From a86c61462b5d0dab3df979b50aed79ad1b055d0a Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 16 Dec 2022 11:01:25 +0100
Subject: [PATCH 01/31] Add CUDA support to software_layer

---
 EESSI-pilot-install-software.sh |  28 ++++
 eb_hooks.py                     | 265 ++++++++++++++++++++++----------
 eessi-2021.12.yml               |  14 ++
 3 files changed, 227 insertions(+), 80 deletions(-)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index 2bc6876965..1698066bf4 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -394,6 +394,34 @@ $EB --from-pr 15885 OpenBLAS-0.3.15-GCC-10.3.0.eb --robot
 $EB SciPy-bundle-2021.05-foss-2021a.eb -r --buildpath /dev/shm/$USER/easybuild_build
 check_exit_code $? "${ok_msg}" "${fail_msg}"
 
+# CUDA support
+
+# install p7zip (to be able to unpack RPMs)
+p7zip_ec="p7zip-17.04-GCCcore-10.3.0.eb"
+echo ">> Installing $p7zip_ec..."
+ok_msg="$p7zip_ec installed, off to a good (?) start!"
+fail_msg="Failed to install $p7zip_ec, woopsie..."
+$EB $p7zip_ec --robot
+check_exit_code $? "${ok_msg}" "${fail_msg}"
+
+# install CUDA (uses eb_hooks.py to only install runtime)
+cuda_ec="CUDA-11.3.1.eb"
+echo ">> Installing $cuda_ec..."
+ok_msg="$cuda_ec installed, off to a good (?) start!"
+fail_msg="Failed to install $cuda_ec, woopsie..."
+$EB $cuda_ec --robot
+check_exit_code $? "${ok_msg}" "${fail_msg}"
+
+# install CUDA samples (requires EESSI support for CUDA)
+# TODO Run EESSI NVIDIA GPU support script here
+# (which unbreaks the symlinks from the runtime installation)
+cuda_samples_ec="CUDA-Samples-11.3-GCC-10.3.0-CUDA-11.3.1.eb"
+echo ">> Installing $cuda_samples_ec..."
+ok_msg="$cuda_ec installed, off to a good (?) start!"
+fail_msg="Failed to install $cuda_samples_ec, woopsie..."
+$EB $cuda_samples_ec --robot --from-pr=16914
+check_exit_code $? "${ok_msg}" "${fail_msg}"
+
 ### add packages here
 
 echo ">> Creating/updating Lmod cache..."
diff --git a/eb_hooks.py b/eb_hooks.py
index df7742f999..c5a0ca9cca 100644
--- a/eb_hooks.py
+++ b/eb_hooks.py
@@ -8,51 +8,56 @@
 from easybuild.tools.systemtools import AARCH64, POWER, X86_64, get_cpu_architecture, get_cpu_features
 from easybuild.tools.toolchain.compiler import OPTARCH_GENERIC
 
-EESSI_RPATH_OVERRIDE_ATTR = 'orig_rpath_override_dirs'
+EESSI_RPATH_OVERRIDE_ATTR = "orig_rpath_override_dirs"
+
+CUDA_ENABLED_TOOLCHAINS = [
+    "fosscuda",
+    "gcccuda",
+    "gimpic",
+    "giolfc",
+    "gmklc",
+    "golfc",
+    "gomklc",
+    "gompic",
+    "goolfc",
+    "iccifortcuda",
+    "iimklc",
+    "iimpic",
+    "intelcuda",
+    "iomklc",
+    "iompic",
+    "nvompic",
+    "nvpsmpic",
+]
 
+PARSE_HOOKS = {
+    "CGAL": cgal_toolchainopts_precise,
+    "fontconfig": fontconfig_add_fonts,
+    "UCX": ucx_eprefix,
+}
 
-def get_eessi_envvar(eessi_envvar):
-    """Get an EESSI environment variable from the environment"""
-
-    eessi_envvar_value = os.getenv(eessi_envvar)
-    if eessi_envvar_value is None:
-        raise EasyBuildError("$%s is not defined!", eessi_envvar)
-
-    return eessi_envvar_value
-
-
-def get_rpath_override_dirs(software_name):
-    # determine path to installations in software layer via $EESSI_SOFTWARE_PATH
-    eessi_software_path = get_eessi_envvar('EESSI_SOFTWARE_PATH')
-    eessi_pilot_version = get_eessi_envvar('EESSI_PILOT_VERSION')
-
-    # construct the rpath override directory stub
-    rpath_injection_stub = os.path.join(
-        # Make sure we are looking inside the `host_injections` directory
-        eessi_software_path.replace(eessi_pilot_version, os.path.join('host_injections', eessi_pilot_version), 1),
-        # Add the subdirectory for the specific software
-        'rpath_overrides',
-        software_name,
-        # We can't know the version, but this allows the use of a symlink
-        # to facilitate version upgrades without removing files
-        'system',
-    )
-
-    # Allow for libraries in lib or lib64
-    rpath_injection_dirs = [os.path.join(rpath_injection_stub, x) for x in ('lib', 'lib64')]
+PRE_CONFIGURE_HOOKS = {
+    "libfabric": libfabric_disable_psm3_x86_64_generic,
+    "MetaBAT": metabat_preconfigure,
+    "WRF": wrf_preconfigure,
+}
 
-    return rpath_injection_dirs
+POST_PACKAGE_HOOKS = {
+    "CUDA": cuda_postpackage,
+}
 
 
 def parse_hook(ec, *args, **kwargs):
     """Main parse hook: trigger custom functions based on software name."""
 
     # determine path to Prefix installation in compat layer via $EPREFIX
-    eprefix = get_eessi_envvar('EPREFIX')
+    eprefix = get_eessi_envvar("EPREFIX")
 
     if ec.name in PARSE_HOOKS:
         PARSE_HOOKS[ec.name](ec, eprefix)
 
+    ec = inject_gpu_property(ec)
+
 
 def pre_configure_hook(self, *args, **kwargs):
     """Main pre-configure hook: trigger custom functions based on software name."""
@@ -74,19 +79,21 @@ def pre_prepare_hook(self, *args, **kwargs):
 
         # update the relevant option (but keep the original value so we can reset it later)
         if hasattr(self, EESSI_RPATH_OVERRIDE_ATTR):
-            raise EasyBuildError("'self' already has attribute %s! Can't use pre_prepare hook.",
-                                 EESSI_RPATH_OVERRIDE_ATTR)
+            raise EasyBuildError(
+                "'self' already has attribute %s! Can't use pre_prepare hook.", EESSI_RPATH_OVERRIDE_ATTR
+            )
 
-        setattr(self, EESSI_RPATH_OVERRIDE_ATTR, build_option('rpath_override_dirs'))
+        setattr(self, EESSI_RPATH_OVERRIDE_ATTR, build_option("rpath_override_dirs"))
         if getattr(self, EESSI_RPATH_OVERRIDE_ATTR):
             # self.EESSI_RPATH_OVERRIDE_ATTR is (already) a colon separated string, let's make it a list
             orig_rpath_override_dirs = [getattr(self, EESSI_RPATH_OVERRIDE_ATTR)]
-            rpath_override_dirs = ':'.join(orig_rpath_override_dirs + mpi_rpath_override_dirs)
+            rpath_override_dirs = ":".join(orig_rpath_override_dirs + mpi_rpath_override_dirs)
         else:
-            rpath_override_dirs = ':'.join(mpi_rpath_override_dirs)
-        update_build_option('rpath_override_dirs', rpath_override_dirs)
-        print_msg("Updated rpath_override_dirs (to allow overriding MPI family %s): %s",
-                  mpi_family, rpath_override_dirs)
+            rpath_override_dirs = ":".join(mpi_rpath_override_dirs)
+        update_build_option("rpath_override_dirs", rpath_override_dirs)
+        print_msg(
+            "Updated rpath_override_dirs (to allow overriding MPI family %s): %s", mpi_family, rpath_override_dirs
+        )
 
 
 def post_prepare_hook(self, *args, **kwargs):
@@ -94,30 +101,78 @@ def post_prepare_hook(self, *args, **kwargs):
 
     if hasattr(self, EESSI_RPATH_OVERRIDE_ATTR):
         # Reset the value of 'rpath_override_dirs' now that we are finished with it
-        update_build_option('rpath_override_dirs', getattr(self, EESSI_RPATH_OVERRIDE_ATTR))
+        update_build_option("rpath_override_dirs", getattr(self, EESSI_RPATH_OVERRIDE_ATTR))
         print_msg("Resetting rpath_override_dirs to original value: %s", getattr(self, EESSI_RPATH_OVERRIDE_ATTR))
         delattr(self, EESSI_RPATH_OVERRIDE_ATTR)
 
 
+def pre_configure_hook(self, *args, **kwargs):
+    """Main pre-configure hook: trigger custom functions based on software name."""
+    if self.name in PRE_CONFIGURE_HOOKS:
+        PRE_CONFIGURE_HOOKS[self.name](self, *args, **kwargs)
+
+
+def post_package_hook(self, *args, **kwargs):
+    """Main post-package hook: trigger custom functions based on software name."""
+    if self.name in POST_PACKAGE_HOOKS:
+        POST_PACKAGE_HOOKS[self.name](self, *args, **kwargs)
+
+
+# Functions used by hooks
+
+
+def get_eessi_envvar(eessi_envvar):
+    """Get an EESSI environment variable from the environment"""
+
+    eessi_envvar_value = os.getenv(eessi_envvar)
+    if eessi_envvar_value is None:
+        raise EasyBuildError("$%s is not defined!", eessi_envvar)
+
+    return eessi_envvar_value
+
+
+def get_rpath_override_dirs(software_name):
+    # determine path to installations in software layer via $EESSI_SOFTWARE_PATH
+    eessi_software_path = get_eessi_envvar("EESSI_SOFTWARE_PATH")
+    eessi_pilot_version = get_eessi_envvar("EESSI_PILOT_VERSION")
+
+    # construct the rpath override directory stub
+    rpath_injection_stub = os.path.join(
+        # Make sure we are looking inside the `host_injections` directory
+        eessi_software_path.replace(eessi_pilot_version, os.path.join("host_injections", eessi_pilot_version), 1),
+        # Add the subdirectory for the specific software
+        "rpath_overrides",
+        software_name,
+        # We can't know the version, but this allows the use of a symlink
+        # to facilitate version upgrades without removing files
+        "system",
+    )
+
+    # Allow for libraries in lib or lib64
+    rpath_injection_dirs = [os.path.join(rpath_injection_stub, x) for x in ("lib", "lib64")]
+
+    return rpath_injection_dirs
+
+
 def cgal_toolchainopts_precise(ec, eprefix):
     """Enable 'precise' rather than 'strict' toolchain option for CGAL on POWER."""
-    if ec.name == 'CGAL':
+    if ec.name == "CGAL":
         if get_cpu_architecture() == POWER:
             # 'strict' implies '-mieee-fp', which is not supported on POWER
             # see https://github.com/easybuilders/easybuild-framework/issues/2077
-            ec['toolchainopts']['strict'] = False
-            ec['toolchainopts']['precise'] = True
-            print_msg("Tweaked toochainopts for %s: %s", ec.name, ec['toolchainopts'])
+            ec["toolchainopts"]["strict"] = False
+            ec["toolchainopts"]["precise"] = True
+            print_msg("Tweaked toochainopts for %s: %s", ec.name, ec["toolchainopts"])
     else:
         raise EasyBuildError("CGAL-specific hook triggered for non-CGAL easyconfig?!")
 
