Add support for Sanitizer tools

[dongahn]

Initial support for AddressSanitizer (ASan) and ThreadSanitizer (TSan).

This is tested with clang-omp-3.5.0 installed at LLNL TOSS2 systems (e.g., Cab). For example,

% use clang-omp-3.5.0
% ./configure CC=clang

Also mildly tested with gcc 4.9.3p. For example,

% use gcc 4.9.3p
% ./configure

[coveralls]

Coverage increased (+0.09%) to 75.233% when pulling 965fbd9 on dongahn:new_san into 3efa493 on flux-framework:master.

[grondo]

Thanks @dongahn!

Minor suggestion for the leak fix commit 965fbd9. Instead of subject leak: Fix a minor leak... we should keep the prefix: to be the "subsystem" of the tree which the patch affects (and idiom we've been loosely following). In this case I'd suggest perhaps libutil: fix leak in nodeset test found by AddressSanitizer or test/libutil: fix leak in nodeset test found by AddressSanitizer.

It is nicer if a commit can be mostly understood (what is fixed/changed and why) from just the commit summary.

[dongahn]

Thanks. Yes, it makes sense. I will change it and force a push.

[dongahn]

OK. Pushed.

[coveralls]

Coverage decreased (-0.004%) to 75.136% when pulling 7edf40b on dongahn:new_san into 3efa493 on flux-framework:master.

[garlick]

This looks good to me. Could we have it rebased on current master? Then it can go in I think.

[coveralls]

Changes Unknown when pulling 2f0a1dd on dongahn:new_san into * on flux-framework:master*.

[garlick]

Ready @dongahn?

[dongahn]

Give me just one min. I am checking one thing.

[dongahn]

I was looking at a compilation error as clang 3.5 was complaining on the new cronodate.c. @garlick @grondo: Do you want me to add a fix to this as part of this PR?

  CC     test/test_cronodate_t-cronodate.o
test/cronodate.c:59:38: error: comparison of constant 0 with expression of type 'bool' is always true
      [-Werror,-Wtautological-constant-out-of-range-compare]
    ok (string_to_tm (expected, &tm) >= 0,
        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^  ~
../../../src/common/libtap/tap.h:48:56: note: expanded from macro 'ok'
#define ok(...)          ok_at_loc(__FILE__, __LINE__, __VA_ARGS__, NULL)
                                                       ^
test/cronodate.c:64:35: error: comparison of constant 0 with expression of type 'bool' is always true
      [-Werror,-Wtautological-constant-out-of-range-compare]
    ok (string_to_tm (start, &tm) >= 0,
        ~~~~~~~~~~~~~~~~~~~~~~~~~ ^  ~
../../../src/common/libtap/tap.h:48:56: note: expanded from macro 'ok'
#define ok(...)          ok_at_loc(__FILE__, __LINE__, __VA_ARGS__, NULL)
                                                       ^
test/cronodate.c:222:51: error: comparison of constant 0 with expression of type 'bool' is always true
      [-Werror,-Wtautological-constant-out-of-range-compare]
    ok (string_to_tm ("2016-06-06 08:00:00", &tm) >= 0, "string_to_tm");
        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^  ~
../../../src/common/libtap/tap.h:48:56: note: expanded from macro 'ok'
#define ok(...)          ok_at_loc(__FILE__, __LINE__, __VA_ARGS__, NULL)
                                                       ^
test/cronodate.c:231:53: error: comparison of constant 0 with expression of type 'bool' is always true
      [-Werror,-Wtautological-constant-out-of-range-compare]
    ok (string_to_tv ("2016-06-06 07:00:00.3", &tv) >= 0, "string_to_tv");
        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^  ~
../../../src/common/libtap/tap.h:48:56: note: expanded from macro 'ok'
#define ok(...)          ok_at_loc(__FILE__, __LINE__, __VA_ARGS__, NULL)
                                                       ^
test/cronodate.c:235:51: error: comparison of constant 0 with expression of type 'bool' is always true
      [-Werror,-Wtautological-constant-out-of-range-compare]
    ok (string_to_tv ("2016-06-06 08:00:00", &tv) >= 0, "string_to_tv");
        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^  ~
../../../src/common/libtap/tap.h:48:56: note: expanded from macro 'ok'
#define ok(...)          ok_at_loc(__FILE__, __LINE__, __VA_ARGS__, NULL)

[garlick]

Oh good catch. No objection if you want to tack a fix on to this PR, or defer till later - your call.

[grondo]

Ooh, yeah @dongahn can you go ahead and fix that? Sloppy conversion on my part from the old approxidate functions I guess (or maybe it was that way all along, in any case that is a good fix)

[coveralls]

Changes Unknown when pulling 7f5331c on dongahn:new_san into * on flux-framework:master*.

[dongahn]

@garlick: Ok. I think this is ready.

[garlick]

Thanks!