On Thu, Nov 21, 2013 at 3:04 PM, Andrew MacLeod <amacl...@redhat.com> wrote: > On 11/21/2013 02:26 PM, Jeff Law wrote: >> >> On 11/21/13 11:15, Andrew MacLeod wrote: >>> >>> >>> Is there anything in particular one needs to do for plugins? I thought I >>> saw a patch somewhere that changed something in the Makefile, but don't >>> know if that is actually required since I never did that for any of the >>> others. Any plugin which used gimple.h probably needs a few more >>> includes... >> >> We need to make sure the header files that are needed by plugins appear in >> Makefile.in::PLUGIN_HEADERS so that they get installed in a place where >> plugins can find them. >> >> > stupid question perhaps, but aren't most header files a potential plugin > header? Why don't we just install them all... > > No one has complained yet, but in theory any .h I split up over the past > couple of months has the potential to be required... maintaining that macro > in Makefile.in seems kinda lame now that we don't maintain the macros for > building. I'm sure its rotted already.
That's true. It's essentially impossible to know what plugins will need. The easy way is the brute-force approach of adding every header that gets factored out of the original. I chose not to do this in the tree.h refactoring because it exposes too much unnecessarily. I exposed as much as was needed to get the plugins in the testsuite. I think plugins authors are going to have to work closely with us as we refactor headers. At some point, it will make sense to offer a better set of headers for them to use. Diego.