From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.9) Gecko/20020513 Description of problem: I am setting up an annex with dumb terminals using a d413-unix emulation and also Kermit-95 using a D217 emulation from my W2k box. These are Date General terminal emulations. Both of these emulations exist in the termcap and terminfo files. The problem is that when I attempt to login using these emulations it immediately disconnects me. If I login with an "unknown" TERM then it lets me in but as soon as I set the TERM it disconnects me (TERM=d413-unix ; export TERM). Below are two examples with the term set right (in these cases, I never get a shell prompt when trying to connect). If I am in BASH, I get a crash and then disconnected: login: Password: Last login <snip> malloc: block on free list clobbered last command: TERM=d413-unix;export TERM Stopping Myself... telnet <snip> closed If I am in TCSH, I get an error and then disconnected: login: Password: Last login <snip> root: Command not found. root: Undefined variable telnet <snip> closed Because both shells do it, I suspect it isn't the shell but something else. I'm not really sure where to search for help on this one... I dug into google groups and found one reference to a bug in RedHat back in 6.x (18 months ago), but nothing since then. http://groups.google.com/groups?hl=en&lr=&ieUTF- &&am&&threadm=1010201152859.AA87584.SM.CWRU.Edu&rnum=4&prev=/groups?q% 3Dtermcap+%22stopping+myself%22&hl=en&lr=&ie=UTF-8&selm% 3D1010201152859.AA87584.SM%40nike.INS.CWRU.Edu&rnum=4 Any ideas? I have worked with Termcap files on DG Unix, but never ran into a disconnect like this. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1.Login with unknown TERM 2.TERM=d413-unix ; export TERM Actual Results: Crashed session Expected Results: Term type set properly Additional info: Bug in our termcap rather... The d413-unix termcap type ends up as a modified d216e+ terminal type. If you read the termcap entry for it, the last line looks like: :sf=^J:ta=^I:up=^^PA:tc=dgkeys+15:tc=d216-dg: The last two bits of these are the interesting ones: tc=dgkeys+15 tc=d216-dg You can't use tc= twice in a single termcap entry. This probably shouldn't segv our binary (and I now hope that we don't have any setuid apps that use termcap; a brief check didn't find anything but someone should check for sure). Anyway, if you remove the "tc=dgkeys+15" bit, this problem goes away. Put this in bugzilla and please include my commentary. In the meantime, you can tell the customer to change that last line to: :sf=^J:ta=^I:up=^^PA:tc=d216-dg: and he should be all set. Erik
Created attachment 123163 [details] libtermcap-2.0.8-44.src.rpm Many thanks for your examples ! 'termcap' seems to be OK (you can use tc= twice). The bug was in libtermcap. Fixed in devel (CURRENT RELEASE). Tested in i386, x86_64, ... (using libtermcap-2.0.8-44 in your scenario).