Marcus Brinkmann <[EMAIL PROTECTED]> writes: > If anybody wants to work on this, he would have to find out why the read on > the file descriptor returns EIO. I am lost at this point. screen does a lot > of funny stuff with fd_sets, too, maybe this is related.
The terminal server itself never returns EIO for a read. However, the library does in one very important Posix case: if the reading process is reading from its controlling terminal, but is not in the foreground process group, then it normally gets SIGTTIN. If SIGTTIN is blocked or ignored, however, then the read returns EIO. I'd check this state completely before the failing call and see if that illuminates anything.
