Test report by @branfosj

Overview of tested easyconfigs (in order)

• SUCCESS GCC-11.1.0.eb
• SUCCESS GCC-10.3.0.eb
• SUCCESS GCC-9.3.0.eb
• SUCCESS GCC-7.3.0-2.30.eb

Build succeeded for 4 out of 4 (4 easyconfigs in total)
bear-pg0211u03a.bear.cluster - Linux RHEL 8.3, x86_64, Intel(R) Xeon(R) Gold 6248 CPU @ 2.50GHz (cascadelake), Python 3.6.8
See https://gist.github.com/1394d137ea572a614e536d3b2fadce38 for a full test report.

Test report by @branfosj

Overview of tested easyconfigs (in order)

• SUCCESS CUDA-11.0.2-GCC-9.3.0.eb
• SUCCESS CUDA-11.0.2-iccifort-2020.1.217.eb
• SUCCESS CUDA-11.1.1-GCC-10.2.0.eb
• SUCCESS CUDA-11.1.1-iccifort-2020.4.304.eb

Build succeeded for 4 out of 4 (4 easyconfigs in total)
bber0501u03a.bb2.cluster - Linux RHEL 8.3, x86_64, Intel(R) Xeon(R) CPU E5-2690 v3 @ 2.60GHz (haswell), Python 3.6.8
See https://gist.github.com/aee048619c2edc046f8f8735221abb8f for a full test report.

wouldn't it make more sense to move to __init__ instead?

wouldn't it make more sense to move to __init__ instead?

The __init__ happens before the dependencies are loaded so get_software_root returns None.

prepare_step then? (after calling super's)

(edit: prepare_step is not skipped when running --module-only)

For this easyblock it has no impact and the change you suggested works. However, it's hard to ensure that any easyblock that derives from this (and that we may not even know about) may not be leveraging self.altroot in another step. For backwards compatability, it is probably best to add

self.altroot = None
self.altversion = None

to the init method and create an initialisation method that you can use both in the config and module methods. I did something similar recently for Tkinter (#2416)

Test report by @ocaisa

Overview of tested easyconfigs (in order)

• SUCCESS GCC-10.2.0.eb

Build succeeded for 1 out of 1 (1 easyconfigs in total)
generoso - Linux centos linux 8.2.2004, x86_64, Intel(R) Xeon(R) CPU E5-2667 v3 @ 3.20GHz, Python 3.6.8
See https://gist.github.com/30131e8a77bc49fdc662070388168483 for a full test report.

Fixes the issue correctly (CMAKE_PREFIX_PATH was added to framework in a later EB release than the original build):

== comparing module file with backup /users/ocaisa/.local/easybuild/modules/all/GCC/10.2.0.bak_20210701090102_2518753; diff is:
--- /users/ocaisa/.local/easybuild/modules/all/GCC/10.2.0.bak_20210701090102_2518753
+++ /users/ocaisa/.local/easybuild/modules/all/GCC/10.2.0.lua
@@ -28,9 +28,10 @@
     load("binutils/2.35-GCCcore-10.2.0")
 end
 
+prepend_path("CMAKE_PREFIX_PATH", root)
 setenv("EBROOTGCC", "/users/ocaisa/.local/easybuild/software/GCCcore/10.2.0")
 setenv("EBVERSIONGCC", "10.2.0")
 setenv("EBDEVELGCC", pathJoin(root, "easybuild/GCC-10.2.0-easybuild-devel"))
 
--- Built with EasyBuild version 4.3.4.dev0-r00b1c75f775c9d423a3ff992c98e0b2c534e9f86
+-- Built with EasyBuild version 4.4.1.dev0-r76b6ab1773faf95f8bac64e2943b1e181d38bdaa
 

== ... (took 1 secs)