add support for %(sysroot)s template value

[xinan1911]

This improves sysroot support. For example this code line in wget can be generalized using sysroot template.
From
preconfigopts = "export PKG_CONFIG_PATH=/usr/lib64/pkgconfig:/usr/lib/pkgconfig:/usr/lib/x86_64-linux-gnu/pkgconfig && "
to
preconfigopts = "export PKG_CONFIG_PATH=%(sysroot)s/usr/lib64/pkgconfig:%(sysroot)s/usr/lib/pkgconfig:%(sysroot)s/usr/lib/x86_64-linux-gnu/pkgconfig && "

This works for both cases when --sysroot is passed and when it's not (i.e. when it defaults to None).

[boegel]

I was looking into this too, here's the test I came up with to test the use of the %(sysroot)s template:

 def test_sysroot_template(self):
       """Test the %(sysroot)s template"""

       test_easyconfigs = os.path.join(os.path.abspath(os.path.dirname(__file__)), 'easyconfigs', 'test_ecs')
       toy_ec = os.path.join(test_easyconfigs, 't', 'toy', 'toy-0.0.eb')

       test_ec = os.path.join(self.test_prefix, 'test.eb')
       test_ec_txt = read_file(toy_ec)
       test_ec_txt += '\nconfigopts = "--sysroot=%(sysroot)s/"'
       write_file(test_ec, test_ec_txt)

       ec = EasyConfig(test_ec)
       self.assertEqual(ec['configopts'], "--sysroot=/")

       update_build_option('sysroot', self.test_prefix)

       ec = EasyConfig(test_ec)
       self.assertEqual(ec['configopts'], "--sysroot=%s/" % self.test_prefix)

[casparvl]

Yep, update_build_option('sysroot', self.test_prefix) is the thing we needed in the test, and didn't know about.

I like that you captured it in one test, we'll update that. I also like how you just modified the toy.eb instead of setting the full text (we stole that from the PR you gave us as example, but it's not really needed here).

I do prefer our approach to using echo's, I find the configopts = "--sysroot=%(sysroot)s/" potentially confusing. It could be mistaken for trying to test if one can pass --sysroot as a configopt, whereas with the echo it's clearly about the template value.

I also think we should stick to setting it in the three pre*opts we did now (again, stolen from the PR you gave us as example), as it verifies that template value is set in all of these stages of (which are probably the most relevant ones in terms of where it might be used) and not just in the configure step.

I'll merge your approach into ours :)

[casparvl]

Ah, wait, I only now see that you are actually not relying on running all steps, you just query the options again. Ok, then I understand why you did the suggestive --sysroot. I'll go with that, it saves some time in the CI since you don't have to actually run the steps in your case.

I'll still do it for configure build and install opts, just to validate that it works in all of those.

[casparvl]

Hm, that's a completely different test. I guess the update_build_option carries over between tests, so we should probably put it back. I thought that build options were reset between tests anyway? Maybe not...

[boegel]

Hm, that's a completely different test. I guess the update_build_option carries over between tests, so we should probably put it back. I thought that build options were reset between tests anyway? Maybe not...

Hmm, if that's the case, then we should definitely fix that (in a separate PR), configuration options shouldn't carry over tests at all...

edit: Ah, you figured it out, some other tests were failing because the sysroot template was introduced, had nothing to do with carrying over of configuration options across tests, good

[boegel]

lgtm