Thread: pgsql: Move security_label test
Move security_label test Rather than have the core security_label regression test depend on the dummy_seclabel module, have that part of the test be executed by dummy_seclabel itself directly. This simplifies the testing rig a bit; in particular it should silence the problems from the MSVC buildfarm phylum, which haven't yet gotten taught how to install src/test/modules. Branch ------ master Details ------- http://git.postgresql.org/pg/commitdiff/df761e3cf79db09d602610ee61e51cb378288382 Modified Files -------------- src/test/modules/dummy_seclabel/Makefile | 2 + .../dummy_seclabel/input/dummy_seclabel.source | 79 +++++++++++++ .../dummy_seclabel/output/dummy_seclabel.source | 87 ++++++++++++++ src/test/regress/GNUmakefile | 16 +-- src/test/regress/expected/security_label.out | 47 ++++++++ src/test/regress/input/security_label.source | 108 ----------------- src/test/regress/output/security_label.source | 123 -------------------- src/test/regress/sql/security_label.sql | 49 ++++++++ 8 files changed, 268 insertions(+), 243 deletions(-)
On Tue, Dec 2, 2014 at 4:13 AM, Alvaro Herrera <alvherre@alvh.no-ip.org> wrote: > Move security_label test > > Rather than have the core security_label regression test depend on the > dummy_seclabel module, have that part of the test be executed by > dummy_seclabel itself directly. This simplifies the testing rig a bit; > in particular it should silence the problems from the MSVC buildfarm > phylum, which haven't yet gotten taught how to install src/test/modules. After this module has been moved to src/test/modules, I am seeing failures with make check when manually kicking the test on at least OSX where I tried because expected/ and sql/ are missing. pg_regress does not like that much. That's annoying because check-world would simply fail. Also, I think that src/test/modules/dummy_seclabel/Makefile should be taught to clean sql/dummy_seclabel.sql and expected/dummy_seclabel.out with EXTRA_CLEAN and that we should have a .gitignore to ignore tmp_check/ results/ and log/ in this path. Per se the patch attached with those things added. Thanks, -- Michael
Attachment
Michael Paquier wrote: > On Tue, Dec 2, 2014 at 4:13 AM, Alvaro Herrera <alvherre@alvh.no-ip.org> wrote: > > Move security_label test > > > > Rather than have the core security_label regression test depend on the > > dummy_seclabel module, have that part of the test be executed by > > dummy_seclabel itself directly. This simplifies the testing rig a bit; > > in particular it should silence the problems from the MSVC buildfarm > > phylum, which haven't yet gotten taught how to install src/test/modules. > After this module has been moved to src/test/modules, I am seeing > failures with make check when manually kicking the test on at least > OSX where I tried because expected/ and sql/ are missing. pg_regress > does not like that much. That's annoying because check-world would > simply fail. I had the directories locally but hadn't git-pushed them. > Also, I think that src/test/modules/dummy_seclabel/Makefile should be > taught to clean sql/dummy_seclabel.sql and expected/dummy_seclabel.out > with EXTRA_CLEAN and that we should have a .gitignore to ignore > tmp_check/ results/ and log/ in this path. Per se the patch attached > with those things added. Pushed the whole thing, thanks. -- Álvaro Herrera http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training & Services
On 12/2/14 9:18 AM, Alvaro Herrera wrote: >> After this module has been moved to src/test/modules, I am seeing >> > failures with make check when manually kicking the test on at least >> > OSX where I tried because expected/ and sql/ are missing. pg_regress >> > does not like that much. That's annoying because check-world would >> > simply fail. > I had the directories locally but hadn't git-pushed them. make check-world still fails.
Peter Eisentraut wrote:
> On 12/2/14 9:18 AM, Alvaro Herrera wrote:
> >> After this module has been moved to src/test/modules, I am seeing
> >> > failures with make check when manually kicking the test on at least
> >> > OSX where I tried because expected/ and sql/ are missing. pg_regress
> >> > does not like that much. That's annoying because check-world would
> >> > simply fail.
> > I had the directories locally but hadn't git-pushed them.
>
> make check-world still fails.
Ah, it needs errcodes.h to be generated. We can add this (or similar)
to the makefiles:
OBJS: | submake-errcodes
submake-errcodes:
$(MAKE) -C $(top_builddir)/src/backend submake-errcodes
but it would be very repetitive to have it in all test modules'
makefiles. Why doesn't contrib need this? Do you have a better idea?
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
Peter Eisentraut wrote: > On 12/2/14 9:18 AM, Alvaro Herrera wrote: > >> After this module has been moved to src/test/modules, I am seeing > >> > failures with make check when manually kicking the test on at least > >> > OSX where I tried because expected/ and sql/ are missing. pg_regress > >> > does not like that much. That's annoying because check-world would > >> > simply fail. > > I had the directories locally but hadn't git-pushed them. > > make check-world still fails. It should work now, I just pushed a couple of fixes. -- Álvaro Herrera http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training & Services
On 12/2/14 3:46 PM, Alvaro Herrera wrote: > Ah, it needs errcodes.h to be generated. We can add this (or similar) > to the makefiles: > > OBJS: | submake-errcodes > > submake-errcodes: > $(MAKE) -C $(top_builddir)/src/backend submake-errcodes > > but it would be very repetitive to have it in all test modules' > makefiles. Why doesn't contrib need this? Do you have a better idea? The top-level GNUmakefile has a dependency to build src/ before contrib/.