reassign 223872 debconf thanks On Tue, 16 Dec 2003, Brian McGroarty wrote:
> On Tue, Dec 16, 2003 at 11:33:47AM -0600, Adam Heath wrote: > > On Mon, 15 Dec 2003, Brian McGroarty wrote: > > > > > On Mon, Dec 15, 2003 at 06:28:37PM -0600, Adam Heath wrote: > > > > On Sat, 13 Dec 2003, Brian McGroarty wrote: > > > > > > > > > Package: dpkg > > > > > Version: 1.10.18 > > > > > Severity: important > > > > > > > > > > dpkg proceeds even if the preconfigure scripts do not successfully > > > > > run. > > > > > > > > dpkg is not running that. debconf is. Which is either called by apt > > > > directly(preconfigure), or by a package's preinst/postinst. If it's the > > > > latter case, it's a bug in the package. > > > > > > > > So, what package are you having problems with, so that this bug can be > > > > reassigned to it. > > > > > > This happened during the preconfigure stage after using dselect. Do > > > you know whether dselect works through apt, or whether should this be > > > a dselect bug? > > > > > > Thanks, Adam. > > > > It depends if you have choosen apt as the access method. Again, this isn't > > a > > bug in dselect, dpkg, apt, or debconf. It's a bug in the exact package's > > maintainer scripts. > > apt is the access method, yes. So it seems this would be an apt bug if > apt is responsible for executing the preconfig step and checking for > success. No, debconf. > > > If dpkg-preconfigure(part of debconf) is not displaying the package name > > when > > errors occur in the config script, then file another bug on debconf in that > > case. > > The error isn't from errors in the config script. It's that the > update/install process isn't stopped when the scripts cannot be run at > all. > > When /tmp is mounted noexec, security prohibits files from being run > on that filesystem. As the preconfig scripts are extracted and run > here, an error is printed when the scripts cannot be run, but > dpkg-preconfigure never reacts to the failure. If I'm understanding > you clearly, apt should be responsible for stopping on this error, > just as it would if the script had run but terminated with an error. >
