Martin, thanks. I've found the DCVAL function. Modified my code & regression tested on z390. Works ok. I've asked John Kalinich to run the same test on HLASM. I hope we'll fare better now.
Thanks for your suggestion. Kind regards, Abe Kornelis. =========
I did run into the same problem with K years ago. To circumvent it there is a function that gives correct result. My guess is that (when ASM-team discovered the problem) there was no way to change legacy behaviour (too much broken pieces) and the new function covers it.