 
 def fontconfig_add_fonts(ec, eprefix):
     """Inject --with-add-fonts configure option for fontconfig."""
-    if ec.name == 'fontconfig':
+    if ec.name == "fontconfig":
         # make fontconfig aware of fonts included with compat layer
-        with_add_fonts = '--with-add-fonts=%s' % os.path.join(eprefix, 'usr', 'share', 'fonts')
-        ec.update('configopts', with_add_fonts)
+        with_add_fonts = "--with-add-fonts=%s" % os.path.join(eprefix, "usr", "share", "fonts")
+        ec.update("configopts", with_add_fonts)
         print_msg("Added '%s' configure option for %s", with_add_fonts, ec.name)
     else:
         raise EasyBuildError("fontconfig-specific hook triggered for non-fontconfig easyconfig?!")
@@ -125,29 +180,23 @@ def fontconfig_add_fonts(ec, eprefix):
 
 def ucx_eprefix(ec, eprefix):
     """Make UCX aware of compatibility layer via additional configuration options."""
-    if ec.name == 'UCX':
-        ec.update('configopts', '--with-sysroot=%s' % eprefix)
-        ec.update('configopts', '--with-rdmacm=%s' % os.path.join(eprefix, 'usr'))
-        print_msg("Using custom configure options for %s: %s", ec.name, ec['configopts'])
+    if ec.name == "UCX":
+        ec.update("configopts", "--with-sysroot=%s" % eprefix)
+        ec.update("configopts", "--with-rdmacm=%s" % os.path.join(eprefix, "usr"))
+        print_msg("Using custom configure options for %s: %s", ec.name, ec["configopts"])
     else:
         raise EasyBuildError("UCX-specific hook triggered for non-UCX easyconfig?!")
 
 
-def pre_configure_hook(self, *args, **kwargs):
-    """Main pre-configure hook: trigger custom functions based on software name."""
-    if self.name in PRE_CONFIGURE_HOOKS:
-        PRE_CONFIGURE_HOOKS[self.name](self, *args, **kwargs)
-
-
 def libfabric_disable_psm3_x86_64_generic(self, *args, **kwargs):
     """Add --disable-psm3 to libfabric configure options when building with --optarch=GENERIC on x86_64."""
-    if self.name == 'libfabric':
+    if self.name == "libfabric":
         if get_cpu_architecture() == X86_64:
-            generic = build_option('optarch') == OPTARCH_GENERIC
-            no_avx = 'avx' not in get_cpu_features()
+            generic = build_option("optarch") == OPTARCH_GENERIC
+            no_avx = "avx" not in get_cpu_features()
             if generic or no_avx:
-                self.cfg.update('configopts', '--disable-psm3')
-                print_msg("Using custom configure options for %s: %s", self.name, self.cfg['configopts'])
+                self.cfg.update("configopts", "--disable-psm3")
+                print_msg("Using custom configure options for %s: %s", self.name, self.cfg["configopts"])
     else:
         raise EasyBuildError("libfabric-specific hook triggered for non-libfabric easyconfig?!")
 
@@ -158,10 +207,10 @@ def metabat_preconfigure(self, *args, **kwargs):
     - take into account that zlib is a filtered dependency,
       and that there's no libz.a in the EESSI compat layer
     """
-    if self.name == 'MetaBAT':
-        configopts = self.cfg['configopts']
+    if self.name == "MetaBAT":
+        configopts = self.cfg["configopts"]
         regex = re.compile(r"\$EBROOTZLIB/lib/libz.a")
-        self.cfg['configopts'] = regex.sub('$EPREFIX/usr/lib64/libz.so', configopts)
+        self.cfg["configopts"] = regex.sub("$EPREFIX/usr/lib64/libz.so", configopts)
     else:
         raise EasyBuildError("MetaBAT-specific hook triggered for non-MetaBAT easyconfig?!")
 
@@ -171,24 +220,80 @@ def wrf_preconfigure(self, *args, **kwargs):
     Pre-configure hook for WRF:
     - patch arch/configure_new.defaults so building WRF with foss toolchain works on aarch64
     """
-    if self.name == 'WRF':
+    if self.name == "WRF":
         if get_cpu_architecture() == AARCH64:
             pattern = "Linux x86_64 ppc64le, gfortran"
             repl = "Linux x86_64 aarch64 ppc64le, gfortran"
-            self.cfg.update('preconfigopts', "sed -i 's/%s/%s/g' arch/configure_new.defaults && " % (pattern, repl))
-            print_msg("Using custom preconfigopts for %s: %s", self.name, self.cfg['preconfigopts'])
+            self.cfg.update("preconfigopts", "sed -i 's/%s/%s/g' arch/configure_new.defaults && " % (pattern, repl))
+            print_msg("Using custom preconfigopts for %s: %s", self.name, self.cfg["preconfigopts"])
     else:
         raise EasyBuildError("WRF-specific hook triggered for non-WRF easyconfig?!")
 
 
-PARSE_HOOKS = {
-    'CGAL': cgal_toolchainopts_precise,
-    'fontconfig': fontconfig_add_fonts,
-    'UCX': ucx_eprefix,
-}
-
-PRE_CONFIGURE_HOOKS = {
-    'libfabric': libfabric_disable_psm3_x86_64_generic,
-    'MetaBAT': metabat_preconfigure,
-    'WRF': wrf_preconfigure,
-}
+def cuda_post_package(self, *args, **kwargs):
+    """Delete CUDA files we are not allowed to ship and replace them with a symlink to a possible installation under host_injections."""
+    print_msg("Replacing CUDA stuff we cannot ship with symlinks...")
+    # read CUDA EULA
+    eula_path = os.path.join(self.installdir, "EULA.txt")
+    tmp_buffer = []
+    with open(eula_path) as infile:
+        copy = False
+        for line in infile:
+            if line.strip() == "2.6. Attachment A":
+                copy = True
+                continue
+            elif line.strip() == "2.7. Attachment B":
+                copy = False
+                continue
+            elif copy:
+                tmp_buffer.append(line)
+    # create whitelist without file extensions, they're not really needed and they only complicate things
+    whitelist = []
+    file_extensions = [".so", ".a", ".h", ".bc"]
+    for tmp in tmp_buffer:
+        for word in tmp.split():
+            if any(ext in word for ext in file_extensions):
+                whitelist.append(word.split(".")[0])
+    whitelist = list(set(whitelist))
+    # iterate over all files in the CUDA path
+    for root, dirs, files in os.walk(self.installdir):
+        for filename in files:
+            # we only really care about real files, i.e. not symlinks
+            if not os.path.islink(os.path.join(root, filename)):
+                # check if the current file is part of the whitelist
+                basename = filename.split(".")[0]
+                if basename not in whitelist:
+                    # if it is not in the whitelist, delete the file and create a symlink to host_injections
+                    source = os.path.join(root, filename)
+                    target = source.replace("versions", "host_injections")
+                    os.remove(source)
+                    # Using os.symlink requires the existence of the target directory, so we use os.system
+                    os.system("ln %s %s" % (target, source))
+
+
+def inject_gpu_property(ec):
+    ec_dict = ec.asdict()
+    # Check if CUDA is in the dependencies, if so add the GPU Lmod tag
+    if (
+        "CUDA" in [dep[0] for dep in iter(ec_dict["dependencies"])]
+        or ec_dict["toolchain"]["name"] in CUDA_ENABLED_TOOLCHAINS
+    ):
+        ec.log.info("[parse hook] Injecting gpu as Lmod arch property and envvar with CUDA version")
+        key = "modluafooter"
+        value = 'add_property("arch","gpu")'
+        cuda_version = 0
+        for dep in iter(ec_dict["dependencies"]):
+            # Make CUDA a build dependency only (rpathing saves us from link errors)
+            if "CUDA" in dep[0]:
+                cuda_version = dep[1]
+                ec_dict["dependencies"].remove(dep)
+                ec_dict["builddependencies"].append(dep) if dep not in ec_dict["builddependencies"] else ec_dict[
+                    "builddependencies"
+                ]
+        value = "\n".join([value, 'setenv("EESSICUDAVERSION","%s")' % cuda_version])
+        if key in ec_dict:
+            if not value in ec_dict[key]:
+                ec[key] = "\n".join([ec_dict[key], value])
+        else:
+            ec[key] = value
+    return ec
diff --git a/eessi-2021.12.yml b/eessi-2021.12.yml
index 977c0f9804..bbfadb9daa 100644
--- a/eessi-2021.12.yml
+++ b/eessi-2021.12.yml
@@ -1,4 +1,14 @@
 software:
+  CUDA:
+    toolchains:
+      SYSTEM:
+        versions: '11.3.1'
+  CUDA-Samples:
+    toolchains:
+      GCC-10.3.0:
+        versions:
+          '11.3':
+            versionsuffix: -CUDA-11.3.1  
   code-server:
     toolchains:
       SYSTEM:
@@ -29,6 +39,10 @@ software:
      toolchains:
        gompi-2020a:
          versions: ['5.6.3']
+  p7zip:
+    toolchains:
+      GCCcore-10.3.0:
+        versions: ['17.04']
   QuantumESPRESSO:
      toolchains:
        foss-2020a:

From 6c41b262d020ce818baa9aedc19c4322d2dee33d Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 16 Dec 2022 13:04:45 +0100
Subject: [PATCH 02/31] singularity install does not seem to install mksquashfs

