On Tue, Jul 05, 2011 at 01:33:22PM +0700, Ivan Shmakov wrote: > >>>>> Ivan Shmakov <i...@gray.siamics.net> writes: > > > I should probably try to compare strace(1)'s??? > > I've grep'ed the strace for ???rt_sig???. The results for the first > and second SIGTSTP are MIME'd. > > And there's a difference: > > ??? first SIGTSTP: > > rt_sigaction(SIGTSTP, {0x7f061efbf800, [], SA_RESTORER|SA_RESTART, > 0x7f061e5461e0}, NULL, 8) = 0 > > ??? second SIGTSTP: > > rt_sigaction(SIGTSTP, {SIG_IGN}, NULL, 8) = 0 > > Do I understand it correctly that the SIGTSTP signal handler is > not properly reset in the latter case?
Perhaps - lynx is calling sigaction, which in turn calls rt_sigaction. But you may be looking at the initialization section, since that's also where SIGWINCH is setup. Lynx is using ncurses for handling SIGTSTP most of the time. It sets/unsets a handler when it is doing a system() call. If your test scenario doesn't run any external programs, then perhaps the problem is in ncurses (or the way lynx uses it). -- Thomas E. Dickey <dic...@invisible-island.net> http://invisible-island.net ftp://invisible-island.net
signature.asc
Description: Digital signature