Bug#238301: After installer has run many debconf questions are not seeded
On Wed, Sep 29, 2004 at 07:13:56PM -0400, Joey Hess wrote: Colin Watson wrote: That's possible ... suggestions for naming of the common class? I'd been thinking the new classes were small enough not to matter too much. Maybe some magic could go into Debconf::Element::Noninteractive. Yes, it seems to me most noninteractive elemenets could just derive from that class though making it work may be too much work/code. OK, let's try this, moving the common show() code into Debconf::Element::Noninteractive: http://patches.ubuntulinux.org/patches/debconf.238301.v2.diff Tested with debootstrap, same changes. I'll probably put this version into Ubuntu for now at least. /var/cache/debconf/config.dat diff attached (for Ubuntu rather than Debian, but hey); /etc is identical except for a popularity-contest configuration file that has the output of uuidgen in it. I guess the couple of Value sets are harmless, since it's just taking the default value from the templates. Right, I think so too. Cheers, -- Colin Watson [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#238301: After installer has run many debconf questions are not seeded
Colin Watson wrote: OK, let's try this, moving the common show() code into Debconf::Element::Noninteractive: http://patches.ubuntulinux.org/patches/debconf.238301.v2.diff Tested with debootstrap, same changes. I'll probably put this version into Ubuntu for now at least. Feel free to merge that into debconf svn. As to whether it should go to the sarge branch.. well, I suppose you could ask Steve. :-) -- see shy jo signature.asc Description: Digital signature
Bug#238301: After installer has run many debconf questions are not seeded
On Thu, Jul 29, 2004 at 02:32:58PM -0400, Joey Hess wrote: Mark Brown wrote: I've found some more instances of this problem. I managed to catch a diff of the config files before and after the upgrade, though I didn't remember to rerun the upgrade or cache the entire database: Now I understand. This is because debootstrap installs packages using the noninteractive frontend which does not set questions to seen as the user has indeed not seen them. It's true that this could lead to some questions that have already set and presumably working values being asked on upgrade. I don't know of a good way to avoid this at this time; debconf's handling of the seen flag for the noninteractive frontend is basically correct; debootstrap's use of the noninteractive frontend is correct, and there's no way d-i can go back after the fact and mark questions as seen. This patch implements a DEBCONF_NONINTERACTIVE_SEEN environment variable that debootstrap can set to make the noninteractive frontend behave a little differently, without risking breaking other uses of the noninteractive frontend. What do you think? http://patches.ubuntulinux.org/patches/debconf.238301.diff -- Colin Watson [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#238301: After installer has run many debconf questions are not seeded
Colin Watson wrote: This patch implements a DEBCONF_NONINTERACTIVE_SEEN environment variable that debootstrap can set to make the noninteractive frontend behave a little differently, without risking breaking other uses of the noninteractive frontend. What do you think? This looks reasonable, I think you could have avoided the duplication of near-identical new classes. I guess you've tested it and verified it has no other effects to a debcootrstrapped system? -- see shy jo signature.asc Description: Digital signature
Bug#238301: After installer has run many debconf questions are not seeded
On Wed, Sep 29, 2004 at 06:25:06PM -0400, Joey Hess wrote: Colin Watson wrote: This patch implements a DEBCONF_NONINTERACTIVE_SEEN environment variable that debootstrap can set to make the noninteractive frontend behave a little differently, without risking breaking other uses of the noninteractive frontend. What do you think? This looks reasonable, I think you could have avoided the duplication of near-identical new classes. That's possible ... suggestions for naming of the common class? I'd been thinking the new classes were small enough not to matter too much. Maybe some magic could go into Debconf::Element::Noninteractive. I guess you've tested it and verified it has no other effects to a debcootrstrapped system? /var/cache/debconf/config.dat diff attached (for Ubuntu rather than Debian, but