test-distributions is broken on remotes/origin/develop

[Ben Goodrich]

It has started trying to compile things like

src/test/agrad/distributions/--help_00000_generated_test.cpp

for some reason.

goodrich@CYBERPOWERPC:/opt/stan$ make CC=clang++ -j8 test-distributions
make[1]: Entering directory `/opt/stan'
src/test/agrad/distributions/no_00000_generated_test.cpp:4:10: fatal error: 'mal_test' file not found
#include <mal_test>
         ^
1 error generated.
src/test/agrad/distributions/normal_00000_generated_test.cpp:4:10: fatal error: 'mal' file not found
#include <mal>
         ^
1 error generated.
src/test/agrad/distributions/--help_00000_generated_test.cpp:4:10: fatal error: 'elp' file not found
#include <elp>
         ^
1 error generated.
make[1]: Leaving directory `/opt/stan'
make[1]: Entering directory `/opt/stan'
src/test/agrad/distributions/normal_00000_generated_test.cpp:4:10: fatal error: 'mal' file not found
#include <mal>
         ^
1 error generated.
src/test/agrad/distributions/no_00000_generated_test.cpp:src/test/agrad/distributions/--help_00000_generated_test.cpp:4:10: 4:fatal error: 10: fatal error: 'mal_test' file not found
'elp' file not found
#include <mal_test>
#include <elp>
         ^         ^

