John E. Malmberg wrote: > Rafael Garcia-Suarez wrote: > >> On 17/03/07, Alexey Tourbin <[EMAIL PROTECTED]> wrote: >> >>> I think that the easiest way to address the prolem is, in perl5db.t, >>> to check for /dev/tty instead of /dev/null. >> >> >> I actually added that condition (as change #30609), thanks. > > > This would also appear to cause the test to be skipped on VMS. > > VMS does not currently have a /dev/tty, which is why the code assumes > SYS\$COMMAND instead. > > That probably could be done better on VMS, but it will be a little > while before I have enough tuits free to look at it. > > The impact of skipping this test on VMS is minor. The test is failing > right now anyway. As near as I can determine, the VMS specific code > is preventing the debugger from accepting a script for input instead > of the command terminal. > > I would also advise any VMS maintainers not to assume that testing for > /dev/tty will fail on VMS. Even on older releases, directory > structures and logical names can cause a false positive. And /dev/tty > may start working in a future release of VMS.
There is an equivalent of /dev/tty of VMS, and that is 'TT:'. SYS$COMMAND can get repointed, but TT will point to the console terminal for the process, if it has one. Batch and detached processes don't have a TT. But if you invoke a script through a pseudoterminal, it will have one.