I already found one in the ntfs library: http://dev.haiku-os.org/browser/haiku/trunk/src/add-ons/kernel/file_systems/ntfs/utils.c#L1056 (another hit actually leads to a wrong comment but the code itself is safe). Seeing other parts of the function actually assign errno and return -1 it seems inconsistent anyway. The ipro1000 has lots of hits but it seems to only be error codes it defined itself so it's ok (seeing how many there are why didn't they #define them < 0 in the first place ? :p). src/add-ons/kernel/drivers/bus/pcmcia/ds.c has several hits too, (do we use it yet ?) src/add-ons/kernel/drivers/bluetooth/h2/h2generic/h2transactions.c has 2 hits, Oliver how could you !? FIXED. src/bin/gdb has some lines that'd deserve attention. src/bin/pcmcia-cs/cardctl.c has some hits too. Seems to be enough, I'd thought there would be more. So much for people saying it's been done everywhere... pcmcia-cs actually comes from linux... so does the ntfs code, it's part of linux-ntfs. So really it seems the infection started there :) François.