---
 .github/workflows/tests_scripts.yml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/.github/workflows/tests_scripts.yml b/.github/workflows/tests_scripts.yml
index 18fcd7b255..24ce26ab61 100644
--- a/.github/workflows/tests_scripts.yml
+++ b/.github/workflows/tests_scripts.yml
@@ -35,6 +35,7 @@ jobs:
         curl -OL https://dl.fedoraproject.org/pub/epel/8/Everything/x86_64/Packages/s/$singularity_rpm
         sudo alien -d $singularity_rpm
         sudo apt install ./singularity*.deb
+        sudo apt install mksquashfs
         singularity --version
 
     - name: test install_software_layer.sh script

From 7d53b030d7acb343baab2866842a913d795fd363 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 16 Dec 2022 13:07:38 +0100
Subject: [PATCH 03/31] Trigger script test

---
 EESSI-pilot-install-software.sh | 1 +
 1 file changed, 1 insertion(+)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index 1698066bf4..16c7c48aa2 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -443,3 +443,4 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 echo ">> Cleaning up ${TMPDIR}..."
 rm -r ${TMPDIR}
+echo nothing

From 58357b9b6a405e3cc3d2f6e4bfc173e5f519021b Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 16 Dec 2022 13:11:24 +0100
Subject: [PATCH 04/31] Revert

---
 EESSI-pilot-install-software.sh | 1 -
 1 file changed, 1 deletion(-)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index 16c7c48aa2..1698066bf4 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -443,4 +443,3 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 echo ">> Cleaning up ${TMPDIR}..."
 rm -r ${TMPDIR}
-echo nothing

From 4b6654dcb5a9e9957784adc9c824fa8081842af7 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 16 Dec 2022 13:13:29 +0100
Subject: [PATCH 05/31] Use the right package name for squash-fs

---
 .github/workflows/tests_scripts.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/tests_scripts.yml b/.github/workflows/tests_scripts.yml
index 24ce26ab61..acacaa54a1 100644
--- a/.github/workflows/tests_scripts.yml
+++ b/.github/workflows/tests_scripts.yml
@@ -35,7 +35,7 @@ jobs:
         curl -OL https://dl.fedoraproject.org/pub/epel/8/Everything/x86_64/Packages/s/$singularity_rpm
         sudo alien -d $singularity_rpm
         sudo apt install ./singularity*.deb
-        sudo apt install mksquashfs
+        sudo apt install squashfs-tools
         singularity --version
 
     - name: test install_software_layer.sh script

From 33ce58497b3cc5beff81b5d8f0265d997a0b4b15 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 16 Dec 2022 14:14:09 +0100
Subject: [PATCH 06/31] Tidy up hooks

---
 eb_hooks.py | 34 +++++++++++++++++-----------------
 1 file changed, 17 insertions(+), 17 deletions(-)

diff --git a/eb_hooks.py b/eb_hooks.py
index c5a0ca9cca..3ab991771f 100644
--- a/eb_hooks.py
+++ b/eb_hooks.py
@@ -30,22 +30,6 @@
     "nvpsmpic",
 ]
 
-PARSE_HOOKS = {
-    "CGAL": cgal_toolchainopts_precise,
-    "fontconfig": fontconfig_add_fonts,
-    "UCX": ucx_eprefix,
-}
-
-PRE_CONFIGURE_HOOKS = {
-    "libfabric": libfabric_disable_psm3_x86_64_generic,
-    "MetaBAT": metabat_preconfigure,
-    "WRF": wrf_preconfigure,
-}
-
-POST_PACKAGE_HOOKS = {
-    "CUDA": cuda_postpackage,
-}
-
 
 def parse_hook(ec, *args, **kwargs):
     """Main parse hook: trigger custom functions based on software name."""
@@ -230,7 +214,7 @@ def wrf_preconfigure(self, *args, **kwargs):
         raise EasyBuildError("WRF-specific hook triggered for non-WRF easyconfig?!")
 
 
-def cuda_post_package(self, *args, **kwargs):
+def cuda_postpackage(self, *args, **kwargs):
     """Delete CUDA files we are not allowed to ship and replace them with a symlink to a possible installation under host_injections."""
     print_msg("Replacing CUDA stuff we cannot ship with symlinks...")
     # read CUDA EULA
@@ -297,3 +281,19 @@ def inject_gpu_property(ec):
         else:
             ec[key] = value
     return ec
+
+PARSE_HOOKS = {
+    "CGAL": cgal_toolchainopts_precise,
+    "fontconfig": fontconfig_add_fonts,
+    "UCX": ucx_eprefix,
+}
+
+PRE_CONFIGURE_HOOKS = {
+    "libfabric": libfabric_disable_psm3_x86_64_generic,
+    "MetaBAT": metabat_preconfigure,
+    "WRF": wrf_preconfigure,
+}
+
+POST_PACKAGE_HOOKS = {
+    "CUDA": cuda_postpackage,
+}

From f1cd893823227a934e5537bba50cc6b5e9358502 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 16 Dec 2022 15:25:32 +0100
Subject: [PATCH 07/31] Force creation of links

---
 eb_hooks.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/eb_hooks.py b/eb_hooks.py
index 3ab991771f..ad55573660 100644
--- a/eb_hooks.py
+++ b/eb_hooks.py
@@ -252,7 +252,7 @@ def cuda_postpackage(self, *args, **kwargs):
                     target = source.replace("versions", "host_injections")
                     os.remove(source)
                     # Using os.symlink requires the existence of the target directory, so we use os.system
-                    os.system("ln %s %s" % (target, source))
+                    os.system("ln -s %s %s" % (target, source))
 
 
 def inject_gpu_property(ec):

From 06a9eaf5eb4118528ac130be31c9b1af26ca4306 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Mon, 19 Dec 2022 23:06:44 +0100
Subject: [PATCH 08/31] Install host_injections CUDA

---
 EESSI-pilot-install-software.sh |  3 ++
 eb_hooks.py                     |  6 ++--
 install_cuda_host_injections.sh | 53 +++++++++++++++++++++++++++++++++
 3 files changed, 60 insertions(+), 2 deletions(-)
 create mode 100755 install_cuda_host_injections.sh

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index 1698066bf4..94b71e14c9 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -412,6 +412,9 @@ fail_msg="Failed to install $cuda_ec, woopsie..."
 $EB $cuda_ec --robot
 check_exit_code $? "${ok_msg}" "${fail_msg}"
 
+# Add the host_injections CUDA so we can actually build CUDA apps
+./install_cuda_host_injections.sh 11.3.1
+
 # install CUDA samples (requires EESSI support for CUDA)
 # TODO Run EESSI NVIDIA GPU support script here
 # (which unbreaks the symlinks from the runtime installation)
diff --git a/eb_hooks.py b/eb_hooks.py
index ad55573660..4d31a5f4b4 100644
--- a/eb_hooks.py
+++ b/eb_hooks.py
@@ -232,7 +232,7 @@ def cuda_postpackage(self, *args, **kwargs):
             elif copy:
                 tmp_buffer.append(line)
     # create whitelist without file extensions, they're not really needed and they only complicate things
-    whitelist = []
+    whitelist = ['eula']
     file_extensions = [".so", ".a", ".h", ".bc"]
     for tmp in tmp_buffer:
         for word in tmp.split():
@@ -252,7 +252,9 @@ def cuda_postpackage(self, *args, **kwargs):
                     target = source.replace("versions", "host_injections")
                     os.remove(source)
                     # Using os.symlink requires the existence of the target directory, so we use os.system
-                    os.system("ln -s %s %s" % (target, source))
+                    system_command="ln -s %s %s" % (target, source)
+                    if os.system(system_command) != 0:
+                        raise EasyBuildError("Failed to create symbolic link: %s" % system_command)
 
 
 def inject_gpu_property(ec):
diff --git a/install_cuda_host_injections.sh b/install_cuda_host_injections.sh
new file mode 100755
index 0000000000..038ff19ff3
--- /dev/null
+++ b/install_cuda_host_injections.sh
@@ -0,0 +1,53 @@
+#!/bin/bash
+
+if [[ $# -eq 0 ]] ; then
+    echo 'You must provide the CUDA version as an argument, e.g.:'
+    echo "  $0 11.3.1"
+    exit 1
+fi
+install_cuda_version=$1
+if [[ -z "${EESSI_SOFTWARE_PATH}" ]]; then
+  echo "This script cannot be used without having first defined EESSI_SOFTWARE_PATH"
+  exit 1
+else
+  # As an installation location just use $EESSI_SOFTWARE_PATH but replacing `versions` with `host_injections`
+  # (CUDA is a binary installation so no need to worry too much about the EasyBuild setup)
+  cuda_install_dir=${EESSI_SOFTWARE_PATH/versions/host_injections}
+fi
+
+# Only install CUDA if specified version is not found.
+# This is only relevant for users, the shipped CUDA installation will
+# always be in versions instead of host_injections and have symlinks pointing
+# to host_injections for everything we're not allowed to ship
+# (existence of easybuild subdir implies a successful install)
+if [ -d ${cuda_install_dir}/software/CUDA/${install_cuda_version}/easybuild ]; then
+  echo "CUDA software found! No need to install CUDA again, proceed with testing."
+else
+  # The install is pretty fat, you need lots of space for download/unpack/install (~3*5GB), need to do a space check before we proceed
+  avail_space=$(df --output=avail ${cuda_install_dir}/ | tail -n 1 | awk '{print $1}')
+  if (( ${avail_space} < 16000000 )); then
+    echo "Need more disk space to install CUDA, exiting now..."
+    exit 1
+  fi
+  if [[ ! -z "${EBROOTEASYBUILD}" ]]; then
+    echo "Loading EasyBuild module to do actual install"
+    module load EasyBuild
+  fi
+  # we need the --rebuild option and a random dir for the module if the module file is shipped with EESSI
+  if [ -f ${EESSI_SOFTWARE_PATH}/modules/all/CUDA/${install_cuda_version}.lua ]; then
+    tmpdir=$(mktemp -d)
+    extra_args="--rebuild --installpath-modules=${tmpdir}"
+  fi
+  # We don't want hooks used in this install, we need a vanilla CUDA installation
+  touch $tmpdir/none.py
+  eb ${extra_args} --hooks=$tmpdir/none.py --installpath=${cuda_install_dir}/ CUDA-${install_cuda_version}.eb
+  ret=$?
+  if [ $ret -ne 0 ]; then
+    echo "CUDA installation failed, please check EasyBuild logs..."
+    exit 1
+  fi
+  # clean up tmpdir if it exists
+  if [ -f ${EESSI_SOFTWARE_PATH}/modules/all/CUDA/${install_cuda_version}.lua ]; then
+    rm -rf ${tmpdir}
+  fi
+fi

From b4e80a153390be28dda32fba288f780a5899d430 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Mon, 19 Dec 2022 23:47:06 +0100
Subject: [PATCH 09/31] Move comments to the right place

---
 EESSI-pilot-install-software.sh | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index 94b71e14c9..b7752d3e83 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -413,11 +413,10 @@ $EB $cuda_ec --robot
 check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 # Add the host_injections CUDA so we can actually build CUDA apps
+# (which unbreaks the symlinks from the runtime installation)
 ./install_cuda_host_injections.sh 11.3.1
 
 # install CUDA samples (requires EESSI support for CUDA)
-# TODO Run EESSI NVIDIA GPU support script here
-# (which unbreaks the symlinks from the runtime installation)
 cuda_samples_ec="CUDA-Samples-11.3-GCC-10.3.0-CUDA-11.3.1.eb"
 echo ">> Installing $cuda_samples_ec..."
 ok_msg="$cuda_ec installed, off to a good (?) start!"