1 error generated.
1 error generated.
clang++ -I src -I lib/eigen_3.1.3 -I lib/boost_1.53.0 -pipe -Wall -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE -DBOOST_DISABLE_ASSERTS -Wno-unused-function -Wno-uninitialized   -c -O3 -DGTEST_USE_OWN_TR1_TUPLE -DGTEST_HAS_PTHREAD=0 -I lib/gtest_1.6.0/include -I lib/gtest_1.6.0 src/test/agrad/distributions/normal_00000_generated_test.cpp -o test/agrad/distributions/normal_00000_generated.o
clang++ -I src -I lib/eigen_3.1.3 -I lib/boost_1.53.0 -pipe -Wall -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE -DBOOST_DISABLE_ASSERTS -Wno-unused-function -Wno-uninitialized   -c -O3 -DGTEST_USE_OWN_TR1_TUPLE -DGTEST_HAS_PTHREAD=0 -I lib/gtest_1.6.0/include -I lib/gtest_1.6.0 src/test/agrad/distributions/--help_00000_generated_test.cpp -o test/agrad/distributions/--help_00000_generated.o
clang++ -I src -I lib/eigen_3.1.3 -I lib/boost_1.53.0 -pipe -Wall -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE -DBOOST_DISABLE_ASSERTS -Wno-unused-function -Wno-uninitialized   -c -O3 -DGTEST_USE_OWN_TR1_TUPLE -DGTEST_HAS_PTHREAD=0 -I lib/gtest_1.6.0/include -I lib/gtest_1.6.0 src/test/agrad/distributions/no_00000_generated_test.cpp -o test/agrad/distributions/no_00000_generated.o
clang++ -I src -I lib/eigen_3.1.3 -I lib/boost_1.53.0 -pipe -Wall -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE -DBOOST_DISABLE_ASSERTS -Wno-unused-function -Wno-uninitialized   -c -O3 -DGTEST_USE_OWN_TR1_TUPLE -DGTEST_HAS_PTHREAD=0 -I lib/gtest_1.6.0/include -I lib/gtest_1.6.0 src/test/agrad/distributions/univariate/continuous/exp_mod_normal_00025_generated_test.cpp -o test/agrad/distributions/univariate/continuous/exp_mod_normal_00025_generated.o
clang++ -I src -I lib/eigen_3.1.3 -I lib/boost_1.53.0 -pipe -Wall -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE -DBOOST_DISABLE_ASSERTS -Wno-unused-function -Wno-uninitialized   -c -O3 -DGTEST_USE_OWN_TR1_TUPLE -DGTEST_HAS_PTHREAD=0 -I lib/gtest_1.6.0/include -I lib/gtest_1.6.0 src/test/agrad/distributions/univariate/continuous/beta_00005_generated_test.cpp -o test/agrad/distributions/univariate/continuous/beta_00005_generated.o
clang++ -I src -I lib/eigen_3.1.3 -I lib/boost_1.53.0 -pipe -Wall -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE -DBOOST_DISABLE_ASSERTS -Wno-unused-function -Wno-uninitialized   -c -O3 -DGTEST_USE_OWN_TR1_TUPLE -DGTEST_HAS_PTHREAD=0 -I lib/gtest_1.6.0/include -I lib/gtest_1.6.0 src/test/agrad/distributions/univariate/continuous/gamma_00004_generated_test.cpp -o test/agrad/distributions/univariate/continuous/gamma_00004_generated.o
clang++ -I src -I lib/eigen_3.1.3 -I lib/boost_1.53.0 -pipe -Wall -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE -DBOOST_DISABLE_ASSERTS -Wno-unused-function -Wno-uninitialized   -c -O3 -DGTEST_USE_OWN_TR1_TUPLE -DGTEST_HAS_PTHREAD=0 -I lib/gtest_1.6.0/include -I lib/gtest_1.6.0 src/test/agrad/distributions/univariate/continuous/pareto_00002_generated_test.cpp -o test/agrad/distributions/univariate/continuous/pareto_00002_generated.o
clang++ -I src -I lib/eigen_3.1.3 -I lib/boost_1.53.0 -pipe -Wall -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE -DBOOST_DISABLE_ASSERTS -Wno-unused-function -Wno-uninitialized   -c -O3 -DGTEST_USE_OWN_TR1_TUPLE -DGTEST_HAS_PTHREAD=0 -I lib/gtest_1.6.0/include -I lib/gtest_1.6.0 src/test/agrad/distributions/univariate/continuous/skew_normal_cdf_00001_generated_test.cpp -o test/agrad/distributions/univariate/continuous/skew_normal_cdf_00001_generated.o
src/test/agrad/distributions/no_00000_generated_test.cpp:4:10: fatal error: 'mal_test' file not found
#include <mal_test>
         ^
src/test/agrad/distributions/--help_00000_generated_test.cpp:4:10: fatal error: 'elp' file not found
#include <elp>
         ^
src/test/agrad/distributions/normal_00000_generated_test.cpp:4:10: fatal error: 'mal' file not found
#include <mal>
         ^
1 error generated.
make[1]: *** [test/agrad/distributions/no_00000_generated.o] Error 1
make[1]: *** Waiting for unfinished jobs....
1 error generated.
make[1]: *** [test/agrad/distributions/--help_00000_generated.o] Error 1
1 error generated.
make[1]: *** [test/agrad/distributions/normal_00000_generated.o] Error 1
make[1]: Leaving directory `/opt/stan'
make: *** [test-distributions] Error 2

Ben

[Ben Goodrich]

On Monday, May 27, 2013 1:58:58 PM UTC-4, Ben Goodrich wrote:

It has started trying to compile things like

src/test/agrad/distributions/--help_00000_generated_test.cpp

for some reason.

test-unit is also messed up, but

make clean-all > /dev/null && make CC=clang++ -j8 O=0 test/agrad/distributions/normal_00000_generated

is sufficient to trigger it. It seems to be related to the test/model/command breakage.

Ben

[Daniel Lee]

Hi Ben,

I'm not seeing any of that behavior. Can you get some more info? And you don't have any local modifications to any of the makefiles?

I don't expect using test/agrad/distributions/normal_00000_generated to work: it's not a valid target. Does this one work?
test/agrad/distributions/univariate/continuous/normal_00000_generated

Also, test/model/command shouldn't work. It should be test/models/command

Can you update and try again? It's working locally for me and on our Windows test box. I'd like to find out under what conditions our makefile is failing.

