Dennis Heimbigner

> The current system is a pain for packagers because all of the filters get installed into DESTDIR and need to get manually removed. This method deletes them automatically. I...

> Would putting the test plugins in [check_LTLIBRARIES](https://www.gnu.org/software/automake/manual/html_node/Libtool-Convenience-Libraries.html) instead of plugin_LTLIBRARIES, lib_LTLIBRARIES, or tmp_LTLIBRARIES work, or does that result in them not getting found during the tests (or cause [the...

> $(DESTDIR)/$(abs_top_builddir) The use of this by the packager is a major problem since abs_top_builddir is itself intended to be an absolute path. I wish noinst_LTLIBRARIES worked when also using...

I think the solution in this situation (for packagers) is to transfer the responsibility for installing the filter libraries to be the done by the packager. This, I think, entails...

I was not aware of check_LTLIBRARIES. I think your solution is correct; I will test. I do note that the requirement for -rpath seems to conflict with the desire above...

If I put the plugin dir ahead of all the test directories, then I think the check libraries should get built at the proper time. I need to do some...

So I did a test in which I used either noinst_LTLIBRARIES or check_LTLIBRARIES instead of tmp_LTLIBRARIES. The problem I saw was that in both cases it built the static library...

Yes, that was it. It also solves the noinst_ problem so I can use noinst_LTLIBRARIES instead of check_LTLIBRARIES

When I run my tests with no -rpath for the installable filters, it fails. See e.g. https://github.com/DennisHeimbigner/netcdf-c/runs/7120535287?check_suite_focus=true

This should be closed in favor of PR https://github.com/Unidata/netcdf-c/pull/2440