>>> On 04/20/2018 05:43 AM, Peter Maydell wrote: >>> Since we can now safely call serial_mm_init(..., chr=NULL, ...), I'll >>> work on simplifying and fixing this. >> >> Correct. There are a bunch of dubious workarounds in the >> codebase which try to handle the problem at different levels, >> but we've decided that the right thing is for the NULL pointer >> to be dealt with in one place at the bottom (in the qemu_chr_fe_* >> functions). > > OK, I'll happily clean that :)
Hmmm removing the "if (serial_hds[0])" check I reach: void serial_realize_core(SerialState *s, Error **errp) { if (!qemu_chr_fe_backend_connected(&s->chr)) { error_setg(errp, "Can't create serial device, empty char device"); return; }