Daniel

 

[Ben Goodrich]

On Monday, May 27, 2013 10:08:32 PM UTC-4, Daniel Lee wrote:

I'm not seeing any of that behavior. Can you get some more info? And you don't have any local modifications to any of the makefiles?

No modifications; just

git checkout remotes/origin/develop
make clean-all > /dev/null
make CC=clang++ -j8 bin/stanc 2&> /dev/null
make CC=clang++ -j8 test-distributions

I don't expect using test/agrad/distributions/normal_00000_generated to work: it's not a valid target. Does this one work?

test/agrad/distributions/univariate/continuous/normal_00000_generated

Also, test/model/command shouldn't work. It should be test/models/command

That is the breakage: It tries to compile targets that don't exist like

test/agrad/distributions/no_00000_generated
test/agrad/distributions/--help_00000_generated

which inevitably result in errors like

src/test/agrad/distributions/no_00000_generated_test.cpp:4:10: fatal error: 'mal_test' file not found
#include <mal_test>
         ^
src/test/agrad/distributions/--help_00000_generated_test.cpp:4:10: fatal error: 'elp' file not found
#include <elp>
         ^

or you can try something like make CC=clang++ test/agrad/distributions/normal_00000_generated and instead of saying that is not a valid make target, it says

src/test/agrad/distributions/normal_00000_generated_test.cpp:4:10: fatal error: 'mal' file not found
#include <mal>

Ben
 

[Daniel Lee]

I can't recreate it on my end. I check out remotes/origin/develop and get this message:

~/dev/stan (develop)> git checkout remotes/origin/develop
Note: checking out 'remotes/origin/develop'.

You are in 'detached HEAD' state. You can look around, make experimental
changes and commit them, and you can discard any commits you make in this
state without impacting any branches by performing another checkout.

If you want to create a new branch to retain commits you create, you may
do so (now or later) by using -b with the checkout command again. Example:

  git checkout -b new_branch_name

HEAD is now at 2c593e4... Merge pull request #86 from randommm/feature/sort

I'm running the distribution tests again, but so far, it's not trying to build any of those targets that don't exist.

What version of make are you running? I'm on GNU Make 3.81.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Ben Goodrich]

On Monday, May 27, 2013 10:55:13 PM UTC-4, Daniel Lee wrote:

I can't recreate it on my end. I check out remotes/origin/develop and get this message:

~/dev/stan (develop)> git checkout remotes/origin/develop
Note: checking out 'remotes/origin/develop'.

Right, that is what is supposed to happen.

What version of make are you running? I'm on GNU Make 3.81.

I also have 3.81. You can pass the -n parameter to make test-distribution to just see what it is planning to do. The runtest/test/agrad/distributions/--help_00000_generated and runtest/test/agrad/distributions/no_00000_generated "targets" are two of the first three.

Ben

[Ben Goodrich]

On Monday, May 27, 2013 11:10:18 PM UTC-4, Ben Goodrich wrote:

I also have 3.81. You can pass the -n parameter to make test-distribution to just see what it is planning to do. The runtest/test/agrad/distributions/--help_00000_generated and runtest/test/agrad/distributions/no_00000_generated "targets" are two of the first three.

Another useful thing is

make CC=clang++ -p test-distributions

which ultimately tells me

# makefile (from `test/agrad/distributions/univariate/continuous/student_t_cdf_00029_generated.d', line 1)
MAKEFILE_LIST :=  makefile make/detect_cc make/os_detect make/os_linux make/libstan bin/stan/agrad/rev/var_stack.d bin/stan/math/matrix.d bin/stan/agrad/matrix.d bin/stan/gm/grammars/expression_grammar_inst.d bin/stan/gm/grammars/whitespace_grammar_inst.d bin/stan/gm/grammars/program_grammar_inst.d bin/stan/gm/grammars/statement_grammar_inst.d bin/stan/gm/grammars/var_decls_grammar_inst.d bin/stan/gm/grammars/term_grammar_inst.d bin/stan/gm/grammars/statement_2_grammar_inst.d bin/stan/gm/ast_def.d make/tests make/models src/stan/model/model_header.hpp.d make/command bin/stan/command/stanc.d make/doxygen make/dist make/manual test/agrad/distributions/normal_00000_generated.d test/agrad/distributions/--help_00000_generated.d test/agrad/distributions/no_00000_generated.d test/agrad/distributions/
...

