honey@allegra.UUCP (07/28/83)
when remote mail is initiated and uux is building the X. file for the next site, it reaches out into thin air to pick on some poor luser to fill in the User field of the U line. (actually, it's deterministic, and even sometimes gets it right, but the mechanics aren't worth describing -- they're wrong.) i have changed allegra's uux as follows: if pred!source sends mail to allegra!succ!sink, the outgoing X. file (actually a D. file on my side, an X. file at succ) will have a U line U pred!source allegra in english, then, i am putting the return address (relative to the local site) in the User field of the U line. ok, so what's broken? your uuxqt is what -- it probably allocates about 10 characters for the User field; i think you can see you'll need more than that. please go look (search for X_USER in uuxqt.c and check array sizes). BUFSIZ should be enough to hold the return address. (sites that talk to wat* have already fixed this bug -- even shortish names like bstempleton overrun the array!) of course, if you don't have a source license, then you have a problem. peter honeyman
dmmartindale@watcgl.UUCP (Dave Martindale) (08/02/83)
Actually, after a few days during which we tortured uuxqt's in several parts of the country, watmath has been very careful to trim outgoing user names to 8 characters. Watcgl, which has been careful to make sure its prospective connections can handle slightly longer names, still never sends out anything longer than 12 characters, and the local versions of uuxqt (embarrasingly) would blow up if fed anything longer than 29. So full return addresses are likely to blow up almost EVERYONE's uuxqt. This would make it so much more likely that errors will get mailed back to where they should go, though, that it's worth the trouble.
honey@allegra.UUCP (08/03/83)
i couldn't take the heat so i backed out of the return address on the U line. however, our (us == redman, nowitz, honeyman) X. files are taking shape into something beautiful. cat one if you dare! peter
jmc@root44.UUCP (John Collins) (08/09/83)
Half this kind of problem would be avoided if more UNIX programs were written to allocate memory dynamically. It is only a little bit more effort, and yet program after program has arrays declared with magic bounds (which are only rarely checked against). Uucp is a particularly bad offender in this respect. I'm for promoting flames directed at this kind of sloppy programming! John Collins ....!vax135!ukc!root44!jmc