On Fri, Sep 06, 2019 at 05:52:15PM +0200, Pascal Hambourg wrote: > Le 06/09/2019 à 16:39, Gene Heskett a écrit : > > On Friday 06 September 2019 10:20:14 Greg Wooledge wrote: > > > > > > I could very easily see amanda itself breaking from usrmerge, if it > > > contains programs that try to invoke commands using their full paths > > > (e.g. /bin/rm -f ...). > > Why would that break ? Old paths are still valid, this is the purpose of the > symlinks. > > Before usr merge : > /bin/rm -> ok > /usr/bin/rm -> ko > > After usr merge : > /bin/rm -> ok > /usr/bin/rm -> ok
I was too hasty in writing my example. An actual example of something that fails is a hard-coded /usr/bin/ command in a buster package, assuming that usrmerge will be performed -- and it fails on systems that did *not* perform it. Nevertheless, if Gene were to present a logfile error with a pathname and a "no such file or directory" message, or "command not found", or something similar, that could be an indicator to look at usrmerge. > Unless something is unable to follow symlinks. > > FAILURE DUMP SUMMARY: > > picnc / lev 0 FAILED [data timeout] > > picnc / lev 0 FAILED [[request failed: No route to host]] > > picnc / lev 0 FAILED [dumper TRYAGAIN: [request failed: No route to > > host]] > > This is a network error (ARP or NDP address resolution failure). I wonder > what it has to do with usr merge. Agreed.