So maybe there is something about student_t_cdf_00029_generated?

Ben

[Daniel Lee]

On Mon, May 27, 2013 at 11:10 PM, Ben Goodrich <goodri...@gmail.com> wrote:

I also have 3.81. You can pass the -n parameter to make test-distribution to just see what it is planning to do. The runtest/test/agrad/distributions/--help_00000_generated and runtest/test/agrad/distributions/no_00000_generated "targets" are two of the first three.

I'm not seeing any of those targets. Can you run it with -n? Here are the first few that I get:

~/dev/stan ((2c593e4...))> make test-distributions -n
make runtest/test/agrad/distributions/univariate/continuous/beta_00000_generated runtest/test/agrad/distributions/univariate/continuous/beta_00001_generated runtest/test/agrad/distributions/univariate/continuous/beta_00002_generated runtest/test/agrad/distributions/univariate/continuous/beta_00003_generated runtest/test/agrad/distributions/univariate/continuous/beta_00004_generated runtest/test/agrad/distributions/univariate/continuous/beta_00005_generated runtest/test/agrad/distributions/univariate/continuous/beta_cdf_00000_generated runtest/test/agrad/distributions/univariate/continuous/beta_cdf_00001_generated runtest/test/agrad/distributions/univariate/continuous/beta_cdf_00002_generated runtest/test/agrad/distributions/univariate/continuous/beta_cdf_00003_generated runtest/test/agrad/distributions/univariate/continuous/beta_cdf_00004_generated runtest/test/agrad/distributions/univariate/continuous/beta_cdf_00005_generated runtest/test/agrad/distributions/univariate/continuous/cauchy_00000_generated runtest/test/agrad/distributions/univariate/continuous/cauchy_00001_generated runtest/test/agrad/distributions/univariate/continuous/cauchy_00002_generated runtest/test/agrad/distributions/univariate/continuous/cauchy_00003_generated runtest/test/agrad/distributions/univariate/continuous/cauchy_00004_generated runtest/test/agrad/distributions/univariate/continuous/cauchy_00005_generated

[Daniel Lee]

I can run student_t_cdf_00029_generated fine.

Just to rule out other weird interactions, your make isn't aliased to encapsulate other commands, is it?

[Ben Goodrich]

On Monday, May 27, 2013 11:20:54 PM UTC-4, Daniel Lee wrote:

I can run student_t_cdf_00029_generated fine.

Just to rule out other weird interactions, your make isn't aliased to encapsulate other commands, is it?

I can also do test/agrad/distributions/univariate/continuous/student_t_cdf_00029_generated and test/agrad/distributions/univariate/continuous/student_t_cdf_00029_generated.d fine. I don't have any aliases on make. Possibly it is a locales or encoding thing? When I do it with -n, my make targets come out in a different order than yours (in addition to having the bogus ones)

