Re: py-tables on sparc64 [sparc64 bulk build report]
On Thu, Jan 16, 2020 at 08:59:31AM -0500, Kurt Mosiejczuk wrote: > On Thu, Jan 16, 2020 at 08:37:56AM -0500, Kurt Mosiejczuk wrote: > > On Thu, Jan 16, 2020 at 01:26:07PM +, Stuart Henderson wrote: > > > > No sparc64 to test here but this sort of failure usually is reproducible. > > > As this is currently built with base-gcc on sparc64, switching to > > > ports-gcc > > > is probably the sanest first step. > > > > COMPILER= base-clang ports-gcc > > OK. Just this change makes it finish compiling fairly quickly on my test > machine. I ran it again just to be sure because it went through so > quickly. Thanks. While there I'd refine tests a bit more. With the diff below: Ran 6355 tests in 1580.823s FAILED (errors=2, skipped=13) The REVISION bump might not be strictly necessary but should I include it anyway? -m Index: Makefile === RCS file: /cvs/ports/math/py-tables/Makefile,v retrieving revision 1.1 diff -u -p -u -p -r1.1 Makefile --- Makefile10 Jan 2020 13:30:41 - 1.1 +++ Makefile18 Jan 2020 18:10:11 - @@ -6,6 +6,7 @@ MODPY_EGG_VERSION= 3.6.1 DISTNAME= tables-${MODPY_EGG_VERSION} PKGNAME= py-${DISTNAME} CATEGORIES=math +REVISION= 0 HOMEPAGE= https://www.pytables.org/ MAINTAINER=Martin Reindl @@ -13,12 +14,13 @@ MAINTAINER= Martin Reindl
Re: py-tables on sparc64 [sparc64 bulk build report]
On Thu, Jan 16, 2020 at 08:37:56AM -0500, Kurt Mosiejczuk wrote: > On Thu, Jan 16, 2020 at 01:26:07PM +, Stuart Henderson wrote: > > No sparc64 to test here but this sort of failure usually is reproducible. > > As this is currently built with base-gcc on sparc64, switching to ports-gcc > > is probably the sanest first step. > > COMPILER= base-clang ports-gcc OK. Just this change makes it finish compiling fairly quickly on my test machine. I ran it again just to be sure because it went through so quickly. --Kurt
Re: py-tables on sparc64 [sparc64 bulk build report]
On Thu, Jan 16, 2020 at 01:26:07PM +, Stuart Henderson wrote: > On 2020/01/16 07:21, Martin Reindl wrote: > > On Wed, Jan 15, 2020 at 07:49:43PM -0700, k...@openbsd.org wrote: > > > Bulk build on sparc64-0.ports.openbsd.org > > > Started : Sun Jan 12 23:18:35 MST 2020 > > > Finished: Wed Jan 15 19:48:36 MST 2020 > > > Duration: 2 Days 20 hours 30 minutes > > > Built using OpenBSD 6.6-current (GENERIC.MP) #184: Sat Jan 11 18:58:57 > > > MST 2020 > > > http://build-failures.rhaalovely.net/sparc64/2020-01-12/math/py-tables.log > > Is this reproducible? > No sparc64 to test here but this sort of failure usually is reproducible. > As this is currently built with base-gcc on sparc64, switching to ports-gcc > is probably the sanest first step. > COMPILER= base-clang ports-gcc I actually tried that yesterday as a quick in-place fix. Seemed to hang just the same. But, that was after a rough few days and a long drive. My test machine (without trying ports-gcc) is hung on: building 'tables.hdf5extension' extension cc -Wno-unused-result -Wsign-compare -DNDEBUG -O2 -pipe -fPIC -O2 -pipe -O2 -pipe -O2 -pipe -I/usr/local/include -fPIC -DNDEBUG=1 -DHAVE_LZO2_LIB=1 -DHAVE_BZ2_LIB=1 -DHAVE_BLOSC_LIB=1 -Ihdf5-blosc/src -I/usr/local/lib/python3.7/site-packages/numpy/core/include -I/usr/local/include/python3.7m -c tables/hdf5extension.c -o /usr/ports/pobj/py-tables-3.6.1/tables-3.6.1/temp.openbsd-6.6-sparc64-3.7/tables/hdf5extension.o -O2 -pipe -I/usr/local/include -Isrc -DH5_USE_18_API -DH5Acreate_vers=2 -DH5Aiterate_vers=2 -DH5Dcreate_vers=2 -DH5Dopen_vers=2 -DH5Eclear_vers=2 -DH5Eprint_vers=2 -DH5Epush_vers=2 -DH5Eset_auto_vers=2 -DH5Eget_auto_vers=2 -DH5Ewalk_vers=2 -DH5E_auto_t_vers=2 -DH5Gcreate_vers=2 -DH5Gopen_vers=2 -DH5Pget_filter_vers=2 -DH5Pget_filter_by_id_vers=2 -DH5Tarray_create_vers=2 -DH5Tget_array_dims_vers=2 -DH5Z_class_t_vers=2 In file included from /usr/local/lib/python3.7/site-packages/numpy/core/include/numpy/ndarraytypes.h:1816, from /usr/local/lib/python3.7/site-packages/numpy/core/include/numpy/ndarrayobject.h:18, from /usr/local/lib/python3.7/site-packages/numpy/core/include/numpy/arrayobject.h:4, from tables/hdf5extension.c:611: /usr/local/lib/python3.7/site-packages/numpy/core/include/numpy/npy_1_7_deprecated_api.h:15:2: warning: #warning "Using deprecated NumPy API, disable it by " "#defining NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION" tables/hdf5extension.c: In function '__pyx_pf_6tables_13hdf5extension_5Array_4_open_array': tables/hdf5extension.c:18511: warning: comparison between signed and unsigned tables/hdf5extension.c: In function '__pyx_pf_6tables_13hdf5extension_7VLArray_10_read_array': tables/hdf5extension.c:26272: warning: comparison between signed and unsigned I'll try the ports-gcc swap and run it again. --Kurt
Re: py-tables on sparc64 [sparc64 bulk build report]
On 2020/01/16 07:21, Martin Reindl wrote: > On Wed, Jan 15, 2020 at 07:49:43PM -0700, k...@openbsd.org wrote: > > Bulk build on sparc64-0.ports.openbsd.org > > > > Started : Sun Jan 12 23:18:35 MST 2020 > > Finished: Wed Jan 15 19:48:36 MST 2020 > > Duration: 2 Days 20 hours 30 minutes > > > > Built using OpenBSD 6.6-current (GENERIC.MP) #184: Sat Jan 11 18:58:57 MST > > 2020 > > http://build-failures.rhaalovely.net/sparc64/2020-01-12/math/py-tables.log > > Is this reproducible? > > -m > No sparc64 to test here but this sort of failure usually is reproducible. As this is currently built with base-gcc on sparc64, switching to ports-gcc is probably the sanest first step. COMPILER= base-clang ports-gcc
Re: py-tables on sparc64 [sparc64 bulk build report]
On Thu, Jan 16, 2020 at 07:21:00AM +0100, Martin Reindl wrote: > On Wed, Jan 15, 2020 at 07:49:43PM -0700, k...@openbsd.org wrote: > > Bulk build on sparc64-0.ports.openbsd.org > > Started : Sun Jan 12 23:18:35 MST 2020 > > Finished: Wed Jan 15 19:48:36 MST 2020 > > Duration: 2 Days 20 hours 30 minutes > > Built using OpenBSD 6.6-current (GENERIC.MP) #184: Sat Jan 11 18:58:57 MST > > 2020 > > http://build-failures.rhaalovely.net/sparc64/2020-01-12/math/py-tables.log > Is this reproducible? I ended up killing this one. It ended up frozen for hours and hours on a single compilation. I even restarted it more than once. I kicked off a build last night on my test machine to see if I could better reproduce it. --Kurt
Re: py-tables on sparc64 [sparc64 bulk build report]
On Wed, Jan 15, 2020 at 07:49:43PM -0700, k...@openbsd.org wrote: > Bulk build on sparc64-0.ports.openbsd.org > > Started : Sun Jan 12 23:18:35 MST 2020 > Finished: Wed Jan 15 19:48:36 MST 2020 > Duration: 2 Days 20 hours 30 minutes > > Built using OpenBSD 6.6-current (GENERIC.MP) #184: Sat Jan 11 18:58:57 MST > 2020 > http://build-failures.rhaalovely.net/sparc64/2020-01-12/math/py-tables.log Is this reproducible? -m