Christophe LYON wrote: > >> >> Weird. Can you run the following, hit ^D and post the log: >> >> truss -o log src/sort >> > Sure, here it is.
ioctl(0, TCGETA, 0xFFBEF154) = 0 read(0, 0xFF2C35EC, 1024) (sleeping...) read(0, "\n", 1024) = 1 read(0, 0xFF2C35EC, 1024) (sleeping...) read(0, 0xFF2C35EC, 1024) = 0 ioctl(1, TCGETA, 0xFFBEF10C) = 0 write(1, "\n", 1) = 1 llseek(0, 0, SEEK_CUR) = 14360747 close(0) = 0 close(1) = 0 close(2) = 0 _exit(0) So that does exit() correctly. Hmm, where is that "\n" comming from? Perhaps that's affecting the expect session. To exclude races, is this prob 100% reproducible with: (cd tests && make check TESTS=misc/tty-eof VERBOSE=yes) Nothing else obvious comes to mind. Perhaps you could add the truss call to the tty-eof script to see what sort is doing when it doesn't exit(). I'd need access to a solaris 8 system to debug further. cheers, Pádraig. _______________________________________________ Bug-coreutils mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-coreutils
