snapshot 20050114 race (on list)

[Pierre A. Humblet]

At 06:42 PM 1/22/2005 -0500, Christopher Faylor wrote:
>
>As it turns out, my concerns were somewhat overblown because you just
>have to make sure that you reset current directory while you still are
>in possession of the write pipe.  That's no problem, as it turns out.

Looks good. On NT, ChangeCurrentDirectory needs x access to the destination,
so the current scheme may fail. Could try other things such as $TEMP, but
that's not perfect either ... I wish there was a way to close that cwd handle.

Also, is there always a C: drive? One could use "\\" instead of "C:\\" and
let Windows decide, but this may magnify the access right issue.

I am still puzzled as to why maybe_set_exit_from_windows is called in
exit and not in spawn:
    case _P_OVERLAY:
      if (!synced)
        HERE

The comment about moving it down two lines is misleading. It's 
kind of the opposite, you would need a new else clause if you move it. 

If you leave it where it is,  self->alert_parent (0) will not be called
when the exit code is actually set in  maybe_set_exit_from_windows.

Pierre