Update u8g_ll_api.c

[kgy2002]

Replace u8g_Begin with u8g_UpdateDimension per comment to prevent 2 calls to u8g_Begin:

  /* On the Arduino Environment this will lead to two calls to u8g_Begin(), the following line will be called first (by U8glib constructors) */
  /* if - in future releases - this is removed, then still call u8g_UpdateDimension() */
  /* if Arduino call u8g_UpdateDimension else u8g_Begin */
  /* issue 146 */

This fixes pin 0 config when u8g_Begin is called twice.
May affect other inits but only tested this change on my board.

See Marlin Issue MarlinFirmware/Marlin#21277 and PR MarlinFirmware/Marlin#21417 for more info.

[thinkyhead]

Well isn't it nice that a comment is already there to point the way…!

[kgy2002]

Looks good to me. Everything still working here.

[rhapsodyv]

We are having similar issue with SMT32....

The issue is: Marlin keeps a global uglib instance, that makes it init in the __libc_init_array, before any system initialisation... and uglib try to do hardware things on init... this is messing up everything...

(gdb) bt
#0  WWDG_IRQHandler () at /Users/keith/.platformio/packages/framework-arduinoststm32/system/Drivers/CMSIS/Device/ST/STM32F1xx/Source/Templates/gcc/startup_stm32f103xe.s:114
#1  <signal handler called>
#2  pinMode (ulPin=ulPin@entry=23, ulMode=ulMode@entry=1) at /Users/keith/.platformio/packages/framework-arduinoststm32/cores/arduino/wiring_digital.c:31
#3  0x08013b10 in u8g_dev_rrd_st7920_128x64_fn (u8g=0x200009e4 <u8g+8>, dev=0x2000006c <u8g_dev_st7920_128x64_rrd_sw_spi>, msg=<optimized out>, arg=0x0) at Marlin/src/lcd/dogm/ultralcd_st7920_u8glib_rrd_AVR.cpp:123
#4  0x0802240a in u8g_InitLL (u8g=u8g@entry=0x200009e4 <u8g+8>, dev=0x2000006c <u8g_dev_st7920_128x64_rrd_sw_spi>) at .pio/libdeps/STM32F103RC_btt_stm32/U8glib-HAL/src/clib/u8g_ll_api.c:53
#5  0x080224fc in u8g_Begin (u8g=u8g@entry=0x200009e4 <u8g+8>) at .pio/libdeps/STM32F103RC_btt_stm32/U8glib-HAL/src/clib/u8g_ll_api.c:191
#6  0x08022544 in u8g_InitSPI (u8g=u8g@entry=0x200009e4 <u8g+8>, dev=<optimized out>, sck=<optimized out>, mosi=<optimized out>, cs=24 '\030', a0=255 '\377', reset=255 '\377') at .pio/libdeps/STM32F103RC_btt_stm32/U8glib-HAL/src/clib/u8g_ll_api.c:255
#7  0x08021f50 in U8GLIB::initSPI (this=this@entry=0x200009dc <u8g>, dev=<optimized out>, sck=sck@entry=25 '\031', mosi=mosi@entry=23 '\027', cs=<optimized out>, cs@entry=24 '\030', a0=<optimized out>, a0@entry=255 '\377', reset=<optimized out>, reset@entry=255 '\377')
    at .pio/libdeps/STM32F103RC_btt_stm32/U8glib-HAL/src/U8glib.cpp:45
#8  0x08012b5e in U8GLIB::U8GLIB (reset=255 '\377', a0=255 '\377', cs=24 '\030', mosi=23 '\027', sck=25 '\031', dev=<optimized out>, this=0x200009dc <u8g>) at .pio/libdeps/STM32F103RC_btt_stm32/U8glib-HAL/src/U8glib.h:75
#9  U8GLIB_ST7920_128X64_RRD::U8GLIB_ST7920_128X64_RRD (reset=255, cs=24, mosi=23, sck=25, this=0x200009dc <u8g>) at Marlin/src/lcd/dogm/HAL_LCD_class_defines.h:63
#10 __static_initialization_and_destruction_0 (__initialize_p=1, __priority=65535) at Marlin/src/lcd/dogm/marlinui_DOGM.cpp:79
#11 _GLOBAL__sub_I_u8g () at Marlin/src/lcd/dogm/marlinui_DOGM.cpp:472
#12 0x08022c30 in __libc_init_array ()
#13 0x08009266 in Reset_Handler () at /Users/keith/.platformio/packages/framework-arduinoststm32/system/Drivers/CMSIS/Device/ST/STM32F1xx/Source/Templates/gcc/startup_stm32f103xe.s:97
Backtrace stopped: previous frame identical to this frame (corrupt stack?)

The correct fix here is:

1. don't have a u8glib global var
2. make u8glib global var a pointer, and init it on the Marlin UI init
3. u8glib don't try init hardware on constructor, having a "init" method to be called on setup

[rhapsodyv]

I just send the fix I proposed to Marlin. It fixes the u8g lib issue we are having on STM32.

@kgy2002 can you test if my fix work for your case too?

MarlinFirmware/Marlin#21496

[kgy2002]

@rhapsodyv
I tested your latest (2nd) commit to the PR and it doesn't fix my issue. In the end you are still setting up the constructor which calls u8g->begin() once then u8g->firstPage() which calls u8g->begin() a second time. The second call is what's breaking the first call's I/O setting.

[rhapsodyv]

Ok, don't worry. We can have two issues and two fixes. :-)

We just need do one more check. Some Hal on marlin do some sort of hack to handle these multiple u8g init calls. They use a static local counter and only init the io when u8g lib is called more than once.

I don't know if this fix will affect this. But if it affect, we need to update the marlin code too.

https://github.com/MarlinFirmware/Marlin/blob/e3a12c3c28c2b6c4882eaa6549c6053b4d38953e/Marlin/src/lcd/dogm/u8g_dev_tft_upscale_from_128x64.cpp#L432 to line 437

This code was inherited by the old upscale dogm code.

I will test it with this changes. And seek for another places with the same hack.

[thinkyhead]

Please test Marlin as it stands now, since I went ahead and implemented delayed initialization directly in this library. If it is working well, then this can be closed.

[kgy2002]

Tested the current version of this library with bugfix-2.0.x pulled today. Everything works fine.