From 2c8697330155ca91a3ea78f2e3de4d46134877cf Mon Sep 17 00:00:00 2001
From: ocaisa <alan.ocais@cecam.org>
Date: Tue, 14 Feb 2023 17:11:18 +0100
Subject: [PATCH 10/31] Reimplement `mkdir -p` reporting where permissions
 break down

This will allow us to log where creating directory structures under `host_injections` is breaking down.
---
 scripts/utils.sh | 68 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 68 insertions(+)

diff --git a/scripts/utils.sh b/scripts/utils.sh
index 5d8455bb68..06fc897912 100644
--- a/scripts/utils.sh
+++ b/scripts/utils.sh
@@ -31,3 +31,71 @@ function check_exit_code {
         fatal_error "${fail_msg}"
     fi
 }
+
+# Reimplement 'mkdir -p' with reporting on where permissions break down
+function create_directory_structure() {
+  # Ensure we are given a single path argument
+  if [ $# -ne 1 ]; then
+    echo "Function requires a single (relative or absolute) path argument" >&2
+    return 1
+  fi
+
+  # set a persistent variable that knows the full structure
+  # (i.e., retains the value upon recursive calls)
+  full_structure="${full_structure:="$1"}"
+
+  local directory_structure="$1"
+
+  # Check if directory exists and is writeable
+  if [ -d "${directory_structure}" ]; then
+    if [ "${directory_structure}" = "${full_structure}" ]; then
+        # release our (unneeded) global variable
+        unset full_structure
+    fi
+    if [ -w "${directory_structure}" ]; then
+      # Nothing to be done
+      return 0
+    else
+      echo "Directory ${directory_structure} exists but is not writeable" >&2
+      return 1
+    fi
+  fi
+
+  local directory_structure_parent=$(dirname "${directory_structure}")
+
+  # If the parent doesn't exist we need to create it
+  if [ ! -d "${directory_structure_parent}" ]; then
+    # Create the parent via a recursive call to this function
+    # (if this doesn't succeed we need to return the error code)
+    if ! create_directory_structure "${directory_structure_parent}"; then
+      if [ "${directory_structure}" = "${full_structure}" ]; then
+        # release our (unneeded) global variable
+        unset full_structure
+      fi
+      return 1
+    fi
+  fi
+
+  # Check the parent is writeable, and create the new subdir
+  if [ -w "${directory_structure_parent}" ]; then
+    if [ "${directory_structure}" = "${full_structure}" ]; then
+      # release our (unneeded) global variable
+      unset full_structure
+    fi
+    if ! mkdir "${directory_structure}"; then
+      echo "'mkdir ${directory_structure}' failed for an unknown reason!" >&2
+      return 1
+    else
+      # Success!
+      return 0
+    fi
+  else
+    echo "Attempt to create ${full_structure} failed," \
+      "${directory_structure_parent} exists but you don't have write permissions." >&2
+    if [ "${directory_structure}" = "${full_structure}" ]; then
+      # release our global variable
+      unset full_structure
+    fi
+    return 1
+  fi
+}

From 9590047d008f3be05fc809b6b4b3bbd53c1810cb Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 24 Feb 2023 15:47:34 +0100
Subject: [PATCH 11/31] Be more agressive on catching errors

---
 EESSI-pilot-install-software.sh | 14 ++++++
 install_cuda_host_injections.sh | 86 +++++++++++++++++++++++----------
 2 files changed, 74 insertions(+), 26 deletions(-)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index 2d1132efdb..d702e96778 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -431,6 +431,16 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 # CUDA support
 
+# Need recent version of EasyBuild
+echo ">> Installing EasyBuild 4.5.1..."
+ok_msg="EasyBuild v4.7.0 installed"
+fail_msg="EasyBuild v4.7.0 failed to install"
+$EB --from-pr 17065 --include-easyblocks-from-pr 2893
+check_exit_code $? "${ok_msg}" "${fail_msg}"
+
+LMOD_IGNORE_CACHE=1 module swap EasyBuild/4.7.0
+check_exit_code $? "Swapped to EasyBuild/4.7.0" "Couldn't swap to EasyBuild/4.7.0"
+
 # install p7zip (to be able to unpack RPMs)
 p7zip_ec="p7zip-17.04-GCCcore-10.3.0.eb"
 echo ">> Installing $p7zip_ec..."
@@ -450,6 +460,10 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 # Add the host_injections CUDA so we can actually build CUDA apps
 # (which unbreaks the symlinks from the runtime installation)
 ./install_cuda_host_injections.sh 11.3.1
+echo ">> Installing $cuda_ec under host_injections..."
+ok_msg="$cuda_ec (re)installed under host_injections!"
+fail_msg="Failed to install $cuda_ec under host_injections, woopsie..."
+check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 # install CUDA samples (requires EESSI support for CUDA)
 cuda_samples_ec="CUDA-Samples-11.3-GCC-10.3.0-CUDA-11.3.1.eb"
diff --git a/install_cuda_host_injections.sh b/install_cuda_host_injections.sh
index 038ff19ff3..59c1d72996 100755
--- a/install_cuda_host_injections.sh
+++ b/install_cuda_host_injections.sh
@@ -1,18 +1,18 @@
 #!/bin/bash
 
+# Initialise our bash functions
+source scripts/utils.sh
+
 if [[ $# -eq 0 ]] ; then
-    echo 'You must provide the CUDA version as an argument, e.g.:'
-    echo "  $0 11.3.1"
-    exit 1
+    fatal_error "You must provide the CUDA version as an argument, e.g.:\n $0 11.3.1"
 fi
 install_cuda_version=$1
 if [[ -z "${EESSI_SOFTWARE_PATH}" ]]; then
-  echo "This script cannot be used without having first defined EESSI_SOFTWARE_PATH"
-  exit 1
+  fatal_error "This script cannot be used without having first defined EESSI_SOFTWARE_PATH"
 else
   # As an installation location just use $EESSI_SOFTWARE_PATH but replacing `versions` with `host_injections`
   # (CUDA is a binary installation so no need to worry too much about the EasyBuild setup)
-  cuda_install_dir=${EESSI_SOFTWARE_PATH/versions/host_injections}
+  cuda_install_parent=${EESSI_SOFTWARE_PATH/versions/host_injections}
 fi
 
 # Only install CUDA if specified version is not found.
@@ -20,34 +20,68 @@ fi
 # always be in versions instead of host_injections and have symlinks pointing
 # to host_injections for everything we're not allowed to ship
 # (existence of easybuild subdir implies a successful install)
-if [ -d ${cuda_install_dir}/software/CUDA/${install_cuda_version}/easybuild ]; then
-  echo "CUDA software found! No need to install CUDA again, proceed with testing."
+if [ -d "${cuda_install_parent}"/software/CUDA/"${install_cuda_version}"/easybuild ]; then
+  echo_green "CUDA software found! No need to install CUDA again, proceed with testing."
 else
-  # The install is pretty fat, you need lots of space for download/unpack/install (~3*5GB), need to do a space check before we proceed
-  avail_space=$(df --output=avail ${cuda_install_dir}/ | tail -n 1 | awk '{print $1}')
-  if (( ${avail_space} < 16000000 )); then
-    echo "Need more disk space to install CUDA, exiting now..."
-    exit 1
-  fi
-  if [[ ! -z "${EBROOTEASYBUILD}" ]]; then
-    echo "Loading EasyBuild module to do actual install"
+  # We need to be able write to the installation space so let's make sure we can
+  if ! create_directory_structure "${cuda_install_parent}"/software/CUDA ; then
+    fatal_error "No write permissions to directory ${cuda_install_parent}/software/CUDA"
+  fi
+
+  # we need a directory we can use for temporary storage
+  if [[ -z "${CUDA_TEMP_DIR}" ]]; then
+    tmpdir=$(mktemp -d)
+  else
+    tmpdir="${CUDA_TEMP_DIR}"/temp
+    if ! mkdir "$tmpdir" ; then
+      fatal_error "Could not create directory ${tmpdir}"
+    fi
+  fi
+
+
+  required_space_in_tmpdir=50000
+  # Let's see if we have sources and build locations defined if not, we use the temporary space
+  if [[ -z "${EASYBUILD_BUILDPATH}" ]]; then
+    export EASYBUILD_BUILDPATH=${tmpdir}/build
+    required_space_in_tmpdir=$((required_space_in_tempdir + 5000000))
+  fi
+  if [[ -z "${EASYBUILD_SOURCEPATH}" ]]; then
+    export EASYBUILD_SOURCEPATH=${tmpdir}/sources
+    required_space_in_tmpdir=$((required_space_in_tempdir + 5000000))
+  fi
+
+  # The install is pretty fat, you need lots of space for download/unpack/install (~3*5GB),
+  # need to do a space check before we proceed
+  avail_space=$(df --output=avail "${tmpdir}"/ | tail -n 1 | awk '{print $1}')
+  if (( avail_space < 5000000 )); then
+    fatal_error "Need at least 5GB disk space to install CUDA under ${cuda_install_parent}, exiting now..."
+  fi
+  avail_space=$(df --output=avail "${cuda_install_parent}"/ | tail -n 1 | awk '{print $1}')
+  if (( avail_space < required_space_in_tmpdir )); then
+    error1="Need at least ${required_space_in_tmpdir} disk space under ${tmpdir}.\n"
+    error2="You can set EASYBUILD_BUILDPATH and/or EASYBUILD_SOURCEPATH to reduce this requirement. Exiting now..."
+    fatal_error "${error1}${error2}"
+  fi
+
+  if [[ -z "${EBROOTEASYBUILD}" ]]; then
+    echo_yellow "Loading EasyBuild module to do actual install"
     module load EasyBuild
   fi
-  # we need the --rebuild option and a random dir for the module if the module file is shipped with EESSI
+
+  # we need the --rebuild option and a (random) dir for the module if the module
+  # file is shipped with EESSI
   if [ -f ${EESSI_SOFTWARE_PATH}/modules/all/CUDA/${install_cuda_version}.lua ]; then
-    tmpdir=$(mktemp -d)
     extra_args="--rebuild --installpath-modules=${tmpdir}"
   fi
   # We don't want hooks used in this install, we need a vanilla CUDA installation
-  touch $tmpdir/none.py
-  eb ${extra_args} --hooks=$tmpdir/none.py --installpath=${cuda_install_dir}/ CUDA-${install_cuda_version}.eb
+  touch "$tmpdir"/none.py
+  eb "${extra_args}" --hooks="$tmpdir"/none.py --installpath="${cuda_install_parent}"/ CUDA-"${install_cuda_version}".eb
   ret=$?
   if [ $ret -ne 0 ]; then
-    echo "CUDA installation failed, please check EasyBuild logs..."
-    exit 1
-  fi
-  # clean up tmpdir if it exists
-  if [ -f ${EESSI_SOFTWARE_PATH}/modules/all/CUDA/${install_cuda_version}.lua ]; then
-    rm -rf ${tmpdir}
+    fatal_error  "CUDA installation failed, please check EasyBuild logs..."
+  else
+    echo_green "CUDA installation at ${cuda_install_parent}/software/CUDA/${install_cuda_version} succeeded!"
   fi
+  # clean up tmpdir
+  rm -rf "${tmpdir}"
 fi

