On Wed, Feb 27, 2008 at 03:16:16PM -0600, Brian Cameron wrote: > Following this thread, it seems that there are a lot of build problems > with SQLite3. I wanted to share the attached spec-file that we were
The problems are mostly specific to SFW integration. The SQLite3 build itself is fine except for the libtool issues and the fact that it doesn't do 32- and 64-bit builds in one go, and the fact that it doesn't dynamically link sqlite3 and libtclsqlite3.so with libsqlite3.so. Of these the only challenging problem is the libtool one; the last problem is trivially fixed in my ws and noone minds it. The other issue is minor but has led to a lot of discussion. > previously using to build this module. We didn't seem to have any > build problems. I'm not sure if this addresses the problems you have > highlighted - perhaps we were building it with RUNPATH issues, etc. and > just didn't notice them? You should see what RUNPATH appears in libtclsqlite3.so. That's the only object with a messed up RUNPATH. And you should check whether it built with a dynamic dependency on libsqlite3.so. If you didn't modify SQLite3's Makefile.in then you're building a sqlite3 command shell and a libtclsqlite3.so that directly include the objects that make up libsqlite3.so. In that case the RUNPATH of libtclsqlite3.so won't be messed up, but libtclsqlite3.so will be much larger than it should be (as will be the sqlite3 command shell). > Also, you might consider trying to use our build system to build > this module, if you find it works better. I know SWF has a different Not in SFW. > build system, but perhaps it might make sense to converge more here? > Not sure. At any rate, if you want to build using these spec files > directly, instructions are here: My main question is: did you patch SQLite3's build system _at all_? I don't see any indication in your spec files that you did. Did you apply any other patches that I should know about? Nico --