hey); /etc is identical except for a popularity-contest configuration file that has the output of uuidgen in it. -- Colin Watson [EMAIL PROTECTED] --- /home/cjwatson/warty-chroot-1/var/cache/debconf/config.dat 2004-09-29 15:38:40.440025160 +0100 +++ /home/cjwatson/warty-chroot-2/var/cache/debconf/config.dat 2004-09-29 17:01:29.619595040 +0100 @@ -314,6 +314,7 @@ Template: console-data/keymap/policy Value: Don't touch keymap Owners: console-common, console-data +Flags: seen Name: console-data/keymap/powerpcadb Template: console-data/keymap/powerpcadb @@ -897,6 +898,7 @@ Name: dash/sh Template: dash/sh +Value: false Owners: dash Name: debconf/frontend @@ -931,10 +933,12 @@ Name: discover/manage_cdrom_devices Template: discover/manage_cdrom_devices +Value: true Owners: discover1 Name: hotplug/ignore_pci_class_display Template: hotplug/ignore_pci_class_display +Value: true Owners: hotplug Name: hotplug/net_agent_policy @@ -944,6 +948,7 @@ Name: hotplug/static_module_list Template: hotplug/static_module_list +Value: Owners: hotplug Variables: usbmodules = @@ -983,6 +988,7 @@ Name: locales/locales_to_be_generated Template: locales/locales_to_be_generated +Value: Owners: locales Variables: locales = aa_DJ ISO-8859-1, aa_ER UTF-8, [EMAIL PROTECTED] UTF-8, aa_ET UTF-8, af_ZA ISO-8859-1, am_ET UTF-8, an_ES ISO-8859-15, ar_AE ISO-8859-6, ar_AE.UTF-8 UTF-8, ar_BH ISO-8859-6, ar_BH.UTF-8 UTF-8, ar_DZ ISO-8859-6, ar_DZ.UTF-8 UTF-8, ar_EG ISO-8859-6, ar_EG.UTF-8 UTF-8, ar_IN UTF-8, ar_IQ ISO-8859-6, ar_IQ.UTF-8 UTF-8, ar_JO ISO-8859-6, ar_JO.UTF-8 UTF-8, ar_KW ISO-8859-6, ar_KW.UTF-8 UTF-8, ar_LB ISO-8859-6, ar_LB.UTF-8 UTF-8, ar_LY ISO-8859-6, ar_LY.UTF-8 UTF-8, ar_MA ISO-8859-6, ar_MA.UTF-8 UTF-8, ar_OM ISO-8859-6, ar_OM.UTF-8 UTF-8, ar_QA ISO-8859-6, ar_QA.UTF-8 UTF-8, ar_SA ISO-8859-6, ar_SA.UTF-8 UTF-8, ar_SD ISO-8859-6, ar_SD.UTF-8 UTF-8, ar_SY ISO-8859-6, ar_SY.UTF-8 UTF-8, ar_TN ISO-8859-6, ar_TN.UTF-8 UTF-8, ar_YE ISO-8859-6, ar_YE.UTF-8 UTF-8, az_AZ.UTF-8 UTF-8, be_BY CP1251, be_BY.UTF-8 UTF-8, bg_BG CP1251, bg_BG.UTF-8 UTF-8, bn_BD UTF-8, bn_IN UTF-8, br_FR ISO-8859-1, [EMAIL PROTECTED] ISO-8859-15, bs_BA ISO-8859-2, byn_ER UTF-8, ca_ES ISO-8859-1, ca_ES.UTF-8 UTF-8, [EMAIL PROTECTED] UTF-8, [EMAIL PROTECTED] ISO-8859-15, cs_CZ ISO-8859-2, cs_CZ.UTF-8 UTF-8, cy_GB ISO-8859-14, cy_GB.UTF-8 UTF-8, da_DK ISO-8859-1, da_DK.ISO-8859-15 ISO-8859-15, da_DK.UTF-8 UTF-8, de_AT ISO-8859-1, de_AT.UTF-8 UTF-8, [EMAIL PROTECTED] UTF-8, [EMAIL PROTECTED] ISO-8859-15, de_BE ISO-8859-1, de_BE.UTF-8 UTF-8, [EMAIL PROTECTED] UTF-8, [EMAIL PROTECTED] ISO-8859-15, de_CH ISO-8859-1, de_CH.UTF-8 UTF-8, de_DE ISO-8859-1, de_DE.UTF-8 UTF-8, [EMAIL PROTECTED] UTF-8, [EMAIL PROTECTED] ISO-8859-15, de_LU ISO-8859-1, de_LU.UTF-8 UTF-8, [EMAIL PROTECTED] UTF-8, [EMAIL PROTECTED] ISO-8859-15, el_GR ISO-8859-7, el_GR.UTF-8 UTF-8, en_AU ISO-8859-1, en_AU.UTF-8 UTF-8, en_BW ISO-8859-1, en_BW.UTF-8 UTF-8, en_CA ISO-8859-1, en_CA.UTF-8 UTF-8, en_DK ISO-8859-1, en_DK.UTF-8 UTF-8, en_GB ISO-8859-1, en_GB.ISO-8859-15 ISO-8859-15, en_GB.UTF-8 UTF-8, en_HK ISO-8859-1, en_HK.UTF-8 UTF-8, en_IE ISO-8859-1, en_IE.UTF-8 UTF-8, [EMAIL PROTECTED] UTF-8, [EMAIL PROTECTED] ISO-8859-15, en_IN UTF-8, en_NZ ISO-8859-1, en_NZ.UTF-8 UTF-8, en_PH ISO-8859-1, en_PH.UTF-8 UTF-8, en_SG ISO-8859-1, en_SG.UTF-8 UTF-8, en_US ISO-8859-1, en_US.ISO-8859-15 ISO-8859-15, en_US.UTF-8 UTF-8, en_ZA ISO-8859-1, en_ZA.UTF-8 UTF-8, en_ZW ISO-8859-1, en_ZW.UTF-8 UTF-8, es_AR ISO-8859-1, es_AR.UTF-8 UTF-8, es_BO ISO-8859-1, es_BO.UTF-8 UTF-8, es_CL ISO-8859-1, es_CL.UTF-8 UTF-8, es_CO ISO-8859-1, es_CO.UTF-8 UTF-8, es_CR ISO-8859-1, es_CR.UTF-8 UTF-8, es_DO ISO-8859-1, es_DO.UTF-8 UTF-8, es_EC ISO-8859-1, es_EC.UTF-8 UTF-8, es_ES ISO-8859-1, es_ES.UTF-8 UTF-8, [EMAIL PROTECTED] UTF-8, [EMAIL PROTECTED] ISO-8859-15, es_GT ISO-8859-1, es_GT.UTF-8 UTF-8, es_HN ISO-8859-1, es_HN.UTF-8 UTF-8, es_MX ISO-8859-1, es_MX.UTF-8 UTF-8, es_NI ISO-8859-1, es_NI.UTF-8 UTF-8, es_PA ISO-8859-1, es_PA.UTF-8 UTF-8, es_PE ISO-8859-1, es_PE.UTF-8 UTF-8, es_PR
Bug#238301: After installer has run many debconf questions are not seeded
Colin Watson wrote: That's possible ... suggestions for naming of the common class? I'd been thinking the new classes were small enough not to matter too much. Maybe some magic could go into Debconf::Element::Noninteractive. Yes, it seems to me most noninteractive elemenets could just derive from that class though making it work may be too much work/code. /var/cache/debconf/config.dat diff attached (for Ubuntu rather than Debian, but hey); /etc is identical except for a popularity-contest configuration file that has the output of uuidgen in it. I guess the couple of Value sets are harmless, since it's just taking the default value from the templates. -- see shy jo signature.asc Description: Digital signature
Bug#238301: After installer has run many debconf questions are not seeded
On Thu, Jul 29, 2004 at 02:32:58PM -0400, Joey Hess wrote: I don't know of a good way to avoid this at this time; debconf's handling of the seen flag for the noninteractive frontend is basically correct; debootstrap's use of the noninteractive frontend is correct, and there's no way d-i can go back after the fact and mark questions as seen. Perhaps it would help if when changing an option that would change what you see Debconf should offer to rerun config scripts so that being prompted for unseen options happens then (when you know why) rather than randomly later? It's pretty gross but it's fairly non-invasive and I guess some people might find it useful. -- You grabbed my hand and we fell into it, like a daydream - or a fever. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#238301: After installer has run many debconf questions are not seeded
Mark Brown wrote: I've found some more instances of this problem. I managed to catch a diff of the config files before and after the upgrade, though I didn't remember to rerun the upgrade or cache the entire database: Now I understand. This is because debootstrap installs packages using the noninteractive frontend which does not set questions to seen as the user has indeed not seen them. It's true that this could lead to some questions that have already set and presumably working values being asked on upgrade. I don't know of a good way to avoid this at this time; debconf's handling of the seen flag for the noninteractive frontend is basically correct; debootstrap's use of the noninteractive frontend is correct, and there's no way d-i can go back after the fact and mark questions as seen. -- see shy jo signature.asc Description: Digital signature
Bug#238301: After installer has run many debconf questions are not seeded
On Tue, Mar 16, 2004 at 10:56:22PM -0500, Joey Hess wrote: Mark Brown wrote: It seems to be happening an awful lot - I'd suggest including some kind of automated reinstall of all the packages in the testing of the installer (next time I do an install I'll probably give this a whirl). I can't imagine how it could happen at all without some great care being taken to make debconf behave in a way it truely does not want to, or a debconf behavior I am not anticipating. Can you provide an example of a debconf database before the upgrade, plus the upgrade with DEBCONF_DEBUG='.*' ? I've found some more instances of this problem. I managed to catch a diff of the config files before and after the upgrade, though I didn't remember to rerun the upgrade or cache the entire database: --- config.dat-old 2004-07-23 20:09:40.0 +0100 +++ config.dat 2004-07-25 21:43:15.207208392 +0100 @@ -3,6 +3,12 @@ Value: true Owners: adduser +Name: alsa-base/alsactl_store_on_shutdown +Template: alsa-base/alsactl_store_on_shutdown +Value: autosave always +Owners: alsa-base +Flags: seen + Name: anacron/jobs_in_crontab Template: anacron/jobs_in_crontab Owners: anacron @@ -1060,11 +1066,13 @@ Template: debconf/frontend Value: Dialog Owners: debconf +Flags: seen Name: debconf/priority Template: debconf/priority -Value: medium +Value: low Owners: d-i, debconf +Flags: seen Name: debian-installer/country Template: debian-installer/country @@ -1085,6 +1093,7 @@ Template: debsums/apt-autogen Value: true Owners: debsums +Flags: seen Name: dictionaries-common/default-ispell Template: dictionaries-common/default-ispell @@ -1423,6 +1432,7 @@ Template: mozilla/locale_auto Value: true Owners: mozilla-browser +Flags: seen Name: mozilla/prefs_note Template: mozilla/prefs_note -- You grabbed my hand and we fell into it, like a daydream - or a fever. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#238301: After installer has run many debconf questions are not seeded