From 1357f763b5f27a0be82ed6c443ce3213a7e8361e Mon Sep 17 00:00:00 2001
From: ocaisa <alan.ocais@cecam.org>
Date: Mon, 27 Feb 2023 16:43:08 +0100
Subject: [PATCH 12/31] `${extra_args}` is actually multiple args not a single
 string

---
 install_cuda_host_injections.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/install_cuda_host_injections.sh b/install_cuda_host_injections.sh
index 59c1d72996..b6f83b4e9d 100755
--- a/install_cuda_host_injections.sh
+++ b/install_cuda_host_injections.sh
@@ -75,7 +75,7 @@ else
   fi
   # We don't want hooks used in this install, we need a vanilla CUDA installation
   touch "$tmpdir"/none.py
-  eb "${extra_args}" --hooks="$tmpdir"/none.py --installpath="${cuda_install_parent}"/ CUDA-"${install_cuda_version}".eb
+  eb ${extra_args} --hooks="$tmpdir"/none.py --installpath="${cuda_install_parent}"/ CUDA-"${install_cuda_version}".eb
   ret=$?
   if [ $ret -ne 0 ]; then
     fatal_error  "CUDA installation failed, please check EasyBuild logs..."

From 8096c5460c583f88d83bb8f2b00f204fb8830d3c Mon Sep 17 00:00:00 2001
From: ocaisa <alan.ocais@cecam.org>
Date: Mon, 27 Feb 2023 16:46:33 +0100
Subject: [PATCH 13/31] Update EESSI-pilot-install-software.sh

---
 EESSI-pilot-install-software.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index d702e96778..5b604deaa7 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -468,7 +468,7 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 # install CUDA samples (requires EESSI support for CUDA)
 cuda_samples_ec="CUDA-Samples-11.3-GCC-10.3.0-CUDA-11.3.1.eb"
 echo ">> Installing $cuda_samples_ec..."
-ok_msg="$cuda_ec installed, off to a good (?) start!"
+ok_msg="$cuda_samples_ec installed, off to a good (?) start!"
 fail_msg="Failed to install $cuda_samples_ec, woopsie..."
 $EB $cuda_samples_ec --robot --from-pr=16914
 check_exit_code $? "${ok_msg}" "${fail_msg}"

From ec31edfbb1954982f197cd44cd296a62e2f47954 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Wed, 1 Mar 2023 14:46:56 +0100
Subject: [PATCH 14/31] Catching echo exit code instead of actual code

---
 EESSI-pilot-install-software.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index 5b604deaa7..a9f6576866 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -459,8 +459,8 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 # Add the host_injections CUDA so we can actually build CUDA apps
 # (which unbreaks the symlinks from the runtime installation)
-./install_cuda_host_injections.sh 11.3.1
 echo ">> Installing $cuda_ec under host_injections..."
+./install_cuda_host_injections.sh 11.3.1
 ok_msg="$cuda_ec (re)installed under host_injections!"
 fail_msg="Failed to install $cuda_ec under host_injections, woopsie..."
 check_exit_code $? "${ok_msg}" "${fail_msg}"

From 0e99db5127e2735828d22a415aeeb194216f1378 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Wed, 1 Mar 2023 14:48:52 +0100
Subject: [PATCH 15/31] Give a full path to the CUDA host injections script

---
 EESSI-pilot-install-software.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index a9f6576866..c715dd55ad 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -460,7 +460,7 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 # Add the host_injections CUDA so we can actually build CUDA apps
 # (which unbreaks the symlinks from the runtime installation)
 echo ">> Installing $cuda_ec under host_injections..."
-./install_cuda_host_injections.sh 11.3.1
+$(dirname "$BASH_SOURCE")/install_cuda_host_injections.sh 11.3.1
 ok_msg="$cuda_ec (re)installed under host_injections!"
 fail_msg="Failed to install $cuda_ec under host_injections, woopsie..."
 check_exit_code $? "${ok_msg}" "${fail_msg}"

From cd11792609b4a637f324748089da61f1456827f5 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Wed, 1 Mar 2023 15:29:02 +0100
Subject: [PATCH 16/31] Add checks for some whitelist entries for CUDA

---
 eb_hooks.py | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/eb_hooks.py b/eb_hooks.py
index 4d31a5f4b4..c7358d5f13 100644
--- a/eb_hooks.py
+++ b/eb_hooks.py
@@ -232,13 +232,18 @@ def cuda_postpackage(self, *args, **kwargs):
             elif copy:
                 tmp_buffer.append(line)
     # create whitelist without file extensions, they're not really needed and they only complicate things
-    whitelist = ['eula']
+    whitelist = ['EULA']
     file_extensions = [".so", ".a", ".h", ".bc"]
     for tmp in tmp_buffer:
         for word in tmp.split():
             if any(ext in word for ext in file_extensions):
                 whitelist.append(word.split(".")[0])
     whitelist = list(set(whitelist))
+    # Do some quick checks for things we should or shouldn't have in the list
+    if "nvcc" in whitelist:
+        raise EasyBuildError("Found 'nvcc' in whitelist: %s" % whitelist)
+    if "libcudart" not in whitelist:
+        raise EasyBuildError("Did not find 'libcudart' in whitelist: %s" % whitelist)
     # iterate over all files in the CUDA path
     for root, dirs, files in os.walk(self.installdir):
         for filename in files:

From f514f8155fd6f02e894fc48e4a7a83d046b08222 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Wed, 1 Mar 2023 15:56:24 +0100
Subject: [PATCH 17/31] Fix failing eb installation

---
 EESSI-pilot-install-software.sh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index c715dd55ad..f97894df74 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -432,10 +432,10 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 # CUDA support
 
 # Need recent version of EasyBuild
-echo ">> Installing EasyBuild 4.5.1..."
+echo ">> Installing EasyBuild 4.7.0..."
 ok_msg="EasyBuild v4.7.0 installed"
 fail_msg="EasyBuild v4.7.0 failed to install"
-$EB --from-pr 17065 --include-easyblocks-from-pr 2893
+$EB --from-pr 17065 --include-easyblocks-from-pr 2893 --try-amend=use_pip=1
 check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 LMOD_IGNORE_CACHE=1 module swap EasyBuild/4.7.0

From be326a1e5842460035ad6dffc580c73395a4c8b2 Mon Sep 17 00:00:00 2001
From: ocaisa <alan.ocais@cecam.org>
Date: Wed, 1 Mar 2023 21:02:01 +0100
Subject: [PATCH 18/31] Make sure we check space in the right places

---
 install_cuda_host_injections.sh | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/install_cuda_host_injections.sh b/install_cuda_host_injections.sh
index b6f83b4e9d..0684c7ca80 100755
--- a/install_cuda_host_injections.sh
+++ b/install_cuda_host_injections.sh
@@ -38,7 +38,6 @@ else
     fi
   fi
 
-
   required_space_in_tmpdir=50000
   # Let's see if we have sources and build locations defined if not, we use the temporary space
   if [[ -z "${EASYBUILD_BUILDPATH}" ]]; then
@@ -52,11 +51,11 @@ else
 
   # The install is pretty fat, you need lots of space for download/unpack/install (~3*5GB),
   # need to do a space check before we proceed
-  avail_space=$(df --output=avail "${tmpdir}"/ | tail -n 1 | awk '{print $1}')
+  avail_space=$(df --output=avail "${cuda_install_parent}"/ | tail -n 1 | awk '{print $1}')
   if (( avail_space < 5000000 )); then
     fatal_error "Need at least 5GB disk space to install CUDA under ${cuda_install_parent}, exiting now..."
   fi
-  avail_space=$(df --output=avail "${cuda_install_parent}"/ | tail -n 1 | awk '{print $1}')
+  avail_space=$(df --output=avail "${tmpdir}"/ | tail -n 1 | awk '{print $1}')
   if (( avail_space < required_space_in_tmpdir )); then
     error1="Need at least ${required_space_in_tmpdir} disk space under ${tmpdir}.\n"
     error2="You can set EASYBUILD_BUILDPATH and/or EASYBUILD_SOURCEPATH to reduce this requirement. Exiting now..."

