If inside openvz container (I did not managed to reproduce in chroot) you build ncurses with USE="gpm" ufed will die with "Segmentation fault" inside screen (or just set TERM=screen) with the following backtrace: #0 0xb7dd2b35 in Gpm_Open (conn=0x809dfb4, flag=0) at lib/liblow.c:262 #1 0xb7f1ac65 in enable_gpm_mouse (enable=<value optimized out>) at /var/tmp/portage/sys-libs/ncurses-5.6-r2/work/ncurses-5.6/ncurses/base/lib_mouse.c:390 #2 0xb7f1b487 in _nc_mouse_init () at /var/tmp/portage/sys-libs/ncurses-5.6-r2/work/ncurses-5.6/ncurses/base/lib_mouse.c:442 #3 0xb7f1b789 in mousemask (newmask=15, oldmask=0xffffffff) at /var/tmp/portage/sys-libs/ncurses-5.6-r2/work/ncurses-5.6/ncurses/base/lib_mouse.c:1264 #4 0x0804a11e in initcurses () at ufed-curses.c:40 #5 0x0804ae9f in main () at ufed-curses-checklist.c:341
Created attachment 144683 [details, diff] gpm-1.20.1-deb-alt-xterm_mouse_support.patch Attached patch I took from the following link: http://www.sisyphus.ru/srpm/Sisyphus/gpm/patches fixes the issue. But seems that this patch possibly could create new problems. At least one program (clines, not in portage http://manticore.2y.net/prj/clines-g.html) loose some formatting (adds additional spaces) with that patch. In Russian the problem was described in the following message: http://lists.altlinux.org/pipermail/devel/2006-August/047745.html Short version in English. cline has the following code initscr(); int fd = Gpm_Open(); if (fd == -2) { Gpm_Close(); } ...........game here................... endwin(); initscr() : opens terminaland sets correct state for the game Gpm_Open(with this patch !!) : calls setupterm(), and this overwrites values save in ncurses endwin() restores the state just before Gpm_Open() , and this restores state correct for the game but not for terminal, thus additional spaces I'll try to investigate if this is correct, but generally I do not use gpm so I'm not sure that I'll return to this problem in the nearest future. So I'm posting this bug generally for documentation purposes.
Created attachment 144684 [details, diff] gpm-1.20.1-mdk-alt-consolename.patch Together with the previous patch this patch also affects same parts of code, but I did not found why this patch is required. This requires investigation too.
i dont suppose people have tested gpm-1.20.5 ?
(In reply to comment #3) > i dont suppose people have tested gpm-1.20.5 ? > currently no but i can test it today on amd64 machine
Thank you Mike for heads up. This issue is really fixed in gpm-1.20.5. At least I'm unable to reproduce this problem.
