... still unfamiliar with the thingy I am reluctant to put it in a bug report.

the startup code goues out to lunch (see below) if _sdcc_external_startup is left out or return zero. If _sdcc_external_startup returns nonzero, all is OK.

manually copying the disassembly from the emulator window:
fairly normal looking code till here
djnz ; the end of the xdata clearing
djnz
mov 29h,#80h
mov 2ah,#81h
mov 2bh,#82h
... adding 1 to both left and right
mov 50h,#f3h
mov 51h,#f4h !!!!!!!!! observe the anomaly (not +1, +1
mov 52h,#f8h
mov 53h,#ffh
setb 2a.4h
setb 2a.5h
setb 2a.6h
.... next bit
setb 2f.3h
mov 7ch,#80h
mov 7dh,#81h
... adding 1 to both left and right
mov 87h,#8eh !!!!!!!!!!!! sets the power down bit

as stated above, due to unfamiliarity I am reluctant to put it in a bug report, if someone 'fluent' in SDCC confirms this, (s) is welcome to add a bug report.

Is it so that not having _sdcc_external_startup is so unusual that this simply has not been detected?

Instead of all this jumping through hoops for the sake of universality, I guess I will just "SDCCize" a regular (Keil) startup code and include that. Since I disallow initialized (to nonzero) global variables, I should be fine.

Erik