Test report by @geimer
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in this PR)
swat - Linux debian 9.7, Intel(R) Xeon(R) CPU X5550 @ 2.67GHz, Python 2.7.13
See https://gist.github.com/1032bfe7d20d732d09ba806a7b2774a4 for a full test report.

Please add the fix to all versions of DBus 1.13.6 and newer that exist in develop branch and we'll run the tests.

@akesandgren Ehm... That's what I've done already. Or am I missing something?

oops, sorry. I though I saw more of them in the ls output :-)

So please add it to anything with a toolchain 2016a or newer. That's what I should have said...

The question is rather whether it is needed or not... The situation is as follows. Currently, there are 9 DBus easyconfigs in develop. I have tested and with this PR fixed the following:

• DBus-1.13.6-GCCcore-6.4.0.eb
• DBus-1.13.6-GCCcore-7.3.0.eb
• DBus-1.13.8-GCCcore-8.2.0.eb

I have also tested

• DBus-1.10.20-GCCcore-6.4.0.eb

but it didn't seem to require the fix; it found the expat library even without pkg-config for whatever reason. Of the remaining configs, only

• DBus-1.13.0-intel-2017b.eb

specifies expat as a dependency, and since it is of the 1.13 series, likely also requires the fix. The others, i.e.,

• DBus-1.10.8-foss-2016a.eb
• DBus-1.10.8-intel-2016a.eb
• DBus-1.10.12-intel-2016b.eb
• DBus-1.11.20-intel-2017a.eb

don't specify expat as a dependency, thus pkg-config is very likely not required if this is really correct (#3697 might be an indication that it is not, but this should be verified).

Bottom line: I'm happy to include fixes for the easyconfigs I'm unable to test, but I'd like to get some indication what is actually needed to avoid adding unnecessary dependencies.

At least dbus 1.10.8 uses pkg-config during configure, so anything from that onwards needs it.

Added. What about the expat dependency? Is it required in all configs or not?

From 1.10.8 yes.

Hmm... Just noticed that the dependency conflict check is failing with a "No conflicts detected" assertion (see https://travis-ci.org/easybuilders/easybuild-easyconfigs/builds/530758391). This seems to be a logic error...

@geimer No, the assertion is that there are no conflicts, and that is failing (because there are conflicts).

It's a bit confusing, but the other way around could be confusing too (we use this same pattern consistently in the test suite)

Test report by @boegel
SUCCESS
Build succeeded for 9 out of 9 (9 easyconfigs in this PR)
node2079.delcatty.os - Linux centos linux 7.6.1810, Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz, Python 2.7.5
See https://gist.github.com/c8c87457c0b208ddd29f667a481fc367 for a full test report.

Thanks a lot for your persistence on this one @geimer!

Going in, thanks @geimer!