Hi, After some investigation, I finally found what was happening in this bug. It follows this sequence :
- template_db_new(), question_db_new(). At this point, both dbs are empty. - frontend_new() creates a new frontend as g_frontend, based on the --frontend parameter or config file. It tries to set debconf/frontend, which doesn't exist yet. - g_config->methods.load() loads the database, including debconf/frontend. - first call to go() (int commands.c), reading debconf/frontend, which is different from the current one. Therefore it shuts it down, and frees it (it is the one g_frontend points to, remember). - Things go on, until the call to cleanup(), which tries to shutdown, and free, g_frontend, for the second time. Hence the segfault. This would also happen when dpkg-reconfigure-ing several packages. This points to several issues : - The templates and questions dbs should be loaded *before* the call to frontend_new(). It would avoid the unexpected change of frontend. The patch is trivial for that. - The change of frontend while running is not handled gracefully. I believe the safest solution would be to update g_frontend with confmodule->frontend in runconfmodule(). I have a patch ready for all this, and will push things in git by tomorrow hopefully. Regis -- To UNSUBSCRIBE, email to debian-boot-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/1308695022.8499.35.camel@x200s
