Hello community, here is the log from the commit of package netcdf for openSUSE:Factory checked in at 2017-10-27 13:47:22 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Comparing /work/SRC/openSUSE:Factory/netcdf (Old) and /work/SRC/openSUSE:Factory/.netcdf.new (New) ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "netcdf" Fri Oct 27 13:47:22 2017 rev:34 rq:536591 version:4.4.1.1 Changes: -------- --- /work/SRC/openSUSE:Factory/netcdf/netcdf.changes 2017-10-24 22:21:07.402737189 +0200 +++ /work/SRC/openSUSE:Factory/.netcdf.new/netcdf.changes 2017-10-27 13:47:25.987247681 +0200 @@ -1,0 +2,7 @@ +Tue Oct 24 20:16:57 UTC 2017 - jjo...@suse.com + +- Added patch netcdf-h5-test-fix-type-mismatch.patch + * Fixes type mismatch in test tst_h_dimscales + (bsc#1064705). + +------------------------------------------------------------------- New: ---- netcdf-h5-test-fix-type-mismatch.patch ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Other differences: ------------------ ++++++ netcdf.spec ++++++ --- /var/tmp/diff_new_pack.sQh5Wb/_old 2017-10-27 13:47:27.083196459 +0200 +++ /var/tmp/diff_new_pack.sQh5Wb/_new 2017-10-27 13:47:27.087196272 +0200 @@ -134,6 +134,7 @@ Patch0: %{pname}-no_date_time.patch # PATCH-FIX-UPSTREAM netcdf-disable-szip-linking.patch badshah...@gmail.com -- Disable building with szip support for now, since test nc_test4/tst_interops5.c fails otherwise; this is likely to be fixed with version 4.5.0 of netcdf, but requires big enough changes that we don't backport it here. Patch1: netcdf-disable-szip-linking.patch +Patch2: netcdf-h5-test-fix-type-mismatch.patch BuildRoot: %{_tmppath}/%{name}-%{version}-build BuildRequires: gawk BuildRequires: libtool @@ -313,6 +314,7 @@ %setup -q -n %{pname}-%{version} %patch0 -p0 -b .no_date_time %patch1 -p1 +%patch2 -p1 m4 libsrc/ncx.m4 > libsrc/ncx.c # Create baselib.conf dynamically (non-HPC build only). ++++++ _multibuild ++++++ --- /var/tmp/diff_new_pack.sQh5Wb/_old 2017-10-27 13:47:27.135194028 +0200 +++ /var/tmp/diff_new_pack.sQh5Wb/_new 2017-10-27 13:47:27.139193841 +0200 @@ -3,5 +3,5 @@ <package>openmpi</package> <package>gnu-hpc</package> <package>gnu-openmpi-hpc</package> - <!-- <package>gnu-mvapich2-hpc</package> --> + <package>gnu-mvapich2-hpc</package> </multibuild> ++++++ netcdf-h5-test-fix-type-mismatch.patch ++++++ Subject: [netcdf] Fix type mismatch in tst_h_dimscales test When iterating through scales, the test passed a pointer to a local int variable for the iterator callback to use. The callback casts the data to the hid_t type, which is incompatible with the int type. This causes stack corruption and inevitably failure. Changing the visitor_data to the hid_t type eliminates the stack overflow and the test succeeds. This was discovered while compiling with the mvapich2 compiler, built using the SUSE-HPC build system. Signed-off by: John Jolly <jjo...@suse.com> --- Index: netcdf-4.4.1.1/h5_test/tst_h_dimscales.c =================================================================== --- netcdf-4.4.1.1.orig/h5_test/tst_h_dimscales.c +++ netcdf-4.4.1.1/h5_test/tst_h_dimscales.c @@ -85,7 +85,7 @@ rec_scan_group(hid_t grpid) } else { - int visitor_data = 0; + hid_t visitor_data = 0; /* Here's how to get the number of scales attached * to the dataset's dimension 0. */ @@ -426,7 +426,7 @@ main() else { char label[STR_LEN+1]; - int visitor_data = 0; + hid_t visitor_data = 0; /* Here's how to get the number of scales attached * to the dataset's dimension 0. */ @@ -617,7 +617,7 @@ main() else { char label[STR_LEN+1]; - int visitor_data = 0; + hid_t visitor_data = 0; /* SHould have these dimensions... */ if (dims[TIME_DIM] != 0 || dims[LAT_DIM] != LAT_LEN ||