Bug#992374: buster-pu: package nvidia-cuda-toolkit/9.2.148-7+deb10u1
Package: release.debian.org
Severity: normal
Tags: buster
User: release.debian.org@packages.debian.org
Usertags: pu
The ppc64el nvidia-cuda-toolkit package in buster does not ship the
customized nvcc.profile configuration file but the unmodified upstream
one which does not work with the FHS layout used by the Debian
packaging. #991592
Since some people actually use the ppc64el stable packages in production
environments, I'd like to fix that.
This problem is specific to the version in buster/ppc64el:
- newer releases (10+) don't provide incremental patch releases, they
just release the full toolkit more frequently
- in newer (10+) toolkit releases, ppc64el has a .run installer like
amd64, no more repacking of upstream .deb files needed
There is also some replacement of $< in debian/rules with the intended
value, since $< does not work for order-only prerequisites. (This only
affects verbose output, which contained a lot of incorrect noise.)
Andreas
diff --git a/debian/changelog b/debian/changelog
index cba0bf5..0bfbae6 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -1,3 +1,10 @@
+nvidia-cuda-toolkit (9.2.148-7+deb10u1) buster; urgency=medium
+
+ * [ppc64el]: Do not overwrite the customized nvcc.profile with the
+ unmodified upstream copy from CUDA 9.2 Patch 1. (Closes: #991592)
+
+ -- Andreas Beckmann <anbe@debian.org> Tue, 17 Aug 2021 18:57:22 +0200
+
nvidia-cuda-toolkit (9.2.148-7) unstable; urgency=medium
* Use debianize.patch to customize paths etc. in nvcc.profile and the
diff --git a/debian/gbp.conf b/debian/gbp.conf
index 92b4ec0..db5c248 100644
--- a/debian/gbp.conf
+++ b/debian/gbp.conf
@@ -7,4 +7,4 @@ pristine-tar = False
compression = gz
tarball-dir = ../tarballs-nvidia-cuda-toolkit/%(version)s
component = [ 'ppc64el', 'amd64-patch1', 'ppc64el-patch1', ]
-debian-branch = master
+debian-branch = buster
diff --git a/debian/rules b/debian/rules
index 7d28500..08ce2b6 100755
--- a/debian/rules
+++ b/debian/rules
@@ -65,7 +65,7 @@ nvidia-cuda-amd64-patch1: | nvidia-cuda-amd64
$(RM) -v $@.tmp/install_patch.pl
$(RM) -v $@.tmp/payload/install_manifest
mv $@.tmp $@
- for f in $$(cd $@ && find . ! -type d) ; do test -e $</$$f || echo "new file in $@: $$f" ; done
+ for f in $$(cd $@ && find . ! -type d) ; do test -e nvidia-cuda-amd64/$$f || echo "new file in $@: $$f" ; done
nvidia-cuda-ppc64el:
dh_testdir
@@ -96,11 +96,12 @@ nvidia-cuda-ppc64el-patch1: | nvidia-cuda-ppc64el
$(RM) -r $@.tmp/ubuntutree/usr/share/doc
$(RM) -r $@.tmp/ubuntutree/usr/share/lintian
$(RM) -v $@.tmp/ubuntutree/etc/ld.so.conf.d/cuda-$(subst .,-,$(CUDA_SOVERSION)).conf
+ cmp -s nvidia-cuda-ppc64el/bin/nvcc.profile $@.tmp/bin/nvcc.profile && $(RM) -v $@.tmp/bin/nvcc.profile
$(RM) $@.tmp/include $@.tmp/lib64
mv $@.tmp/targets/ppc64le-linux/include $@.tmp/include
mv $@.tmp/targets/ppc64le-linux/lib $@.tmp/lib64
mv $@.tmp $@
- for f in $$(cd $@ && find . ! -type d) ; do test -e $</$$f || echo "new file in $@: $$f" ; done
+ for f in $$(cd $@ && find . ! -type d) ; do test -e nvidia-cuda-ppc64el/$$f || echo "new file in $@: $$f" ; done
unpack-stamp: $(filter debian/patches%,$(AUTOGEN))
unpack-stamp: nvidia-cuda-$(DEB_HOST_ARCH)
Reply to: