We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.
Hello,
I am writing an I2C master transmitter object. The problem I am having is that the IRQ Handler for I2C1 is not firing. According to the manual, it should fire after the start condition is sent and the status register is 0x08. When I step through the code with the debugger, the status code is 0x08. I am using an MCB1700 development board. Why is it not firing? Did I not configure something properly? For tests, All I am doing is toggling an LED on when the interrupt fires. Once I solve this problem, I will write the rest of the code.
#include <stdio.h> #include <LPC17XX.h> #include "GPIO1Output.h" GPIO1Output out(P1_28); __irq void I2C1_IRQHandler(void) { if (LPC_I2C1->I2STAT == 0x08) { out.Set(); } } int main(void) { // Initialize I2C. LPC_PINCON->PINSEL0 = 0x0F; LPC_PINCON->PINMODE_OD0 = 0x03; NVIC_EnableIRQ(I2C1_IRQn); LPC_I2C1->I2CONSET = 0x40; LPC_I2C1->I2CONSET = 0x20; int status = LPC_I2C1->I2STAT; }
There is a reason why an embedded developer should have at least some knowledge about assembler for the target architecture. The way to tell is to locate the interrupt vectors in the startup file, and then follow the path until you either get to your ISR, or somewhere you do not want to reach - like your infinite branch.
The datasheet for your processor - actually "User Manual" is the name NXP users - does a good job of telling you where the interrupt vectors are.
Another thing you really have to know about is the difference between single-stepping C source code and single-stepping processor instructions. It is possible to single-step in the startup file too. It's your single-stepping source lines (which indirectly means that the compiler sets a breakpoint on the next source line) that have had you not notice the infinite loop until now. When suspecting interrupt problems, it's often needed to perform instruction-level single-stepping.
I use the startup file provided with Keil. Do they want me to set up the routines manually?
I just don't understand. In the example projects the interrupts work, and the startup and configuration files are no different (I looked at examples using the most recent files too). Im stuck and I seriously doubt I am doing anything wrong as I cannot figure out from the examples.
Thanks, Ben
Note that the NVIC function does not name the ISR. The mapping to ISR is just by having the correct name of the ISR in relation to the startup file. And any ISR not supplied by you have a default ISR supplied during linking.
So what did you find out when you single-stepped through the startup file? When your ISR was not called - what other function was called? Was it just a question of Keil having changed the default name for the ISR function?
Stepping through the startup file only just initializes the reset handler and the stack heap. Then it jumps into the infinate loop that is supposed to be the interrupt (the B instruction).
It appears as if the default ISR was called.
OK I figured it out. I am using C++ and this forum posting delivered the answer I was looking for. It talks about how the overloading of C++ functions could cause weak references to the intrrupt handlers.
electronics.stackexchange.com/.../bug-in-keil-arm-compiler-with-interrupt-handlers-and-c
All I have to do is put a wrapper around the interrupt functions and that fixes it.
extern "C" { ... Interrupt Functions. }
Is there a better way of doing this or do you always need the wrapper when using C++?
No, C++ does not cause weak references to interrupt handlers.
But C++ uses name mangling, which means the external name the linker sees contains extra characters encoding the returned data type, and the parameter list.
So a C++ file with a function void hello(void) will not let the linker see a function "hello". You don't need to use the braces to wrap. Standard C++ documentation tells you that you can specifiy directly for the function that it should use the C calling convention and have C bindings.
If I Recall/Remember Correctly, For Cortex-M3, or I should say NXP LPC17xx, There are two kind of Keil examples, 1. Older, not CMSIS style; it has two important source files; startup_LPC17xx.s, system_LPC17xx.c 2. Newer, CMSIS style; it has three important source files; startup_LPC17xx.s, system_LPC17xx.c, core_cm3.c Newer Keil MDK provides the CMSIS style source, startup_LPC17xx.s, system_LPC17xx.c, core_cm3.c. If you mix-up the [not CMSIS style source] with [CMSIS style source], you will have some problems to fix.
In your Difference Report->system_LPC17xx_c.htm
Left:
#define PLL0CFG_Val 0x0000000B
Right:
#define PLL0CFG_Val 0x00050063
As per your Difference Report, I believe that If you mix-up different versions of the [CMSIS style source], you will have some problems to fix.
There are also examples from NXP
I don't know, but I would not expect arbitrary files taken from NXP examples to "just work" with Keil examples - or vice-versa.