Hi again Julian, I wonder which library the getopt(_long) calls are taken from - is it from the kpse library or the libc++ library? If the prototypes are different, are the behaviours different as well? If so, might it be wiser to rename the functions so that the behaviour is consistent across platforms?
Given the number of people who have touched this code since I originally adapted it for kpse, I no longer know exactly what happens without intensive investigation; it's not easy to tell from the maze of #ifdef's in getopt.h. Anyway, I doubt the system vs. kpse getopt behavior is different in any way that matters to us. I originally included getopt (and a lot of other things) in kpathsea precisely so that the TeX programs could have the same behavior across platforms and not have to repeat all this system-dependent ugliness in multiple places. When getopt(_long) was in libc, it was used from there. Otherwise, it was pulled in from kpathsea. These days, what with the plethora of distros doing crazy things, new versions with incompatible definitions, compiling C code under C++, etc., etc., there are a corresponding plethora of new and useless problems. We can do something to cut the Gordian knot, since clearly it is a pain to keep having to adjust with the never-ending new system "improvements", but like I said in my last mail: it requires investigating/changing all uses of getopt(_long) everywhere in the source. It would make sense in the long term, but don't want to go there now, just for the sake of a trivial mismatch under a brand-new compiler version that we don't even want to use for TL'12. What's simple is to omit the declaration when compiling under C++, since that is where the problem occurs. I have just tested that omission of the declaration, since no one else was jumping at doing the work, and it did compile everything fine under 4.7.0. Now I'm running a compile under an old version to see how that goes. I'll commit something when I have some actual evidence. karl -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org