From 103f5fa34ea9c4cc740088d25a6d4ce6b6ca78bb Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 09:38:12 +0100
Subject: [PATCH 19/31] Simply wrap `mkdir -p` for better error reporting

---
 EESSI-pilot-install-software.sh |  4 +-
 install_cuda_host_injections.sh | 87 ---------------------------------
 scripts/utils.sh                | 82 ++++++-------------------------
 3 files changed, 19 insertions(+), 154 deletions(-)
 delete mode 100755 install_cuda_host_injections.sh

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index f97894df74..d648162c31 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -431,6 +431,8 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 # CUDA support
 
+cuda_version="11.3.1"
+
 # Need recent version of EasyBuild
 echo ">> Installing EasyBuild 4.7.0..."
 ok_msg="EasyBuild v4.7.0 installed"
@@ -460,7 +462,7 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 # Add the host_injections CUDA so we can actually build CUDA apps
 # (which unbreaks the symlinks from the runtime installation)
 echo ">> Installing $cuda_ec under host_injections..."
-$(dirname "$BASH_SOURCE")/install_cuda_host_injections.sh 11.3.1
+$(dirname "$BASH_SOURCE")/gpu_support/cuda_utils/install_cuda_host_injections.sh 11.3.1
 ok_msg="$cuda_ec (re)installed under host_injections!"
 fail_msg="Failed to install $cuda_ec under host_injections, woopsie..."
 check_exit_code $? "${ok_msg}" "${fail_msg}"
diff --git a/install_cuda_host_injections.sh b/install_cuda_host_injections.sh
deleted file mode 100755
index b6f83b4e9d..0000000000
--- a/install_cuda_host_injections.sh
+++ /dev/null
@@ -1,87 +0,0 @@
-#!/bin/bash
-
-# Initialise our bash functions
-source scripts/utils.sh
-
-if [[ $# -eq 0 ]] ; then
-    fatal_error "You must provide the CUDA version as an argument, e.g.:\n $0 11.3.1"
-fi
-install_cuda_version=$1
-if [[ -z "${EESSI_SOFTWARE_PATH}" ]]; then
-  fatal_error "This script cannot be used without having first defined EESSI_SOFTWARE_PATH"
-else
-  # As an installation location just use $EESSI_SOFTWARE_PATH but replacing `versions` with `host_injections`
-  # (CUDA is a binary installation so no need to worry too much about the EasyBuild setup)
-  cuda_install_parent=${EESSI_SOFTWARE_PATH/versions/host_injections}
-fi
-
-# Only install CUDA if specified version is not found.
-# This is only relevant for users, the shipped CUDA installation will
-# always be in versions instead of host_injections and have symlinks pointing
-# to host_injections for everything we're not allowed to ship
-# (existence of easybuild subdir implies a successful install)
-if [ -d "${cuda_install_parent}"/software/CUDA/"${install_cuda_version}"/easybuild ]; then
-  echo_green "CUDA software found! No need to install CUDA again, proceed with testing."
-else
-  # We need to be able write to the installation space so let's make sure we can
-  if ! create_directory_structure "${cuda_install_parent}"/software/CUDA ; then
-    fatal_error "No write permissions to directory ${cuda_install_parent}/software/CUDA"
-  fi
-
-  # we need a directory we can use for temporary storage
-  if [[ -z "${CUDA_TEMP_DIR}" ]]; then
-    tmpdir=$(mktemp -d)
-  else
-    tmpdir="${CUDA_TEMP_DIR}"/temp
-    if ! mkdir "$tmpdir" ; then
-      fatal_error "Could not create directory ${tmpdir}"
-    fi
-  fi
-
-
-  required_space_in_tmpdir=50000
-  # Let's see if we have sources and build locations defined if not, we use the temporary space
-  if [[ -z "${EASYBUILD_BUILDPATH}" ]]; then
-    export EASYBUILD_BUILDPATH=${tmpdir}/build
-    required_space_in_tmpdir=$((required_space_in_tempdir + 5000000))
-  fi
-  if [[ -z "${EASYBUILD_SOURCEPATH}" ]]; then
-    export EASYBUILD_SOURCEPATH=${tmpdir}/sources
-    required_space_in_tmpdir=$((required_space_in_tempdir + 5000000))
-  fi
-
-  # The install is pretty fat, you need lots of space for download/unpack/install (~3*5GB),
-  # need to do a space check before we proceed
-  avail_space=$(df --output=avail "${tmpdir}"/ | tail -n 1 | awk '{print $1}')
-  if (( avail_space < 5000000 )); then
-    fatal_error "Need at least 5GB disk space to install CUDA under ${cuda_install_parent}, exiting now..."
-  fi
-  avail_space=$(df --output=avail "${cuda_install_parent}"/ | tail -n 1 | awk '{print $1}')
-  if (( avail_space < required_space_in_tmpdir )); then
-    error1="Need at least ${required_space_in_tmpdir} disk space under ${tmpdir}.\n"
-    error2="You can set EASYBUILD_BUILDPATH and/or EASYBUILD_SOURCEPATH to reduce this requirement. Exiting now..."
-    fatal_error "${error1}${error2}"
-  fi
-
-  if [[ -z "${EBROOTEASYBUILD}" ]]; then
-    echo_yellow "Loading EasyBuild module to do actual install"
-    module load EasyBuild
-  fi
-
-  # we need the --rebuild option and a (random) dir for the module if the module
-  # file is shipped with EESSI
-  if [ -f ${EESSI_SOFTWARE_PATH}/modules/all/CUDA/${install_cuda_version}.lua ]; then
-    extra_args="--rebuild --installpath-modules=${tmpdir}"
-  fi
-  # We don't want hooks used in this install, we need a vanilla CUDA installation
-  touch "$tmpdir"/none.py
-  eb ${extra_args} --hooks="$tmpdir"/none.py --installpath="${cuda_install_parent}"/ CUDA-"${install_cuda_version}".eb
-  ret=$?
-  if [ $ret -ne 0 ]; then
-    fatal_error  "CUDA installation failed, please check EasyBuild logs..."
-  else
-    echo_green "CUDA installation at ${cuda_install_parent}/software/CUDA/${install_cuda_version} succeeded!"
-  fi
-  # clean up tmpdir
-  rm -rf "${tmpdir}"
-fi
diff --git a/scripts/utils.sh b/scripts/utils.sh
index 035a914445..c1342c21fc 100644
--- a/scripts/utils.sh
+++ b/scripts/utils.sh
@@ -14,7 +14,7 @@ ANY_ERROR_EXITCODE=1
 function fatal_error() {
     echo_red "ERROR: $1" >&2
     if [[ $# -gt 1 ]]; then
-      exit $2
+      exit "$2"
     else
       exit "${ANY_ERROR_EXITCODE}"
     fi
@@ -32,79 +32,29 @@ function check_exit_code {
     fi
 }
 
-# Reimplement 'mkdir -p' with reporting on where permissions break down
 function create_directory_structure() {
   # Ensure we are given a single path argument
   if [ $# -ne 1 ]; then
-    echo "Function requires a single (relative or absolute) path argument" >&2
-    return 1
+    echo_red "Function requires a single (relative or absolute) path argument" >&2
+    return $ANY_ERROR_EXITCODE
   fi
-
-  # set a persistent variable that knows the full structure
-  # (i.e., retains the value upon recursive calls)
-  full_structure="${full_structure:="$1"}"
-
-  local directory_structure="$1"
-
-  # Check if directory exists and is writeable
-  if [ -d "${directory_structure}" ]; then
-    if [ "${directory_structure}" = "${full_structure}" ]; then
-        # release our (unneeded) global variable
-        unset full_structure
-    fi
-    if [ -w "${directory_structure}" ]; then
-      # Nothing to be done
-      return 0
-    else
-      echo "Directory ${directory_structure} exists but is not writeable" >&2
-      return 1
-    fi
-  fi
-
-  local directory_structure_parent=$(dirname "${directory_structure}")
-
-  # If the parent doesn't exist we need to create it
-  if [ ! -d "${directory_structure_parent}" ]; then
-    # Create the parent via a recursive call to this function
-    # (if this doesn't succeed we need to return the error code)
-    if ! create_directory_structure "${directory_structure_parent}"; then
-      if [ "${directory_structure}" = "${full_structure}" ]; then
-        # release our (unneeded) global variable
-        unset full_structure
-      fi
-      return 1
-    fi
-  fi
-
-  # Check the parent is writeable, and create the new subdir
-  if [ -w "${directory_structure_parent}" ]; then
-    if [ "${directory_structure}" = "${full_structure}" ]; then
-      # release our (unneeded) global variable
-      unset full_structure
-    fi
-    if ! mkdir "${directory_structure}"; then
-      echo "'mkdir ${directory_structure}' failed for an unknown reason!" >&2
-      return 1
-    else
-      # Success!
-      return 0
-    fi
-  else
-    echo "Attempt to create ${full_structure} failed," \
-      "${directory_structure_parent} exists but you don't have write permissions." >&2
-    if [ "${directory_structure}" = "${full_structure}" ]; then
-      # release our global variable
-      unset full_structure
-    fi
-    return 1
+  dir_structure="$1"
+
+  # Attempt to create the directory structure
+  error_message=$(mkdir -p  2>&1)
+  return_code=$?
+  # If it fails be explicit about the error
+  if [ ${return_code} -ne 0 ]; then
+    echo_red "Creating ${dir_structure} failed with\n${error_message}" >&2
   fi
+  return $return_code
 }
 
 function get_path_for_tool {
     tool_name=$1
     tool_envvar_name=$2
 
-    which_out=$(which ${tool_name} 2>&1)
+    which_out=$(which "${tool_name}" 2>&1)
     exit_code=$?
     if [[ ${exit_code} -eq 0 ]]; then
         echo "INFO: found tool ${tool_name} in PATH (${which_out})" >&2
@@ -136,7 +86,7 @@ function get_host_from_url {
     url=$1
     re="(http|https)://([^/:]+)"
     if [[ $url =~ $re ]]; then
-        echo ${BASH_REMATCH[2]}
+        echo "${BASH_REMATCH[2]}"
         return 0
     else
         echo ""
@@ -148,7 +98,7 @@ function get_port_from_url {
     url=$1
     re="(http|https)://[^:]+:([0-9]+)"
     if [[ $url =~ $re ]]; then
-        echo ${BASH_REMATCH[2]}
+        echo "${BASH_REMATCH[2]}"
         return 0
     else
         echo ""
@@ -158,7 +108,7 @@ function get_port_from_url {
 
 function get_ipv4_address {
     hname=$1
-    hipv4=$(grep ${hname} /etc/hosts | grep -v '^[[:space:]]*#' | cut -d ' ' -f 1)
+    hipv4=$(grep "${hname}" /etc/hosts | grep -v '^[[:space:]]*#' | cut -d ' ' -f 1)
     # TODO try other methods if the one above does not work --> tool that verifies
     #      what method can be used?
     echo "${hipv4}"

