On Sun, Aug 22, 2010 at 07:36:40PM +0300, Niko Tyni wrote: > > The failing test seems to be breaking at the point where it calls > > gtk_new_attach (which is where the language is set). In this case, it calls > > it with a NULL "lang" setting, which causes gtkspell to try looking at the > > LANG environment variable. If this is not set, it falls back to "en". > > It's looking for an English dictionary, such as in the myspell-en-us > or hunspell-en-us packages. Building with either of those makes the > tests succeed.
Aha.. good catch. So it sounds like we just need to Build-Depend on an english dictionary package. Is it worth adding alternative build-deps on all suitable packages[0], or just a specific one do you think? > It would probably also be good to set LC_ALL=C for the test suite > so that user locale settings don't affect the build. (Setting LC_ALL=en_US > or something like that works too, but then we need the locales package > or perl will complain about unknown locale settings.) Yes, that sounds like a good idea. [0] Perhaps: "myspell-en-us | myspell-dictionary-en | hunspell-dictionary-en", a specific package plus the virtual packages provided by myspell and hunspell dictionaries. -- Chris Butler <chr...@debian.org> GnuPG Key ID: 4096R/49E3ACD3 -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org