On Sunday 02 March 2008, Frans Pop wrote: > The attached patch series is what IMO should be backported. It includes > the following commits (this is on top of the patch that was already > backported in the 1.5.11etch1 release): > * debconf-apt-progress: Add --no-progress option intended to be used > by apt-install in d-i. > * debconf-apt-progress: Notice if the child exited abnormally, and exit > nonzero. > * add --dlwaypoint option (without the cancel changes) > * debconf-apt-progress sometimes gets captured by buggy daemons, > causing tasksel to hang because $debconf_command_eof never becomes > true. STATUS_READ should be the last fd to close, so checking > $status_eof is sufficient (LP: #141601). > The last patch also looks interesting from a stability PoV.
In addition to these commit r2266 should probably be included as it avoids
some annoying errors in the syslog:
* Fix use of 'next' rather than 'return' in debconf-apt-progress
handle_status.
> As these patches cannot be backported in time for beta1, I have also
> prepared a new version of the etch-support udeb that includes an updated
> debconf-apt-progress with these patches and that diverts the Etch version
> debconf-apt-progress for the duration of the installation (from
> post-base-installer to finish-install).
This has been done, but Phil later discovered that this solution does not
cover one particular case: when stable is installed using LVM or RAID,
apt-install is called just before the divert is done and this makes the
progress bar disappear during the last part of base-installer. So a stable
update is very much desired.
Cheers,
FJP
signature.asc
Description: This is a digitally signed message part.