On Tue, Mar 16, 2004 at 10:56:22PM -0500, Joey Hess wrote: I can't imagine how it could happen at all without some great care being taken to make debconf behave in a way it truely does not want to, or a debconf behavior I am not anticipating. Can you provide an example of a debconf database before the upgrade, plus the upgrade with DEBCONF_DEBUG='.*' ? I'll try to find another machine to do an install on but I don't know when that'll be. -- You grabbed my hand and we fell into it, like a daydream - or a fever. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#238301: After installer has run many debconf questions are not seeded
Package: debian-installer Version: As of 20040412 Severity: normal During the installer run many packages (for example, setserial) appear to get installed without their debconf configuration scripts being run. This means that when the packages get upgraded they prompt for configuration. While this is probably mostly OK for a stable release since upgrades are infrequent I can see this creating problems with security updates and point releases as well which is probably undesirable. -- System Information: Debian Release: testing/unstable APT prefers unstable APT policy: (500, 'unstable') Architecture: powerpc (ppc) Kernel: Linux 2.4.22-powerpc Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#238301: After installer has run many debconf questions are not seeded
Mark Brown wrote: During the installer run many packages (for example, setserial) appear to get installed without their debconf configuration scripts being run. This means that when the packages get upgraded they prompt for configuration. Incorrect, they are installed in the normal way but with the noninteractive debconf frontend. The config scripts run. Any package that behaves as you describe when installed this way is broken; this is doubly true for packages in the base system which have all been installed in this way since before the release of woody. Please clone and/or reassign this bug to the broken packages. -- see shy jo signature.asc Description: Digital signature
Bug#238301: After installer has run many debconf questions are not seeded
clone 238301 -1 reassign -1 setserial thanks On Tue, Mar 16, 2004 at 01:14:19PM -0500, Joey Hess wrote: Mark Brown wrote: During the installer run many packages (for example, setserial) appear to get installed without their debconf configuration scripts being run. This means that when the packages get upgraded they prompt for configuration. Incorrect, they are installed in the normal way but with the noninteractive debconf frontend. The config scripts run. Any package that behaves as you describe when installed this way is broken; this is setserial is one of the packages I remember doing this to me on upgrade - it prompted me for autosave-types. -- You grabbed my hand and we fell into it, like a daydream - or a fever. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Processed: Re: Bug#238301: After installer has run many debconf questions are not seeded
Processing commands for [EMAIL PROTECTED]: clone 238301 -1 Bug#238301: After installer has run many debconf questions are not seeded Bug 238301 cloned as bug 238390. reassign -1 setserial Bug#238390: After installer has run many debconf questions are not seeded Bug reassigned from package `debian-installer' to `setserial'. thanks Stopping processing here. Please contact me if you need assistance. Debian bug tracking system administrator (administrator, Debian Bugs database) -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#238301: After installer has run many debconf questions are not seeded
Mark Brown wrote: On Tue, Mar 16, 2004 at 01:14:19PM -0500, Joey Hess wrote: Mark Brown wrote: During the installer run many packages (for example, setserial) appear to get installed without their debconf configuration scripts being run. This means that when the packages get upgraded they prompt for configuration. Incorrect, they are installed in the normal way but with the noninteractive debconf frontend. The config scripts run. Any package that behaves as you describe when installed this way is broken; this is It seems to be happening an awful lot - I'd suggest including some kind of automated reinstall of all the packages in the testing of the installer (next time I do an install I'll probably give this a whirl). I can't imagine how it could happen at all without some great care being taken to make debconf behave in a way it truely does not want to, or a debconf behavior I am not anticipating. Can you provide an example of a debconf database before the upgrade, plus the upgrade with DEBCONF_DEBUG='.*' ? -- see shy jo signature.asc Description: Digital signature