It also seems reportbug left my problem report blank. Here is the original problem that should have appeared in the original bug report:
------ I have used mgetty for many years. Recently we began using the debian package cyclades-serial-client to connect external modems. This package allows you to create devices that connect to remote modems via tcp/ip. One thing this product does is create symbolic links in /dev/ that point to corresponding pseduo tty's in /dev/pts. This causes a bit of confusion for programs like pppd which may be spawned by mgetty. I configure mgetty to use the symlinked /dev entries (Like /dev/ttyC01). All seems to run. The modem is accessed properly by mgetty, but the problems start when a PPP session is found and pppd is spawned. In our case we spawn a shell script that eventually launches pppd. pppd can end up calling getlogin() to determine the user of the controlling tty. getlogin() queries the utmp records of the controlling tty. The controlling tty that pppd (or any spwaned process sees) is the non-symlinked device entry, not the original symlinked device that mgetty may have used. This mismatch causes pppd to fail when it calls getlogin() because it looks up the actual device name, while mgetty wrote the symlinked entry into utmp with the symlinked name. I believe that utmp/wtmp should have the real device names written to them, so that spawned programs can query with getlogin() and have returned the appropriate user. This patch follows any device that is a symlink to its real device name, and records that in the utmp/wtmp records. In our case, the failure causes pppd to determine the wrong user of the controlling tty, and then used a default that caused the pppd environment variables to be set incorrectly. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
