Enrico Scholz <enrico.scholz(a)wirtschaft.tu-chemnitz.de&gt; writes: See my comments below... on the suspect ioctls. I still think it could be some symbolversioning problem. The main change glibc 2.1 made was in the FILE * structure. In this case the open() call is done by the X libraries (which I assume were not recompiled with glibc 2.1), but XEmacs is doing the F_SETSIG stuff which is only supported if you opened the file with the new open. Or maybe it is something else altogether (it would be interesting to find out if these two failing ioctl's (see below) are indeed the code in request_sigio_on_device. Maybe you can set a breakpoint on that function to see which settings XEmacs does? Then we can compare this with a glibc 2.0 build. Here is your trace, annotated by me. I am learning by doing here so please correct me if I am writing gibberish. execve("/usr/bin/xemacs", ["xemacs", "-vanilla", "-eval", "(call-process \"sleep\" nil nil nil \"3\")"], [/* 41 vars */]) = 0 brk(0) = 0x839e000 [... init code] socket(PF_UNIX, SOCK_STREAM, 0) = 11 uname({sys="Linux", node="ultra.csn.tu-chemnitz.de", ...}) = 0 uname({sys="Linux", node="ultra.csn.tu-chemnitz.de", ...}) = 0 connect(11, {sun_family=AF_UNIX, sun_path="/tmp/.X11-unix/X0"}, 19) = 0 [open up the X connection...] uname({sys="Linux", node="ultra.csn.tu-chemnitz.de", ...}) = 0 fcntl(11, F_SETFD, FD_CLOEXEC) = 0 access("/home/ensc/.Xauthority", R_OK) = 0 open("/home/ensc/.Xauthority", O_RDONLY) = 12 [...] close(12) = 0 [..] write(11, "\20\1\6\0\20\0\0\0SCREEN_RESOURC"..., 24) = 24 read(11, "\1\0\10\0\0\0\0\0T\0\0\0\0\0\0\0"..., 32) = 32 write(11, "\24\0\6\0%\0\0\0T\0\0\0\37\0\0\0"..., 24) = 24 [..] read(11, "\1\0\t\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 32) = 32 read(11, "\1\2=\0\4\0\0\0\270\317P\10\0\0\0"..., 32) = 32 readv(11, [{"2>B\0%m＠qM\0t\0suN\0", 16}, {"", 0}], 2) = 16 fcntl(11, F_GETOWN) = 0 getpid() = 2285 fcntl(11, F_SETOWN, 2285) = 0 ioctl(11, 0x7, 0xbfffef38) = -1 EINVAL (Invalid argument) ioctl(11, 0x6, 0x4) = -1 EINVAL (Invalid argument) This seems suspect to me.... It does look suspiciously like ioctl (filedesc, I_GETSIG, &events); ioctl (filedesc, I_SETSIG, events | S_INPUT); from sysdep.c (The solaris manpage says this what you need to do there to get a signal on input). Could you grep the glibc2.1 include files to compare the definition of I_GETSIG, I_SETSIG, S_INPUT with 0x7, 0x6, and 0x4 ? [...] write(11, "-\0\17\0\t\0\200\5/\0AR-*-courie"..., 68) = 68 getdents(12, /* 0 entries */, 3933) = 0 close(12) = 0 stat("/bin/sleep", {st_mode=0, st_size=0, ...}) = 0 access("/bin/sleep", X_OK) = 0 open("/dev/null", O_RDONLY) = 12 pipe([13, 14]) = 0 fork() = 2286 close(14) = 0 close(12) = 0 close(14) = -1 EBADF (Bad file descriptor) lseek(13, 0, SEEK_CUR) = -1 ESPIPE (Illegal seek) read(13, "", 512) = 0 --- SIGCHLD (Der Kind-Prozeß ist beendet) --- write(10, "\0", 1) = 1