On Mon, Sep 16, 2002 at 11:46:41AM +0200, Martin Sj?gren wrote: ... > When I ran ../debconf test.config (using the gtk2 fe) and tried using > the Back button it segfaulted: > > 236 *pq = (*pq)->prev->prev; > (gdb) p (*pq)->prev > $2 = (struct question *) 0xe8 > (gdb) p (*pq)->prev->prev > Cannot access memory at address 0x108 >
Great! Thanks for the test case! I'll take a look at this asap. > So it seems that the ->prev->prev pointer points to garbage. But why use > ->prev->prev? If we want to go back, should we not use ->prev? Or is it > a hack so that in the next iteration, you can use q->next? > yes. since it's in a loop that it iterating over the questions. > Bear in mind that for the first element of a double-linked list, the > invariant curr->prev->next == curr doesn't hold, since curr->prev > doesn't exist... > Of course. I'll fix it soon. > Also, I really dislike buttons that disappear and come back. If there's > going to be a Back button in the next dialog, please have a disabled > Back button in the first one too. Oh, another issue. Using Back and Next > isn't that good IMO. Previous and Next, or Back and Forward, but not > Back and Next... > That's a very good point. > > When I get time to hack some more on it, I'll see if I can make the > "main loop" more intelligent when it comes to backing up. > > > Regards, > Martin I think we also have the larger problem that there will be multiple packages asking lists of questions, so finish isn't really appropriate. We need a mechanism to go back to a previously configured packae. -- michael cardenas | lead software engineer lindows.com hyperpoem.net | GNU/Linux software developer debian.org "The secrets of this earth are not for all men to see, but only for those who will seek them" - Ayn Rand, Anthem
msg22209/pgp00000.pgp
Description: PGP signature
