http://bugzilla.suse.com/show_bug.cgi?id=1132282 http://bugzilla.suse.com/show_bug.cgi?id=1132282#c10 --- Comment #10 from Daniel Dadap --- Created attachment 803053 --> http://bugzilla.suse.com/attachment.cgi?id=803053&action=edit Debug builds of initscrtest Apologies; I neglected to build with -g. Attached new binaries built with -g and -O0, though it hardly really matters, since the issue demonstrated by initscrtest is really secondary to the primary issue (which is demonstrated by the previous hello-ncurses test program). Good find with the macros vs. functions difference. It seems that this particular difference comes from NCURSES_OPAQUE. Indeed, if I build a non-reentrant ncurses configured with --enable-opaque-ncurses and build/link initscrtest.c against that, I get the desired behavior that the reentrant ncurses exhibits. So this at least identifies the minimum that we want to use for nvidia-installer to ensure binary compatibility with both reentrant and non-reentrant ncurses: avoid referencing stdscr directly (including avoiding the macros that assume stdscr as the default WINDOW * for the APIs that operate on WINDOW *), and build/link against ncurses with NCURSES_OPAQUE set. However, the underlying problem could still be problematic for other binary programs that are built and linked against a non-reentrant, non-opaque ncurses 6. Again, I don't know how common such programs may be, but it's worth consideration nevertheless. -- You are receiving this mail because: You are on the CC list for the bug.