On Fri, Aug 07, 2020 at 09:41:01PM +0000, Thorsten Glaser wrote: > Marcel Partap dixit: > > >in my user .tmux.conf, I have put following mechanism > > This is a very complex setup; I cannot easily reproduce what exactly > makes this fail for you. > > >https://www.mail-archive.com/tech@openbsd.org/msg49105.html > > As I said in my earlier mail, this patch combines multiple changes > and is… tricky. > > Matthieu, can you have a look at it as well as the change > http://cvsweb.openbsd.org/cgi-bin/cvsweb/xenocara/app/xidle/xidle.c.diff?r1=1.3&r2=1.4 > possibly introducing the problem for our user?
Hi, I'm currently on vacation. I will look at this when I come back later next week. > > I’m seeing multiple things here: > > • closing stdout and stderr: I agree with that output should go > to .xsession-errors (if open) and would remove that, redirecting > only stdin from /dev/null > > • using execvp: this makes me cringe, both the change and the > current code also. I’d propose introducing two new ways of > specifying the command to run. One would take one argument, > like -program does, but actually run sh -c <argument> (with > shell interpolation, pipes, I/O redirection, etc. possible); > the other (probably a double dash) would collect all remaining > arguments and create an argument vector from them (whitespace- > safe), perhaps with an -argv0 option (like exec -a in some shells) > and retire -program entirely (or keep it for a while but document > it as deprecated, to avoid breaking users right now) > > • calling setsid(2): is this deliberate? What are the effects of > doing so vs. not doing so for the programs run? > > If this is deliberate, perhaps we can introduce a -keepsession > flag that would omit the call to setsid(2) only, for use cases > like Marcel’s. > > >noticed the introduction of a setsid() call, which seem to be the source of > >the > > Are you sure about this? That is, if you locally recompile xidle with > just the call to setsid removed, does your scenatio work? > > Thanks, > //mirabilos > -- > (gnutls can also be used, but if you are compiling lynx for your own use, > there is no reason to consider using that package) > -- Thomas E. Dickey on the Lynx mailing list, about OpenSSL -- Matthieu Herrb