From 793ba29d5bb1359a6ef5d631371cdf445b1147a5 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 09:42:57 +0100
Subject: [PATCH 20/31] Simply wrap `mkdir -p` for better error reporting

---
 .../install_cuda_host_injections.sh           | 86 +++++++++++++++++++
 1 file changed, 86 insertions(+)
 create mode 100755 gpu_support/cuda_utils/install_cuda_host_injections.sh

diff --git a/gpu_support/cuda_utils/install_cuda_host_injections.sh b/gpu_support/cuda_utils/install_cuda_host_injections.sh
new file mode 100755
index 0000000000..0684c7ca80
--- /dev/null
+++ b/gpu_support/cuda_utils/install_cuda_host_injections.sh
@@ -0,0 +1,86 @@
+#!/bin/bash
+
+# Initialise our bash functions
+source scripts/utils.sh
+
+if [[ $# -eq 0 ]] ; then
+    fatal_error "You must provide the CUDA version as an argument, e.g.:\n $0 11.3.1"
+fi
+install_cuda_version=$1
+if [[ -z "${EESSI_SOFTWARE_PATH}" ]]; then
+  fatal_error "This script cannot be used without having first defined EESSI_SOFTWARE_PATH"
+else
+  # As an installation location just use $EESSI_SOFTWARE_PATH but replacing `versions` with `host_injections`
+  # (CUDA is a binary installation so no need to worry too much about the EasyBuild setup)
+  cuda_install_parent=${EESSI_SOFTWARE_PATH/versions/host_injections}
+fi
+
+# Only install CUDA if specified version is not found.
+# This is only relevant for users, the shipped CUDA installation will
+# always be in versions instead of host_injections and have symlinks pointing
+# to host_injections for everything we're not allowed to ship
+# (existence of easybuild subdir implies a successful install)
+if [ -d "${cuda_install_parent}"/software/CUDA/"${install_cuda_version}"/easybuild ]; then
+  echo_green "CUDA software found! No need to install CUDA again, proceed with testing."
+else
+  # We need to be able write to the installation space so let's make sure we can
+  if ! create_directory_structure "${cuda_install_parent}"/software/CUDA ; then
+    fatal_error "No write permissions to directory ${cuda_install_parent}/software/CUDA"
+  fi
+
+  # we need a directory we can use for temporary storage
+  if [[ -z "${CUDA_TEMP_DIR}" ]]; then
+    tmpdir=$(mktemp -d)
+  else
+    tmpdir="${CUDA_TEMP_DIR}"/temp
+    if ! mkdir "$tmpdir" ; then
+      fatal_error "Could not create directory ${tmpdir}"
+    fi
+  fi
+
+  required_space_in_tmpdir=50000
+  # Let's see if we have sources and build locations defined if not, we use the temporary space
+  if [[ -z "${EASYBUILD_BUILDPATH}" ]]; then
+    export EASYBUILD_BUILDPATH=${tmpdir}/build
+    required_space_in_tmpdir=$((required_space_in_tempdir + 5000000))
+  fi
+  if [[ -z "${EASYBUILD_SOURCEPATH}" ]]; then
+    export EASYBUILD_SOURCEPATH=${tmpdir}/sources
+    required_space_in_tmpdir=$((required_space_in_tempdir + 5000000))
+  fi
+
+  # The install is pretty fat, you need lots of space for download/unpack/install (~3*5GB),
+  # need to do a space check before we proceed
+  avail_space=$(df --output=avail "${cuda_install_parent}"/ | tail -n 1 | awk '{print $1}')
+  if (( avail_space < 5000000 )); then
+    fatal_error "Need at least 5GB disk space to install CUDA under ${cuda_install_parent}, exiting now..."
+  fi
+  avail_space=$(df --output=avail "${tmpdir}"/ | tail -n 1 | awk '{print $1}')
+  if (( avail_space < required_space_in_tmpdir )); then
+    error1="Need at least ${required_space_in_tmpdir} disk space under ${tmpdir}.\n"
+    error2="You can set EASYBUILD_BUILDPATH and/or EASYBUILD_SOURCEPATH to reduce this requirement. Exiting now..."
+    fatal_error "${error1}${error2}"
+  fi
+
+  if [[ -z "${EBROOTEASYBUILD}" ]]; then
+    echo_yellow "Loading EasyBuild module to do actual install"
+    module load EasyBuild
+  fi
+
+  # we need the --rebuild option and a (random) dir for the module if the module
+  # file is shipped with EESSI
+  if [ -f ${EESSI_SOFTWARE_PATH}/modules/all/CUDA/${install_cuda_version}.lua ]; then
+    extra_args="--rebuild --installpath-modules=${tmpdir}"
+  fi
+  # We don't want hooks used in this install, we need a vanilla CUDA installation
+  touch "$tmpdir"/none.py
+  eb ${extra_args} --hooks="$tmpdir"/none.py --installpath="${cuda_install_parent}"/ CUDA-"${install_cuda_version}".eb
+  ret=$?
+  if [ $ret -ne 0 ]; then
+    fatal_error  "CUDA installation failed, please check EasyBuild logs..."
+  else
+    echo_green "CUDA installation at ${cuda_install_parent}/software/CUDA/${install_cuda_version} succeeded!"
+  fi
+  # clean up tmpdir
+  rm -rf "${tmpdir}"
+fi

From c0a12470de4ba2c4523f084edd8b516d6fb62235 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 09:46:57 +0100
Subject: [PATCH 21/31] Make CUDA version a variable

---
 EESSI-pilot-install-software.sh | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index d648162c31..f535634631 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -452,7 +452,7 @@ $EB $p7zip_ec --robot
 check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 # install CUDA (uses eb_hooks.py to only install runtime)
-cuda_ec="CUDA-11.3.1.eb"
+cuda_ec="CUDA-${cuda_version}.eb"
 echo ">> Installing $cuda_ec..."
 ok_msg="$cuda_ec installed, off to a good (?) start!"
 fail_msg="Failed to install $cuda_ec, woopsie..."
@@ -462,9 +462,9 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 # Add the host_injections CUDA so we can actually build CUDA apps
 # (which unbreaks the symlinks from the runtime installation)
 echo ">> Installing $cuda_ec under host_injections..."
-$(dirname "$BASH_SOURCE")/gpu_support/cuda_utils/install_cuda_host_injections.sh 11.3.1
-ok_msg="$cuda_ec (re)installed under host_injections!"
-fail_msg="Failed to install $cuda_ec under host_injections, woopsie..."
+$(dirname "$BASH_SOURCE")/gpu_support/cuda_utils/install_cuda_host_injections.sh ${cuda_version}
+ok_msg="CUDA $cuda_version (re)installed under host_injections!"
+fail_msg="Failed to install CUDA $cuda_version under host_injections, woopsie..."
 check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 # install CUDA samples (requires EESSI support for CUDA)

From 5e82923f01ff3089074e0760a1f64cfa5d3a5b2d Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 09:51:37 +0100
Subject: [PATCH 22/31] Use TOPDIR, be more descriptive

---
 EESSI-pilot-install-software.sh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/EESSI-pilot-install-software.sh b/EESSI-pilot-install-software.sh
index f535634631..035f851d61 100755
--- a/EESSI-pilot-install-software.sh
+++ b/EESSI-pilot-install-software.sh
@@ -461,8 +461,8 @@ check_exit_code $? "${ok_msg}" "${fail_msg}"
 
 # Add the host_injections CUDA so we can actually build CUDA apps
 # (which unbreaks the symlinks from the runtime installation)
-echo ">> Installing $cuda_ec under host_injections..."
-$(dirname "$BASH_SOURCE")/gpu_support/cuda_utils/install_cuda_host_injections.sh ${cuda_version}
+echo ">> Re-installing CUDA $cuda_version under host_injections (to un-break symlinks in EESSI installation)..."
+"${TOPDIR}"/gpu_support/cuda_utils/install_cuda_host_injections.sh ${cuda_version}
 ok_msg="CUDA $cuda_version (re)installed under host_injections!"
 fail_msg="Failed to install CUDA $cuda_version under host_injections, woopsie..."
 check_exit_code $? "${ok_msg}" "${fail_msg}"

From 8384b25a9658944fa57de03da7dcac84ae5b0a8e Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 10:06:45 +0100
Subject: [PATCH 23/31] Add missing argument

---
 scripts/utils.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/scripts/utils.sh b/scripts/utils.sh
index c1342c21fc..ef9e4095c2 100644
--- a/scripts/utils.sh
+++ b/scripts/utils.sh
@@ -41,7 +41,7 @@ function create_directory_structure() {
   dir_structure="$1"
 
   # Attempt to create the directory structure
-  error_message=$(mkdir -p  2>&1)
+  error_message=$(mkdir -p "$dir_structure" 2>&1)
   return_code=$?
   # If it fails be explicit about the error
   if [ ${return_code} -ne 0 ]; then

From 98fe2a747fef928e123288fba157433322473dac Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 11:49:48 +0100
Subject: [PATCH 24/31] Improve error messages in new bash function

---
 scripts/utils.sh | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/scripts/utils.sh b/scripts/utils.sh
index ef9e4095c2..b9aad997e1 100644
--- a/scripts/utils.sh
+++ b/scripts/utils.sh
@@ -45,8 +45,18 @@ function create_directory_structure() {
   return_code=$?
   # If it fails be explicit about the error
   if [ ${return_code} -ne 0 ]; then
-    echo_red "Creating ${dir_structure} failed with\n${error_message}" >&2
+    real_dir=$(readlink -m "$dir_structure")
+    echo_red "Creating ${dir_structure} (real path ${real_dir}) failed with:\n ${error_message}" >&2
+  else
+    # If we're creating it, our use case is that we want to be able to write there
+    # (this is a check in case the directory already existed)
+    if [ ! -w "${dir_structure}" ]; then
+      real_dir=$(readlink -m "$dir_structure")
+      echo_red "You do not have (required) write permissions to ${dir_structure} (real path ${real_dir})!"
+      return $ANY_ERROR_EXITCODE
+    fi
   fi
+
   return $return_code
 }
 

