Updated to latest IDF

[MrSurly]

Not entirely sure CONFIG_ESP32_XTAL_FREQ_AUTO needs to be 1 or y

[vandys]

I'm pretty sure those are booleans from the config file, the "y" is for purposes of the config UI.
1 would be a better bet for both bools in your changes.

[MrSurly]

I'm pretty sure those are booleans from the config file, the "y" is for purposes of the config UI.
1 would be a better bet for both bools in your changes.

Done.

[dpgeorge]

Does this switch the code to use dual core? If so is dual core working ok now, did you test it?

[MrSurly]

Had to remove that to get the IDF to compile, since some (new) functions are hard-coded to use 2 cores. I did test it, and it's working fine. If you're interested in the exact area, I can try compiling with CONFIG_FREERTOS_UNICORE restored to 1.

Depends on what you mean by "test," hence my question in #88 WRT running tests.

I was just in "make it work" mode.

I did a make menuconfig on one of the IDF examples, and it un-defines this.

[dpgeorge]

some (new) functions are hard-coded to use 2 cores

Can you give examples of these? I'd be surprised if the IDF changed so much that it no longer works with single core.

I was just in "make it work" mode.

The esp32 port is still beta so as long as things work then it's ok. I assume you checked that wifi and your BT stuff works.

[MrSurly]

Dug into this:

Here's what you get if you restore CONFIG_FREERTOS_UNICORE:

/home/epoulsen/workspaces/esp-idf//components/esp32/dport_access.c: In function 'dport_access_init_core1':
/home/epoulsen/workspaces/esp-idf//components/esp32/dport_access.c:164:21: error: array subscript is above array bounds [-Werror=array-bounds]
     dport_access_ref[core_id] = 0;
                     ^
/home/epoulsen/workspaces/esp-idf//components/esp32/dport_access.c:165:23: error: array subscript is above array bounds [-Werror=array-bounds]
     dport_access_start[core_id] = 0;
                       ^
/home/epoulsen/workspaces/esp-idf//components/esp32/dport_access.c:166:21: error: array subscript is above array bounds [-Werror=array-bounds]
     dport_access_end[core_id] = 0;
                     ^
/home/epoulsen/workspaces/esp-idf//components/esp32/dport_access.c:167:21: error: array subscript is above array bounds [-Werror=array-bounds]
     dport_core_state[core_id] = DPORT_CORE_STATE_RUNNING;
                     ^
cc1: all warnings being treated as errors
make: *** [build//home/epoulsen/workspaces/esp-idf//components/esp32/dport_access.o] Error 1
make: *** Waiting for unfinished jobs....
epoulsen@banana ~/workspaces/micropython-esp32/esp32 $ 

Code in question:

Note that dport_access_init_core1 is not guarded by any #ifdef. Evidently the toolchain or the IDF code is set up to do static evaluation of assert() statements where possible.

esp-idf/components/esp32/dport_access.c:

static void dport_access_init_core1(void *arg)
{
    int core_id = xPortGetCoreID();

    assert(core_id == 1);

    ESP_INTR_DISABLE(ETS_DPORT_INUM);
    intr_matrix_set(core_id, ETS_FROM_CPU_INTR3_SOURCE, ETS_DPORT_INUM);
    ESP_INTR_ENABLE(ETS_DPORT_INUM);

    dport_access_ref[core_id] = 0;
    dport_access_start[core_id] = 0;
    dport_access_end[core_id] = 0;
    dport_core_state[core_id] = DPORT_CORE_STATE_RUNNING;

    vTaskDelete(NULL);
}

esp-idf/components/freertos/include/freertos/FreeRTOSConfig.h:

/* ESP31 and ESP32 are dualcore processors. */
#ifndef CONFIG_FREERTOS_UNICORE
#define portNUM_PROCESSORS 2
#else 
#define portNUM_PROCESSORS 1
#endif

If you removeesp-idf/components/esp32/dport_access.c from the Makefile, you get a whole bunch of link errors:

LINK build/application.elf
build//home/epoulsen/workspaces/esp-idf//components/freertos/xtensa_vectors.o:(.iram1.literal+0x54): undefined reference to `dport_access_start'
build//home/epoulsen/workspaces/esp-idf//components/freertos/xtensa_vectors.o:(.iram1.literal+0x58): undefined reference to `dport_access_end'
build//home/epoulsen/workspaces/esp-idf//components/newlib/time.o:(.literal.esp_setup_time_syscalls+0xc): undefined reference to `rtc_time_get'
build//home/epoulsen/workspaces/esp-idf//components/newlib/time.o: In function `esp_setup_time_syscalls':

... over 100 more lines ...

[dpgeorge]

Compiling the example get-started/hello_world with FreeRTOS/UNICORE set to "yes" succeeds. Try adding -Wno-error=array-bounds to CFLAGS for the dport_access.c file.

[MrSurly]

Try adding -Wno-error=array-bounds to CFLAGS for the dport_access.c file.

I'll try it, but that function is unconditionally a part of the IDF, however, I don't know if anything links to it.

The example I used was the gatt_server_service_table, though all I did was let it write the defaults for sdkconfig.

[MrSurly]

Try adding -Wno-error=array-bounds to CFLAGS for the dport_access.c file.

Yes, this worked, with CONFIG_FREERTOS_UNICORE restored to 1, basic tests passed.

Wifi works well, per the towel.blinkenlights.nl telnet test.

@dpgeorge : Should I update this PR with those changes?

[MrSurly]

The esp32 port is still beta so as long as things work then it's ok. I assume you checked that wifi and your BT stuff works.

The whole point was to update the IDF to resolve espressif/esp-idf#520, because I kept getting hammered with crashes while testing BT.

Let me test the WiFi thoroughly tomorrow (actually at home -- don't have any ESP32 here), but I was playing around with WiFi today, and it seemed OK.

I'll post results here tomorrow.

[MrSurly]

... and your BT stuff works.

Speaking of which, I have a new branch (BT implementation) that is rebased off of this branch (IDF update), so if this PR is merged, then I'll have to retract #86 (BT implementation on old IDF), and resubmit, since the differences are substantial enough that a simple rebase won't suffice.

[dpgeorge]

Thanks, this was squashed and merged in d7049f3. Follow up commits were made in e276161 and 48b5f4f