This currently breaks the build due to configure not being executable by the looks of it. Also, updating libffi is sadly not a trivial task: #3137

@yorickpeterse, sorry about that, I fixed and it worked fine.
I've been using the libffi 3.1 for some time, to make sure it works, and is working well in my environment.

Ah, thanks. If I recall there are also some custom changes we've made to libffi (#3137 (comment)). We'll need to make sure those are either re-applied or simply no longer needed.

@yorickpeterse, I did a diff and I saw that the only modification is the removal of FFI_MMAP_EXEC_SELINUX macro.
You want that I do some specific test to see if this modification is still necessary?

Apparently that macro was disabled in 06e672b. I'm not sure if we still need this. @brixen @dbussink any thoughts on this?

@yorickpeterse Was gone on vacation to 🇯🇵, but yeah, we still need that macro. Otherwise FFI tries to enable workarounds when SELinux is enabled for executable memory that breaks stuff. We need executable memory anyway for the JIT so it doesn't change any requirements.

Aha, I had a hunch that might've been the case. @gustavotemple could you re-apply the patch discussed in 06e672b and add that to the PR? Once done it should probably be good to go.

@yorickpeterse, @dbussink, @brixen, I submitted the both commits and they were builded OK.

Also looks like 3.2 was released just today, maybe makes sense to immediately update to that. Also the MMAP workaround only seems to add comments, so I wonder if that is still enabled in libffi.

@dbussink: Done !

@gustavotemple thanks for working on this!

@brixen: welcome! :)