Attempting to build packaged libsqlite unexpectedly links against system libraries

[benjunmun]

I have run into an issue installing this gem with the compiled native extension. When I run a build, if there are any sqlite3 headers and libraries installed on my system, the sqlite3_native.so library ends up being dynamically linked against that system library. This is the case even though I am not setting --enable-system-libraries or even if I set --disable-system-libraries explicitly. If I uninstall the library from the system, then the gem extension does link statically against the packaged libsqlite3 the way I expect.

Originally, I ran into this issue on Rocky Linux 9 with Ruby 3.0, sqlite3-ruby 2.0.4. Ancient ruby/rubygems mean that it can't use the precompiled gems, but I really needed the newer packaged version of libsqlite3. I have since been able to reproduce this issue with the latest version of sqlite3-ruby and ruby 3.3.7.

In my mind, this is unexpected behavior, since the gem is going to the trouble of building sqlite3, and since the default is --disable-system-libraries. For my own purposes I've hacked the extconf.rb to static link against the packaged libsqlite3 by default. I'm not an expert with ruby extension building or this gem; If this would be useful for the project, please let me know and I can open a pull request. Also willing to rework it if there is a better approach that anyone is aware of. benjunmun/sqlite3-ruby@31abd7c

[flavorjones]

Hi @benjunmun, sorry you're running into this. It sounds likely that you're running into the problem discussed at flavorjones/mini_portile#118 which is related to the behavior of pkg-config on red hat (and RHEL-derived distros like Rocky).

There's a long-running attempt to clean up some of this up at flavorjones/mini_portile#131 which works OK with the branch on this repo named origin/flavorjones-try-new-mini-portile. Would you be interested in kicking the tires on that branch if I rebase it off of current main?

[benjunmun]

I suspect it's not exactly the same thing - my modern-ruby replication was actually done on Ubuntu 25.04 with
libsqlite3-dev installed. But, I'm also 100% willing to try the branch and report back.

[benjunmun]

Even the current state of flavorjones-try-new-mini-portile looks promising! I can see in the mkmf.log and resulting .so that it's linking the way I expect.

Consider this my vote for following through with the new mini-portile. I can definitely test again with that branch rebased on main, and let me know if there's any other way I could assist. I appreciate your work on this gem!

Built from branch:

ldd ./vendor/ruby/3.3.0/bundler/gems/extensions/x86_64-linux-gnu/3.3.0/sqlite3-ruby-ae5c13996f93/sqlite3/sqlite3_native.so
        linux-vdso.so.1 (0x00007104c1e58000)
        libruby-3.3.so.3.3 => /lib/x86_64-linux-gnu/libruby-3.3.so.3.3 (0x00007104c1600000)
        libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007104c1c9c000)
        libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007104c1ba5000)
        libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007104c1200000)
        libgmp.so.10 => /lib/x86_64-linux-gnu/libgmp.so.10 (0x00007104c1b1d000)
        libcrypt.so.1 => /lib/x86_64-linux-gnu/libcrypt.so.1 (0x00007104c15c1000)
        /lib64/ld-linux-x86-64.so.2 (0x00007104c1e5a000)

Current undesired results:

./vendor/ruby/3.3.0/gems/sqlite3-2.7.3/lib/sqlite3/sqlite3_native.so
        linux-vdso.so.1 (0x000073f05c289000)
        libsqlite3.so.0 => /lib/x86_64-linux-gnu/libsqlite3.so.0 (0x000073f05c0d9000)
        libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x000073f05bfe2000)
        libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x000073f05bfc4000)
        libruby-3.3.so.3.3 => /lib/x86_64-linux-gnu/libruby-3.3.so.3.3 (0x000073f05ba00000)
        libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x000073f05b600000)
        /lib64/ld-linux-x86-64.so.2 (0x000073f05c28b000)
        libgmp.so.10 => /lib/x86_64-linux-gnu/libgmp.so.10 (0x000073f05bf3c000)
        libcrypt.so.1 => /lib/x86_64-linux-gnu/libcrypt.so.1 (0x000073f05befd000)

[flavorjones]

@benjunmun Thanks for trying that branch out. I have a pretty big backlog of open source work at the moment, but I will schedule time to try to merge and release that branch.