... that this behaviour is simply not specified in the "bible" (whichever version - at the end of the day, both the Philips/NXP and Atmel version are verbatim, most probably OCR-ed, copies of the Intel original). There is no mention of whether the said flags can be written by software at all!

So, IMHO, it is legitimate to ask this question, and I know of no source of ultimate answer. And while the absolutely correct answer is "it is not specified", as some of you pointed out, there are "hints" of what may happen. But, you can see those "hints" there only because you are experienced enough to know that this is the usual way of these things! How could a newcomer know that, if it is NOT written down anywhere in the relevant literature? (And I don't care if it was a homework or not).

And, no Stefan, I did not try... at least I don't remember. I've surely triggered interrupts in this way several times, but I don't remember if it was specifically the external interrupt in '51. A couple of years ago I would sit down and construct a test - a series of tests, checking not only what happens if I write the flag in both settings of the level/edge flag, but also what happens if I write to the flag while interrupt is set to level and after that change the setting of the level/edge flag. And I would try it on half-a-dozen of '51 derivatives (remember my PUSH SP thread?). But I am too tired now, sorry.

Jan