From bbe7df210da6594ba94a27cf6ce072c7042cb24b Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 11:53:09 +0100
Subject: [PATCH 25/31] Stick with return_code

---
 scripts/utils.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/scripts/utils.sh b/scripts/utils.sh
index b9aad997e1..b052e0a1ec 100644
--- a/scripts/utils.sh
+++ b/scripts/utils.sh
@@ -53,7 +53,7 @@ function create_directory_structure() {
     if [ ! -w "${dir_structure}" ]; then
       real_dir=$(readlink -m "$dir_structure")
       echo_red "You do not have (required) write permissions to ${dir_structure} (real path ${real_dir})!"
-      return $ANY_ERROR_EXITCODE
+      return_code=$ANY_ERROR_EXITCODE
     fi
   fi
 

From 95dc24509434d38e353f21a54413d771ebfdc401 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 11:56:02 +0100
Subject: [PATCH 26/31] Use realpath to be consistent with other scripts

---
 scripts/utils.sh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/scripts/utils.sh b/scripts/utils.sh
index b052e0a1ec..099c1712ef 100644
--- a/scripts/utils.sh
+++ b/scripts/utils.sh
@@ -45,13 +45,13 @@ function create_directory_structure() {
   return_code=$?
   # If it fails be explicit about the error
   if [ ${return_code} -ne 0 ]; then
-    real_dir=$(readlink -m "$dir_structure")
+    real_dir=$(realpath -s "$dir_structure")
     echo_red "Creating ${dir_structure} (real path ${real_dir}) failed with:\n ${error_message}" >&2
   else
     # If we're creating it, our use case is that we want to be able to write there
     # (this is a check in case the directory already existed)
     if [ ! -w "${dir_structure}" ]; then
-      real_dir=$(readlink -m "$dir_structure")
+      real_dir=$(realpath -s "$dir_structure")
       echo_red "You do not have (required) write permissions to ${dir_structure} (real path ${real_dir})!"
       return_code=$ANY_ERROR_EXITCODE
     fi

From a1270f2359d64f9ee19f7ac10645e90d71ee2437 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 11:58:22 +0100
Subject: [PATCH 27/31] Wrong realpath flag

---
 scripts/utils.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/scripts/utils.sh b/scripts/utils.sh
index 099c1712ef..b5aa430985 100644
--- a/scripts/utils.sh
+++ b/scripts/utils.sh
@@ -51,7 +51,7 @@ function create_directory_structure() {
     # If we're creating it, our use case is that we want to be able to write there
     # (this is a check in case the directory already existed)
     if [ ! -w "${dir_structure}" ]; then
-      real_dir=$(realpath -s "$dir_structure")
+      real_dir=$(realpath -m "$dir_structure")
       echo_red "You do not have (required) write permissions to ${dir_structure} (real path ${real_dir})!"
       return_code=$ANY_ERROR_EXITCODE
     fi

From aba486de42f250b7ada1d7459188ba579337540b Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 11:59:06 +0100
Subject: [PATCH 28/31] Wrong realpath flag

---
 scripts/utils.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/scripts/utils.sh b/scripts/utils.sh
index b5aa430985..f043ba0ca6 100644
--- a/scripts/utils.sh
+++ b/scripts/utils.sh
@@ -45,7 +45,7 @@ function create_directory_structure() {
   return_code=$?
   # If it fails be explicit about the error
   if [ ${return_code} -ne 0 ]; then
-    real_dir=$(realpath -s "$dir_structure")
+    real_dir=$(realpath -m "$dir_structure")
     echo_red "Creating ${dir_structure} (real path ${real_dir}) failed with:\n ${error_message}" >&2
   else
     # If we're creating it, our use case is that we want to be able to write there

From d2d1fc3b4a6b4e5372f1218b3da86f621f09f674 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 12:06:58 +0100
Subject: [PATCH 29/31] Fix typo

---
 .../cuda_utils/install_cuda_host_injections.sh     | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/gpu_support/cuda_utils/install_cuda_host_injections.sh b/gpu_support/cuda_utils/install_cuda_host_injections.sh
index 0684c7ca80..4264f06b83 100755
--- a/gpu_support/cuda_utils/install_cuda_host_injections.sh
+++ b/gpu_support/cuda_utils/install_cuda_host_injections.sh
@@ -42,11 +42,11 @@ else
   # Let's see if we have sources and build locations defined if not, we use the temporary space
   if [[ -z "${EASYBUILD_BUILDPATH}" ]]; then
     export EASYBUILD_BUILDPATH=${tmpdir}/build
-    required_space_in_tmpdir=$((required_space_in_tempdir + 5000000))
+    required_space_in_tmpdir=$((required_space_in_tmpdir + 5000000))
   fi
   if [[ -z "${EASYBUILD_SOURCEPATH}" ]]; then
     export EASYBUILD_SOURCEPATH=${tmpdir}/sources
-    required_space_in_tmpdir=$((required_space_in_tempdir + 5000000))
+    required_space_in_tmpdir=$((required_space_in_tmpdir + 5000000))
   fi
 
   # The install is pretty fat, you need lots of space for download/unpack/install (~3*5GB),
@@ -57,9 +57,10 @@ else
   fi
   avail_space=$(df --output=avail "${tmpdir}"/ | tail -n 1 | awk '{print $1}')
   if (( avail_space < required_space_in_tmpdir )); then
-    error1="Need at least ${required_space_in_tmpdir} disk space under ${tmpdir}.\n"
-    error2="You can set EASYBUILD_BUILDPATH and/or EASYBUILD_SOURCEPATH to reduce this requirement. Exiting now..."
-    fatal_error "${error1}${error2}"
+    error="Need at least ${required_space_in_tmpdir} disk space under ${tmpdir}.\n"
+    error="${error}You can set EASYBUILD_BUILDPATH and/or EASYBUILD_SOURCEPATH "
+    error="${error}to reduce this requirement. Exiting now..."
+    fatal_error "${error}"
   fi
 
   if [[ -z "${EBROOTEASYBUILD}" ]]; then
@@ -69,11 +70,12 @@ else
 
   # we need the --rebuild option and a (random) dir for the module if the module
   # file is shipped with EESSI
-  if [ -f ${EESSI_SOFTWARE_PATH}/modules/all/CUDA/${install_cuda_version}.lua ]; then
+  if [ -f "${EESSI_SOFTWARE_PATH}"/modules/all/CUDA/"${install_cuda_version}".lua ]; then
     extra_args="--rebuild --installpath-modules=${tmpdir}"
   fi
   # We don't want hooks used in this install, we need a vanilla CUDA installation
   touch "$tmpdir"/none.py
+  # shellcheck disable=SC2086  # Intended splitting of extra_args
   eb ${extra_args} --hooks="$tmpdir"/none.py --installpath="${cuda_install_parent}"/ CUDA-"${install_cuda_version}".eb
   ret=$?
   if [ $ret -ne 0 ]; then

From 562e94b7beba8e9e4fc71c84a6ad5b154ad9283f Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 12:11:49 +0100
Subject: [PATCH 30/31] Always add the rebuild option if we get to the point
 where we actually install CUDA

---
 gpu_support/cuda_utils/install_cuda_host_injections.sh | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/gpu_support/cuda_utils/install_cuda_host_injections.sh b/gpu_support/cuda_utils/install_cuda_host_injections.sh
index 4264f06b83..0d68759c1e 100755
--- a/gpu_support/cuda_utils/install_cuda_host_injections.sh
+++ b/gpu_support/cuda_utils/install_cuda_host_injections.sh
@@ -68,11 +68,10 @@ else
     module load EasyBuild
   fi
 
-  # we need the --rebuild option and a (random) dir for the module if the module
-  # file is shipped with EESSI
-  if [ -f "${EESSI_SOFTWARE_PATH}"/modules/all/CUDA/"${install_cuda_version}".lua ]; then
-    extra_args="--rebuild --installpath-modules=${tmpdir}"
-  fi
+  # we need the --rebuild option and a (random) dir for the module since we are
+  # fixing the broken links of the EESSI-shipped installation
+  extra_args="--rebuild --installpath-modules=${tmpdir}"
+
   # We don't want hooks used in this install, we need a vanilla CUDA installation
   touch "$tmpdir"/none.py
   # shellcheck disable=SC2086  # Intended splitting of extra_args

From b4ae5f027f77fd38cf215b0fdf04b8136760d414 Mon Sep 17 00:00:00 2001
From: Alan O'Cais <alan.ocais@cecam.org>
Date: Fri, 3 Mar 2023 12:42:31 +0100
Subject: [PATCH 31/31] Expose CUDA_TEMP_DIR

---
 gpu_support/cuda_utils/install_cuda_host_injections.sh | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/gpu_support/cuda_utils/install_cuda_host_injections.sh b/gpu_support/cuda_utils/install_cuda_host_injections.sh
index 0d68759c1e..1ddccf4e82 100755
--- a/gpu_support/cuda_utils/install_cuda_host_injections.sh
+++ b/gpu_support/cuda_utils/install_cuda_host_injections.sh
@@ -58,7 +58,8 @@ else
   avail_space=$(df --output=avail "${tmpdir}"/ | tail -n 1 | awk '{print $1}')
   if (( avail_space < required_space_in_tmpdir )); then
     error="Need at least ${required_space_in_tmpdir} disk space under ${tmpdir}.\n"
-    error="${error}You can set EASYBUILD_BUILDPATH and/or EASYBUILD_SOURCEPATH "
+    error="${error}Set the environment variable CUDA_TEMP_DIR to a location with adequate space to pass this check."
+    error="${error}You can alternatively set EASYBUILD_BUILDPATH and/or EASYBUILD_SOURCEPATH "
     error="${error}to reduce this requirement. Exiting now..."
     fatal_error "${error}"
   fi
@@ -75,7 +76,7 @@ else
   # We don't want hooks used in this install, we need a vanilla CUDA installation
   touch "$tmpdir"/none.py
   # shellcheck disable=SC2086  # Intended splitting of extra_args
-  eb ${extra_args} --hooks="$tmpdir"/none.py --installpath="${cuda_install_parent}"/ CUDA-"${install_cuda_version}".eb
+  eb --prefix="$tmpdir" ${extra_args} --hooks="$tmpdir"/none.py --installpath="${cuda_install_parent}"/ CUDA-"${install_cuda_version}".eb
   ret=$?
   if [ $ret -ne 0 ]; then
     fatal_error  "CUDA installation failed, please check EasyBuild logs..."