Is this bug in AVR porting?

hi
i wrote this program for test:
in port.c:
    void SIG_OUTPUT_COMPARE1A( void ) __attribute__ ( ( signal, naked ) );
    void SIG_OUTPUT_COMPARE1A( void )
    {
        PORTA++;vPortYieldFromTick();PORTA--;
        asm volatile ( "reti" );
    }
=================================================================
/* Priority definitions for most of the tasks in the demo application.  Some
tasks just use the idle priority. */
#define mainLED_TASK_PRIORITY           ( tskIDLE_PRIORITY )
void LED1( void *pvParameters )
{
    DDRB = 0xff;
    while(1)
    {
        PORTB |= (1<<0);
        _delay_us(1);
        PORTB &= ~(1<<0);
        _delay_us(1);
    }
}
void LED2( void *pvParameters )
{
    DDRC = 0xff;
    while(1)
    {
        PORTC |= (1<<1);
        _delay_us(1);
        PORTC &= ~(1<<1);
        _delay_us(1);
    }
}
portSHORT main(void)
{
    DDRA=0xff;
    PORTD++;
    xTaskCreate( LED1, ( signed char * ) "LED1", 100, NULL, 0, NULL );
    xTaskCreate( LED2, ( signed char * ) "LED2", 100, NULL, 0, NULL );

    vTaskStartScheduler();
    return 0;
}
note that i used PORTA to see timer interrupt flow.
after simulating i got this result on output:
you can see task1 (LED1) runs until first timer interrupt , and after first call of timer interrupt task switched to task2 (LED2) and it runs to ever. i worked on it many hours and finally i found a way.
if you change timer interrupt in ‘port.c’ file to this :
    void SIG_OUTPUT_COMPARE1A( void ) __attribute__ ( ( signal, naked ) );
    void SIG_OUTPUT_COMPARE1A( void )
    {
        asm volatile ( "sei" );
        PORTA++;vPortYieldFromTick();PORTA--;
        asm volatile ( "reti" );
    }
you will see this result :

you can see that the program loops in interrupt handler until next interrupt tick and there are an overlap on firing interrupt.
but both task are working and we have auto switch between tasks.

Is this bug in AVR porting?

I don’t know what you are trying to do, what _delay_us() does (does it call a FreeRTOS function?), or what the “sei” instruction does, but the first thing I notice here is your calling PORTA++; from a naked function. That is probably not a good idea because the naked function will not have a stack frame of any kind, and it will change the MCU context before it has been saved. I would have to check the asm code generated by the compiler to be completely sure, but suspect it will break the scheduler.

Is this bug in AVR porting?

thanks for your reply
i made my board and get real test, it works fine and the problem is in my simulator (proteus)