tools/gen-device-svd: Increment lsb/msb last in bitfield ranges.

[bemasher]

Fixes #5712 in gen-device-svd.

Registers with bitfield ranges were incorrectly generated. This corrects _Pos, _Msk constants, and accessors in the following device families and peripherals:

Family	Peripheral
ESP32 (H2, S2, S3, C3, C6)	RMT
Kendryte K210	APU, DMAC, GPIO, GPIOHS
STM32C0	ADC, DMA, DMAMUX, GPIO, TIM
STM32F0	ADC, CAN, CAN_RX, CAN_TX, DAC, DMA, GPIO, TIM, TSC
STM32F1	ADC, CAN, CAN_RX, CAN_TX, DAC, DMA, EXTI, GPIO, TIM
STM32F2	ADC, CAN, CAN_RX, CAN_TX, DAC, DCMI, EXTI, FLASH, GPIO, RTC, TIM
STM32F3	ADC, CAN, CAN_RX, CAN_TX, DAC, DMA, GPIO, HRTIM, RTC, TIM, TSC
STM32F4	ADC, CAN, CAN_RX, CAN_TX, DAC, DCMI, DFSDM, DFSDM_FLT, DSIHOST, EXTI, FLASH, GPIO, RTC, TIM
STM32F7	ADC, CAN, CAN_RX, CAN_TX, DAC, DCMI, DFSDM, DFSDM_FLT, DSI, EXTI, FLASH, GPIO, RTC, TIM
STM32G0	ADC, DAC, DMA, DMAMUX, GPIO, RTC, TIM
STM32G4	ADC, DAC, DMA, DMAMUX, FMC, GPIO, HRTIM, RTC, SAI, TIM
STM32H5	ADC, DAC, DCMI, FMC, GPDMA, GPIO, GTZC, RTC, TIM
STM32H7	ADC, BDMA, DAC, DCMI, DFSDM, DFSDM_FLT, DMAMUX, DSIHOST, GPDMA, GPIO, HPDMA, HRTIM, HSEM, MDMA, RTC, SAI, TIM
STM32L0	ADC, DAC, DMA, GPIO, RTC, TIM, TSC
STM32L1	ADC, DAC, DMA, EXTI, GPIO, RTC, TIM
STM32L4	ADC, CAN, CAN_RX, CAN_TX, DAC, DCMI, DFSDM, DFSDM_FLT, DMA, DMAMUX, DSI, FMC, GPIO, RTC, TIM, TSC
STM32L5	ADC, DAC, DFSDM, DFSDM_FLT, DMA, DMAMUX, FMC, GPIO, RTC, SAI, TIM, TSC
STM32MP	DSIHOST1, GPIOA, GPIOB, GPIOC, GPIOD, GPIOE, GPIOF, GPIOG, GPIOH, GPIOI, GPIOJ, GPIOK, RTC
STM32N6	BSEC, GPIO
STM32U0	DMA, DMAMUX, GPIO, RTC, TIM, TSC
STM32U5	ADC, DAC, DCMI, FMC, GPDMA, GPIO, LPDMA, LPGPIO, RTC, TIM, TSC
STM32WB	ADC, DMA, DMAMUX, GPDMA, GPIO, HSEM, RTC, SAI, TIM, TSC
STM32WL	ADC, DMA, DMAMUX, GPIO, HSEM, RTC, TIM

[davecheney]

@bemasher TYVM, please consider pulling this test bemasher#1

[davecheney]

LGTM. thank you. Someone else must approve

[deadprogram]

Thanks for tracking this down @bemasher and thanks @davecheney for the test. I regenerated the device files before and after this change and took a good look. The following is edited from an automated review:

1. rangeElems already passes the per element offset as the second callback argument, and this is its only caller. Perhaps use it instead of mutating lsb/msb? This gives the same output for stm32 and Espressif and drops the da != nil check.

da.rangeElems(func(ia int, offset uint32) bool {
	lsb, msb := lsb+offset, msb+offset

2. The newer Espressif SVDs also change (esp32c3, esp32c6, esp32h2, esp32s2, esp32s3), all in the RMT interrupt fields. No machine code uses those. Perhaps mention them in the description?

3. This also fixes some machine code that used the wrong constants. In machine_stm32_moder_gpio.go the I2C pin setup used GPIO_OTYPER_OT0_Msk (was 0x2), so it cleared the open drain bit of the next pin. The F1/F4 ADC sample time code used ADC_SMPR2_SMP1_Pos (was 6, now 3). The local workarounds in machine_stm32_tim.go and the ADC_SQR1_SQ1_Pos note in machine_stm32_adc_u5.go can be removed in a follow up.

How does that sound?

[bemasher]

Thanks for the suggestions.

I've updated the original description of the pull request to describe the families and peripherals affected from parsing git's diff for the generated device files and used the offset parameter to set lsb and msb in each call.

I'll review machine source for affected accessor and constant use to see if any need fixes in a separate pull request.