[Intel-gfx] [PATCH i-g-t 2/2] tests: use standard install prefix for programs, scripts and data
Thomas Wood
thomas.wood at intel.com
Thu Apr 9 08:56:04 PDT 2015
On 9 April 2015 at 07:45, Joonas Lahtinen
<joonas.lahtinen at linux.intel.com> wrote:
> On ke, 2015-04-08 at 14:56 +0100, Thomas Wood wrote:
>> Use the pkglibexec and pkgdata prefixes rather than setting bindir and
>> datadir. This also removes the extra 'tests' directory from within the
>> package libexec and data directories.
>>
>> Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
>> Signed-off-by: Thomas Wood <thomas.wood at intel.com>
>> ---
>> tests/Makefile.am | 15 +++++++++++----
>> tests/Makefile.sources | 15 ---------------
>> 2 files changed, 11 insertions(+), 19 deletions(-)
>>
>> diff --git a/tests/Makefile.am b/tests/Makefile.am
>> index d6de373..dc864f4 100644
>> --- a/tests/Makefile.am
>> +++ b/tests/Makefile.am
>> @@ -27,18 +27,25 @@ multi-tests.txt: Makefile.sources
>> @echo ${multi_kernel_tests} >> $@
>> @echo END TESTLIST >> $@
>>
>> -igt_tests_bin_PROGRAMS += \
>> +noinst_PROGRAMS = \
>> + $(HANG) \
>> + $(TESTS_testsuite) \
>> + $(NULL)
>> +
>> +pkglibexec_PROGRAMS = \
>> + gem_alive \
>> + gem_stress \
>> $(TESTS_progs) \
>> $(TESTS_progs_M) \
>> $(NULL)
>
> Reasoning for the tests directory was that if the tests just sit under
> pkglibexec, they might be mistaken for tools (as the i-g-t package name
> suggests) by packagers or really anybody.
I think pklibexec is sufficient since user tools ought to be install
in bin anyway. The test list should also be installed in pkgdatadir to
identify the test binaries.
>
> If it's so important not to have the tests directory, I'd rather suffix
> all the programs with "_test" during build. It would cause some
> cascading changes too, so why do you want to get rid of the tests
> directory in the first place?
>
>>
>> -dist_igt_tests_bin_SCRIPTS = \
>> +dist_pkglibexec_SCRIPTS = \
>> $(TESTS_scripts) \
>> $(TESTS_scripts_M) \
>> $(scripts) \
>> $(NULL)
>>
>
> This one was giving me complaints with automake 1.14.1 that SCRIPTS
> don't belong straight to pkglibexec, so this would have to be kept
> indirect like it is, be it with tests directory or not.
I also have automake 1.14.1 and it doesn't seem to complain. Does it
prevent automake from completing successfully?
>
>> -dist_igt_tests_data_DATA = \
>> +dist_pkgdata_DATA = \
>> $(IMAGES) \
>> $(NULL)
>>
>> @@ -52,7 +59,7 @@ AM_CFLAGS = $(DRM_CFLAGS) $(CWARNFLAGS) $(DEBUG_CFLAGS)\
>> -I$(srcdir)/../lib \
>> -include "$(srcdir)/../lib/check-ndebug.h" \
>> -DIGT_SRCDIR=\""$(abs_srcdir)"\" \
>> - -DIGT_DATADIR=\""$(igt_tests_datadir)"\" \
>> + -DIGT_DATADIR=\""$(pkgdatadir)"\" \
>> $(LIBUNWIND_CFLAGS) \
>> $(NULL)
>>
>> diff --git a/tests/Makefile.sources b/tests/Makefile.sources
>> index 59a06e9..4bf11bf 100644
>> --- a/tests/Makefile.sources
>> +++ b/tests/Makefile.sources
>> @@ -1,18 +1,3 @@
>> -igt_tests_bindir = $(pkglibexecdir)/tests
>> -igt_tests_datadir = $(pkgdatadir)/tests
>> -
>> -noinst_PROGRAMS = \
>> - $(HANG) \
>> - $(TESTS_testsuite) \
>> - $(NULL)
>> -
>> -igt_tests_bin_PROGRAMS = \
>> - gem_alive \
>> - gem_stress \
>> - $(TESTS_progs) \
>> - $(TESTS_progs_M) \
>> - $(NULL)
>> -
>> NOUVEAU_TESTS_M = \
>> prime_nv_api \
>> prime_nv_pcopy \
More information about the Intel-gfx
mailing list