make runtest/test/agrad/distributions/normal_00000_generated runtest/test/agrad/distributions/--help_00000_generated runtest/test/agrad/distributions/no_00000_generated runtest/test/agrad/distributions/univariate/continuous/exp_mod_normal_00025_generated runtest/test/agrad/distributions/univariate/continuous/beta_00005_generated runtest/test/agrad/distributions/univariate/continuous/gamma_00004_generated runtest/test/agrad/distributions/univariate/continuous/pareto_00002_generated runtest/test/agrad/distributions/univariate/continuous/skew_normal_cdf_00001_generated runtest/test/agrad/distributions/univariate/continuous/uniform_00001_generated runtest/test/agrad/distributions/univariate/continuous/skew_normal_cdf_00030_generated runtest/test/agrad/distributions/univariate/continuous/skew_normal_cdf_00033_generated runtest/test/agrad/distributions/univariate/continuous/double_exponential_00004_generated runtest/test/agrad/distributions/univariate/continuous/exp_mod_normal_cdf_00009_generated runtest/test/agrad/distributions/univariate/continuous/cauchy_00000_generated runtest/test/agrad/distributions/univariate/continuous/exp_mod_normal_00024_generated runtest/test/agrad/distributions/univariate/continuous/student_t_cdf_00016_generated runtest/test/agrad/distributions/univariate/continuous/student_t_00020_generated runtest/test/agrad/distributions/univariate/continuous/normal_cdf_00002_generated runtest/test/agrad/distributions/univariate/continuous/exp_mod_normal_00022_generated runtest/test/agrad/distributions/univariate/continuous/exp_mod_normal_00033_generated runtest/test/agrad/distributions/univariate/continuous/cauchy_cdf_00000_generated runtest/test/agrad/distributions/univariate/continuous/skew_normal_cdf_00026_generated runtest/test/agrad/distributions/univariate/continuous/exp_mod_normal_cdf_00017_generated runtest/test/agrad/distributions/univariate/continuous/exp_mod_normal_cdf_00033_generated runtest/test/agrad/distributions/univariate/continuous/student_t_00018_generated ...

Ben

[Daniel Lee]

Yikes! If you take a look at the test-distributions target, you'll see that things should be generated in alphabetical order. This is the block in the make/tests makefile that is relevant.


GENERATED_AGRAD_TESTS := $(shell find src/test/agrad/distributions -type f -name '*_test.hpp')
GENERATED_AGRAD_TESTS := $(GENERATED_AGRAD_TESTS:%_test.hpp=%_00000_generated_test.cpp)
##
# Agrad distributions test generator
##
test/agrad/distributions/generate_tests$(EXE) : src/test/agrad/distributions/generate_tests.cpp
    @mkdir -p $(dir $@)
    $(LINK.c) -O$(O_STANC) $(CFLAGS) $< $(OUTPUT_OPTION)

src/test/agrad/distributions/%_00000_generated_test.cpp : test/agrad/distributions/generate_tests$(EXE) src/test/agrad/distributions/%_test.hpp
    @echo "--- Generating tests for $(word 2,$^) ---"
    test/agrad/distributions/generate_tests$(EXE) $(word 2,$^)

.PHONY: test-distributions
test-distributions: $(GENERATED_AGRAD_TESTS)
    @$(MAKE) $(patsubst src/%_test.cpp,runtest/%,$(shell find src/test/agrad/distributions -type f -name '*_generated_test.cpp'))

Can you run a find from your shell? Does it find the right things? Could something not be escaped properly? Is the use of $(MAKE) within the test-distributions target messing things up? Could you try without -j8?



Daniel

[Ben Goodrich]

Can you run a find from your shell? Does it find the right things? Could something not be escaped properly? Is the use of $(MAKE) within the test-distributions target messing things up? Could you try without -j8?

I doubt $(MAKE) is a problem; usually the problem is the absence of a $(MAKE). My GENERATED_AGRAD_TESTS is not in a recognizeable order, regardless of the -j, but it is still trying to compile a corrupted file from a bogus target

goodrich@CYBERPOWERPC:/opt/stan$ make CC=g++ -p test-distributions > /tmp/make_p.txt
src/test/agrad/distributions/normal_00000_generated_test.cpp:4:15: fatal error: mal: No such file or directory
 #include <mal>
               ^
compilation terminated.

