On 2005-05-20, 01:44, Reuben Thomas wrote: > On Thu, 19 May 2005, Simon Kågström wrote: > > >zile crashes with a segmentation fault when I start it in GDB in emacs. It > >seems like zile lacks some checks for certain terminals. The bug with emacs > >occurs in term_termcap.c:135 (term_refresh), where tgoto() returns NULL, > >but that return value is never checked. > > Thanks for this. If tgoto doesn't return a string, there's not much > point in Zile continuing, so what I anticipate doing is wrapping tgoto > so that its return value is checked, and if NULL, then Zile exits with a > fatal error. Does this seem reasonable? If Zile can't do cursor > positioning, there doesn't seem to be much point continuing.
Yes, I guess that's reasonable. I just tried vi from within Emacs/GUD (I'll have the police after me now! ;-)), and vi starts but does not look correct (does not clear the window). It is possible to quit it normally though and it doesn't crash but it's very hard to edit anything. I guess this is because Emacs/GUD doesn't behave like a real terminal, and I would prefer if zile would just quit with an error message in situations like this. // Simon
