The STATIC macro was introduced a very long time ago in commit
d5df6cd44a. The original reason for this was
to have the option to define it to nothing so that all static functions
become global functions and therefore visible to certain debug tools, so
one could do function size comparison and other things.
This STATIC feature is rarely (if ever) used. And with the use of LTO and
heavy inline optimisation, analysing the size of individual functions when
they are not static is not a good representation of the size of code when
fully optimised.
So the macro does not have much use and it's simpler to just remove it.
Then you know exactly what it's doing. For example, newcomers don't have
to learn what the STATIC macro is and why it exists. Reading the code is
also less "loud" with a lowercase static.
One other minor point in favour of removing it, is that it stops bugs with
`STATIC inline`, which should always be `static inline`.
Methodology for this commit was:
1) git ls-files | egrep '\.[ch]$' | \
xargs sed -Ei "s/(^| )STATIC($| )/\1static\2/"
2) Do some manual cleanup in the diff by searching for the word STATIC in
comments and changing those back.
3) "git-grep STATIC docs/", manually fixed those cases.
4) "rg -t python STATIC", manually fixed codegen lines that used STATIC.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
- Completes a longstanding TODO in the code, to not ignore
the optional family, type, proto and flags arguments to
socket.getaddrinfo().
- Note that passing family=socket.AF_INET6 will now cause queries
to fail (OSError -202). Previously this argument was ignored so
IPV4 results were returned instead.
- Optional 'type' argument is now always copied into the result. If not
set, results have type SOCK_STREAM.
- Fixes inconsistency where previously querying mDNS local suffix (.local)
hostnames returned results with socket type 0 (invalid), but all other
queries returned results with socket type SOCK_STREAM (regardless of
'type' argument).
- Optional proto argument is now returned in the result tuple, if supplied.
- Optional flags argument is now passed through to lwIP. lwIP has handling
for AI_NUMERICHOST, AI_V4MAPPED, AI_PASSIVE (untested, constants for
these are not currently exposed in the esp32 socket module).
- Also fixes a possible memory leak in an obscure code path
(lwip_getaddrinfo apparently sometimes returns a result structure with
address "0.0.0.0" instead of failing, and this structure would have been
leaked.)
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
If the hard socket limit (default 16) is reached then it's possible that
socket allocation fails but garbage collection would allow it to succeed.
Perform a GC pass and try again before giving up, similar to the logic
elsewhere in MicroPython that tries a GC pass before raising MemoryError.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
LWIP doesn't implement a timeout for blocking connect(), and such a timeout
is not required by POSIX. However, CPython will use the socket timeout for
blocking connect on most platforms. The "principle of least surprise"
suggests we should support it on ESP32 as well (not to mention it's
useful!).
This provides the additional improvement that external exceptions (like
KeyboardInterrupt) are now handled immediately if they happen during
connect(). Previously Ctrl-C would not terminate a blocking connect until
connect() returned, but now it will.
Fixes issue #8326.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
This implements support for SO_BINDTODEVICE, which allows telling a socket
to use a specific interface instead of lwIP automatically selecting one.
This allows devices that have multiple connections (for example cellular
over PPP in addition to WLAN) to explicitly choose which data is send over
which connection, which may have different reliability and or (mobile data)
costs associated with using them.
The used lwIP network stack already has support for this, so all that was
needed was to expose this functionality in MicroPython. This commit
exposes a new constant SO_BINDTODEVICE which can be set as an socket
option. As a value it expects the name of the interface to bind to. These
names can be retrieved using `.config('ifname')` implemented on each
interface type (including adding in this commit a `.config()` method to
PPP, which it didn't have before), which returns a string with the
interface name:
>>> import machine
>>> import network
>>> network.WLAN(network.AP_IF).config('ifname')
'lo0'
>>> wlan = network.WLAN(network.AP_IF)
>>> wlan.active(True) and wlan.config('ifname')
'ap1'
>>> wlan = network.WLAN(network.STA_IF)
>>> wlan.active(True) and wlan.config('ifname')
'st1'
>>> ppp = network.PPP(machine.UART(0))
>>> ppp.active(True) and ppp.config('ifname')
'pp1'
>>> ppp = network.PPP(machine.UART(0))
>>> ppp.active(True) and ppp.config('ifname')
'pp2'
>>> ppp = network.PPP(machine.UART(0))
>>> ppp.active(True) and ppp.config('ifname')
'pp3'
Note that lo0 seems to be returned by lwIP if the interface is not yet
active. The method can also return None in the case of PPP where the
entire lwIP interface doesn't yet exist before being activated. Currently
no effort is made to unify those cases; it is expected that whatever we
receive from lwIP is valid.
When the socket option is set, this forces using a specific device:
import socket
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.setsockopt(socket.SOL_SOCKET, socket.SO_BINDTODEVICE, 'st1')
setsockopt will throw (OSError: [Errno 19] ENODEV) if the specified
interface does not exist.
Tested with LAN, WLAN, and PPP; can specify which interface should be used
and when testing with, for example, HTTP requests to ifconfig.co the
returned IP address confirms a specific interface was used.
Signed-off-by: Daniël van de Giessen <daniel@dvdgiessen.nl>
This commit updates the esp32 port to work exclusively with ESP-IDF v5.
IDF v5 is needed for some of the newer ESP32 SoCs to work, and it also
cleans up a lot of the inconsistencies between existing SoCs (eg S2, S3,
and C3).
