest will be padded
with 0s. Right now no exception is raised when that happens. For values
smaller than 8, no address is sent. Also no exception then.
Tested with a VL6180 sensor, which has 16-bit register addresses.
Due to code refactoring, this patch reduces stmhal and esp8266 builds
by about 50 bytes.
When the clock is too fast for the i2c slave, it can temporarily hold
down the scl line to signal to the master that it needs to wait. The
master should check the scl line when it is releasing it after
transmitting data, and wait for it to be released.
This change has been tested with a logic analyzer and an i2c slace
implemented on an atmega328p using its twi peripheral, clocked at 8Mhz.
Without the change, the i2c communication works up to aboy 150kHz
frequency, and above that results in the slave stuck in an unresponsive
state. With this change, communication has been tested to work up to
400kHz.
Adds horizontal scrolling. Right now, I'm just leaving the margins
created by the scrolling as they were -- so they will repeat the
edge of the framebuf. This is fast, and the user can always fill
the margins themselves.
There was a bug in `framebuf1_fill` function, that makes it leave a few
lines unfilled at the bottom if the height is not divisible by 8.
A similar bug is fixed in the scroll method.
The idea is that all ports can use these helper methods and only need to
provide initialisation of the SPI bus, as well as a single transfer
function. The coding pattern follows the stream protocol and helper
methods.
This is an object-oriented approach, where uos is only a proxy for the
methods on the vfs object. Some internals had to be exposed (the STATIC
keyword removed) for this to work.
Fixes#2338.
In `btree_seq()`, when `__bt_seq()` gets called with invalid
`flags` argument it will return `RET_ERROR` and it won't
initialize `val`. If field `data` of uninitialized `val`
is passed to `mp_obj_new_bytes()` it causes a segfault.
This goes bit against websocket nature (message-based communication),
as it ignores boundaries bertween messages, but may be very practical
to do simple things with websockets.
In the sense that while GET_FILE transfers its data, REPL still works.
This is done by requiring client to send 1-byte block before WebREPL
server transfers next block of data.
Storing a chain of pbuf was an original design of @pfalcon's lwIP socket
module. The problem with storing just one, like modlwip does is that
"peer closed connection" notification is completely asynchronous and out of
band. So, there may be following sequence of actions:
1. pbuf #1 arrives, and stored in a socket.
2. pbuf #2 arrives, and rejected, which causes lwIP to put it into a
queue to re-deliver later.
3. "Peer closed connection" is signaled, and socket is set at such status.
4. pbuf #1 is processed.
5. There's no stored pbufs in teh socket, and socket status is "peer closed
connection", so EOF is returned to a client.
6. pbuf #2 gets redelivered.
Apparently, there's no easy workaround for this, except to queue all
incoming pbufs in a socket. This may lead to increased memory pressure,
as number of pending packets would be regulated only by TCP/IP flow
control, whereas with previous setup lwIP had a global overlook of number
packets waiting for redelivery and could regulate them centrally.
Allows to translate C-level pin API to Python-level pin API. In other
words, allows to implement a pin class and Python which will be usable
for efficient C-coded algorithms, like bitbanging SPI/I2C, time_pulse,
etc.
The time stamp is taken from the RTC for all newly generated
or changed files. RTC must be maintained separately.
The dummy time stamp of Jan 1, 2000 is set in vfs.stat() for the
root directory, avoiding invalid time values.
The call to stat() returns a 10 element tuple consistent to the os.stat()
call. At the moment, the only relevant information returned are file
type and file size.
Using usual method of virtual method tables. Single virtual method,
ioctl, is defined currently for all operations. This universal and
extensible vtable-based method is also defined as a default MPHAL
GPIO implementation, but a specific port may override it with its
own implementation (e.g. close-ended, but very efficient, e.g. avoiding
virtual method dispatch).
Make dupterm subsystem close a term stream object when EOF or error occurs.
There's no other party than dupterm itself in a better position to do this,
and this is required to properly reclaim stream resources, especially if
multiple dupterm sessions may be established (e.g. as networking
connections).
Both read and write operations support variants where either a) a single
call is made to the undelying stream implementation and returned buffer
length may be less than requested, or b) calls are repeated until requested
amount of data is collected, shorter amount is returned only in case of
EOF or error.
These operations are available from the level of C support functions to be
used by other C modules to implementations of Python methods to be used in
user-facing objects.
The rationale of these changes is to allow to write concise and robust
code to work with *blocking* streams of types prone to short reads, like
serial interfaces and sockets. Particular object types may select "exact"
vs "once" types of methods depending on their needs. E.g., for sockets,
revc() and send() methods continue to be "once", while read() and write()
thus converted to "exactly" versions.
These changes don't affect non-blocking handling, e.g. trying "exact"
method on the non-blocking socket will return as much data as available
without blocking. No data available is continued to be signaled as None
return value to read() and write().
From the point of view of CPython compatibility, this model is a cross
between its io.RawIOBase and io.BufferedIOBase abstract classes. For
blocking streams, it works as io.BufferedIOBase model (guaranteeing
lack of short reads/writes), while for non-blocking - as io.RawIOBase,
returning None in case of lack of data (instead of raising expensive
exception, as required by io.BufferedIOBase). Such a cross-behavior
should be optimal for MicroPython needs.
Calling it from lwIP accept callback will lead incorrect functioning
and/or packet leaks if Python callback has any networking calls, due
to lwIP non-reentrancy. So, instead schedule "poll" callback to do
that, which will be called by lwIP when it does not perform networking
activities. "Poll" callback is called infrequently though (docs say
every 0.5s by default), so for better performance, lwIP needs to be
patched to call poll callback soon after accept callback, but when
current packet is already processed.
While just a websocket is enough for handling terminal part of WebREPL,
handling file transfer operations requires demultiplexing and acting
upon, which is encapsulated in _webrepl class provided by this module,
which wraps a websocket object.
Damien George
Christopher Cooper
Paul Sokolovsky
Eric Poulsen
Alex Robbins
Javier Candeira
Alexander Steffen
Radomir Dopieralski
Tom Collins
Ville Skyttä
Peter Hinch
Jan Pochyla
Andrew Gatt
Oleg Korsak
w4kpm
Alex March
Pavol Rusnak
Stefan Agner
Krzysztof Blazewicz
Robert HH