ensure --review-pr can find dependencies included in PR

[jfgrimm]

(created using eb --new-pr)

[casparvl]

Well, I can confirm that I had a similar issue with --review-pr:

casparl@software1:~$ eb --review-pr 15132
== Temporary log file in case of crash /scratch/casparl/eb-mpZaHN/easybuild-l3LhZo.log
== found valid index for /sw/noarch/Debian10/2021/software/EasyBuild/4.5.3/easybuild/easyconfigs, so using it...
ERROR: Failed to process easyconfig /scratch/casparl/eb-mpZaHN/tmpqnbSC7/l/loompy/loompy-3.0.7-intel-2021b.eb: Failed to determine minimal toolchain for dep numba 0.54.1

For me, it seemed to occur because we use --minimal-toolchains. Running with --disable-minimal-toolchains works:

casparl@software1:~$ eb --review-pr 15132 --disable-minimal-toolchains
== Temporary log file in case of crash /scratch/casparl/eb-_A6yn7/easybuild-yWomfF.log
== found valid index for /sw/noarch/Debian10/2021/software/EasyBuild/4.5.3/easybuild/easyconfigs, so using it...
Comparing loompy-3.0.7-intel-2021b.eb with loompy-3.0.7-intel-2021b.eb
=====
(no diff)
=====
Comparing numba-0.54.1-intel-2021b.eb with numba-0.54.1-intel-2021b.eb
=====
(no diff)
=====

But, running with --minimal-toolchains with this PR also works:

casparl@software1:~$ eb --review-pr 15132 --minimal-toolchains
== Temporary log file in case of crash /scratch/casparl/eb-0CqDCz/easybuild-_fNPPg.log
Comparing loompy-3.0.7-intel-2021b.eb with loompy-3.0.7-intel-2021b.eb
=====
(no diff)
=====
Comparing numba-0.54.1-intel-2021b.eb with numba-0.54.1-intel-2021b.eb
=====
(no diff)
=====

So, in terms of functionality, I'm happy with this PR. Now only thing todo is make the CI happy too ;-)

[boegel]

@jfgrimm Before making the requested change, please also remove the two commits related to the container test workflow using git push --force, otherwise the commit that was merged via #3985 may get reverted...

[boegel]

@jfgrimm Please update to:

if from_prs and review_pr and review_pr not in from_prs:
    from_prs = from_prs + [review_pr]

Two reasons:

• We shouldn't append to the from_prs list that is passed to alt_easyconfig_paths, since that's a side effect (that change will stick even outside of the scope of the alt_easyconfig_paths call);
• The if ... else construct here doesn't make sense, in particular the else from_prs part doesn't (since we're not assigning the result of that expression)

[boegel]

lgtm