Support for IDF v4 is dropped because it's a lot of effort to maintain both
versions at the same time.
The following components have been verified to work on the various SoCs:
ESP32 ESP32-S2 ESP32-S3 ESP32-C3
build pass pass pass pass
SPIRAM pass pass pass N/A
REPL (UART) pass pass pass pass
REPL (USB) N/A pass pass N/A
filesystem pass pass pass pass
GPIO pass pass pass pass
SPI pass pass pass pass
I2C pass pass pass pass
PWM pass pass pass pass
ADC pass pass pass pass
WiFi STA pass pass pass pass
WiFi AP pass pass pass pass
BLE pass N/A pass pass
ETH pass -- -- --
PPP pass pass pass --
sockets pass pass pass pass
SSL pass ENOMEM pass pass
RMT pass pass pass pass
NeoPixel pass pass pass pass
I2S pass pass pass N/A
ESPNow pass pass pass pass
ULP-FSM pass pass pass N/A
SDCard pass N/A N/A pass
WDT pass pass pass pass
Signed-off-by: Damien George <damien@micropython.org>
Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
This renames the builtin-modules, such that help('modules') and printing
the module object will show "module" rather than "umodule".
This work was funded through GitHub Sponsors.
Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
Instead of being an explicit field, it's now a slot like all the other
methods.
This is a marginal code size improvement because most types have a make_new
(100/138 on PYBV11), however it improves consistency in how types are
declared, removing the special case for make_new.
Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
Previously the desired output type was specified. Now make the type part
of the function name. Because this function is used in a few places this
saves code size due to smaller call-site.
This makes `mp_obj_new_str_type_from_vstr` a private function of objstr.c
(which is almost the only place where the output type isn't a compile-time
constant).
This saves ~140 bytes on PYBV11.
Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
It's no longer needed because this macro is now processed after
preprocessing the source code via cpp (in the qstr extraction stage), which
means unused MP_REGISTER_MODULE's are filtered out by the preprocessor.
Signed-off-by: Damien George <damien@micropython.org>
This follows the CPython change: https://bugs.python.org/issue21455
Socket listen backlog defaults to 2 if not given, based on most bare metal
targets not having many resources for a large backlog. On UNIX it defaults
to SOMAXCONN or 128, whichever is less.
For an unconnected TCP socket, poll should return WR|HUP and read should
raise ENOTCONN. This is implemented by this commit and now the following
tests pass on esp32: extmod/usocket_tcp_basic.py,
net_hosted/connect_poll.py.
Signed-off-by: Damien George <damien@micropython.org>
This commit fixes the behaviour of socket.getaddrinfo on the ESP32 so it
raises an OSError when the name resolution fails instead of returning a []
or a resolution for 0.0.0.0.
Tests are added (generic and ESP32-specific) to verify behaviour consistent
with CPython, modulo the different types of exceptions per MicroPython
documentation.
Previous behaviour is when this argument is set to "true", in which case
the function will raise any pending exception. Setting it to "false" will
cancel any pending exception.
This commit adds support for a second supported hash (currently set to the
4.0-beta1 tag). When this hash is detected, the relevant changes are
applied.
This allows to start using v4 features (e.g. BLE with Nimble), and also
start doing testing, while still supporting the original, stable, v3.3 IDF.
Note: this feature is experimental, not well tested, and network.LAN and
network.PPP are currently unsupported.
They are both enabled by default, but can be disabled by defining
MICROPY_HW_ENABLE_MDNS_QUERIES and/or MICROPY_HW_ENABLE_MDNS_RESPONDER to
0. The hostname for the responder is currently taken from
tcpip_adapter_get_hostname() but should eventually be configurable.
If there are many short reads to a socket in a row (eg by readline) then
releasing and acquiring the GIL each time will give very poor throughput.
So first poll the socket to see if it has data, and if it does then don't
release the GIL.
If a socket is cleanly shut down by the peer then reads on this socket
should continue to return zero bytes. The lwIP socket API does not have
this behaviour (it only returns zero once, then blocks on subsequent calls)
so this patch adds explicit checks and logic for peer closed sockets.
The esp8266 uses modlwip.c for its usocket implementation, which allows to
easily support callbacks on socket events (like when a socket becomes ready
for reading). This is not as easy to do for the esp32 which uses the
ESP-IDF-provided lwIP POSIX socket API. Socket events are needed to get
WebREPL working, and this patch provides a way for such events to work by
explicitly polling registered sockets for readability, and then calling the
associated callback if the socket is readable.
This patch moves the implementation of stream closure from a dedicated
method to the ioctl of the stream protocol, for each type that implements
closing. The benefits of this are:
1. Rounds out the stream ioctl function, which already includes flush,
seek and poll (among other things).
2. Makes calling mp_stream_close() on an object slightly more efficient
because it now no longer needs to lookup the close method and call it,
rather it just delegates straight to the ioctl function (if it exists).
3. Reduces code size and allows future types that implement the stream
protocol to be smaller because they don't need a dedicated close method.
Code size reduction is around 200 bytes smaller for x86 archs and around
30 bytes smaller for the bare-metal archs.
Currently only the first 2 args are used, but this patch should at least
make getaddrinfo() signature-compatible with CPython and other bare-metal
ports that use the lwip bindings.
Angus Gratton