make[1]: *** [test/agrad/distributions/normal_00000_generated.o] Error 1

make: *** [test-distributions] Error 2
goodrich@CYBERPOWERPC:/opt/stan$ grep -F "GENERATED_AGRAD_TESTS" /tmp/make_p.txt
GENERATED_AGRAD_TESTS := src/test/agrad/distributions/univariate/continuous/beta_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/student_t_cdf_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/pareto_cdf_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/pareto_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/normal_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/uniform_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/student_t_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/logistic_cdf_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/lognormal_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/inv_chi_square_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/double_exponential_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/inv_gamma_cdf_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/logistic_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/exp_mod_normal_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/inv_gamma_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/skew_normal_00000_generated_test.cpp src/test/agrad/distributions/univariate/continuous/beta_cdf_00000_generated_test.cp

But that is how it comes out of my find.

goodrich@CYBERPOWERPC:/opt/stan$ find src/test/agrad/distributions -type f -name '*_test.hpp' -print | head
src/test/agrad/distributions/univariate/continuous/beta_test.hpp
src/test/agrad/distributions/univariate/continuous/student_t_cdf_test.hpp
src/test/agrad/distributions/univariate/continuous/pareto_cdf_test.hpp
src/test/agrad/distributions/univariate/continuous/pareto_test.hpp
src/test/agrad/distributions/univariate/continuous/normal_test.hpp
src/test/agrad/distributions/univariate/continuous/uniform_test.hpp
src/test/agrad/distributions/univariate/continuous/student_t_test.hpp
src/test/agrad/distributions/univariate/continuous/logistic_cdf_test.hpp
src/test/agrad/distributions/univariate/continuous/lognormal_test.hpp
src/test/agrad/distributions/univariate/continuous/inv_chi_square_test.hpp

Ben

[Daniel Lee]

Do you have access to a box that I can remote into? I'm running out of ideas. 

[Ben Goodrich]

On Tuesday, May 28, 2013 12:16:22 AM UTC-4, Daniel Lee wrote:

Do you have access to a box that I can remote into? I'm running out of ideas. 

I'm pretty convinced that this is triggered by some change on my system since I keep getting the same errors if I go further back. Unfortunately, I have no idea what that might be and don't really log such changes all that carefully. I'll try to figure it out tomorrow.

Ben

[Ben Goodrich]

Upgrading to make 3.82 didn't help. This is going to be a mess to figure out. Here are the three, 4-line test files that generate_tests generates in (not under) src/test/agrad/distributions/ before bombing:

--help_00000_generated_test.cpp

#include <gtest/gtest.h>
#include <boost/mpl/vector.hpp>
#include <test/agrad/distributions/test_fixture.hpp>
#include <elp>

no_00000_generated_test.cpp

#include <gtest/gtest.h>
#include <boost/mpl/vector.hpp>
#include <test/agrad/distributions/test_fixture.hpp>
#include <mal_test>

normal_00000_generated_test.cpp

#include <gtest/gtest.h>
#include <boost/mpl/vector.hpp>
#include <test/agrad/distributions/test_fixture.hpp>
#include <mal>

But generate_tests also generates a lot of .cpp files that seem to be right (i.e., they are 300 lines long). Does anyone have an idea about what would cause generate_tests to mess up?

Ben

[Ben Goodrich]

On Tuesday, May 28, 2013 7:03:32 PM UTC-4, Ben Goodrich wrote:

On Tuesday, May 28, 2013 1:45:49 AM UTC-4, Ben Goodrich wrote:

On Tuesday, May 28, 2013 12:16:22 AM UTC-4, Daniel Lee wrote:

Do you have access to a box that I can remote into? I'm running out of ideas. 

I'm pretty convinced that this is triggered by some change on my system since I keep getting the same errors if I go further back. Unfortunately, I have no idea what that might be and don't really log such changes all that carefully. I'll try to figure it out tomorrow.

Upgrading to make 3.82 didn't help. This is going to be a mess to figure out. Here are the three, 4-line test files that generate_tests generates in (not under) src/test/agrad/distributions/ before bombing:

I just manually deleted those files, and now test-distributions is fine. We might never figure out what caused them to get generated in the first place.

Ben

[Daniel Lee]

I'm ok with not knowing. Let me know if you end up generating them again.