Sound like the order is reversed :-)
Usually,
putchar() is implemented through a ring buffer connected to the UART ISR
As of printf(), the library implements it over putchar().

anyway, it's fine if it works actually well.

But when you are handling the UART in interrupt, why "TI0=1;" is required at the initialization stage on the main()?
It evokes the UART interrupt uselessly.
And also, it triggers the trouble as you said on your first post.

In usual implementation, just when the UART pauses TX session, putchar() kicks the UART by
SBUF0 = first_byte_on_ring_buffer;
OR
TI0=1;

Is the ring buffer pointers (indices) initialized properly before enabling interrupt (EA = 1;) ?
Where is this initialization?


As you are using "#pragma LARGE", and you don't use any memory specifier, all variables go to XRAM space, including function parameters and internal working of the functions. Then, the IRAM space is almost empty. I don't think stack overflow occurs at just moment after the initialization under this condition.

Tsuneo