[svn-r27713] Added -qflag=w:w to config/ibm-flags. This will suppress informational flags
on AIX, suppressing thousands of "__attribute__ is non-portable" messages
in the AIX nightly test output.
Tested on: Nothing. I have no access to an AIX machine.
[svn-r27695] HDFFV-0: Bug fix.
Problem: #if-#the-#else-#endif were added incorrectly to test_file_lock_concur()
such that the test would fail but did not execute a return statment. So, some
arbitary value is returned, causing confusion in the routine that called it.
Solution:
The #if-#the-#elif-... is unncessary. Fixed the #if-#endif so that they
are much shorter in scope and allowed the original return code to be
executed.
Tested: AIX ADA where it was discovered.
[svn-r27694] HDFFV-9521: arithmetic on a void* flagged as error by AIX XL compiler
Problem: image is a *void and xlc rejected the code of image+len.
Arithmentic on a void* is not legal in C but GCC extension allows it.
Solution: define image as a *uint_8 according to how it is used in the
code.
Tested: passed in AIX ADA machine.
[svn-r27693] HDFFV-0: bug fix
Problem: the script tried to launch the test command as "a.out" but not all userss
would have "." included in their $PATH for safety concerns.
Solution: Changed the mere "a.out" to "./a.out" so that the script knows how to launch
the test program.
Tested: AIX Ada (where I don't have '.' in my $PATH)
[svn-r27633] Removed pid check from Pflock() in H5system.c. This field is only
used when getting lock status.
tested on: emu (only system that uses fcntl in lieu of flock)
[svn-r27571] - h5watch tests are now skipped when the test VFD does not support SWMR.
- General improvements to the h5watch Makefile.am.
Tested on: jam w/ HDF5_DRIVER=stdio and sec2
[svn-r27522] CMake files match those of the trunk but still need to be updated with changes in this branch.
Tested on: NONE (CMake is still broken in this branch)
[svn-r27490] Disable flushrefresh test when VFD does not support SWMR.
Part of SWMR-23 and SWMR-25.
Tested on: Linux (jam), Darwin (wren) w/ threadsafe & autotools
[svn-r27486] Changed the "unsupported" VFD in test/swmr.c to be the stdio VFD.
We were previously using the family VFD and that should be
supported soon. I'd also like to explicitly test that the stdio
VFD is unsupported since it can never be supported under SWMR
due to the unpredictable nature of the buffering layer.
Added a comment to the stdio VFD query function to indicate
why SWMR is not supported.
Tested on: local linux VM (very minor change)
[svn-r27482] Fixes for revise_chunks daily test failures:
(1) test/swmr.c: test for H5_HAVE_FORK in test_file_lock_swmr_concur() as in test_file_lock_concur().
(2) test/cache_logging.c: (a) fix to catch proper return from the test (b) use h5_fixname() so check-vfd won't fail.
[svn-r27476] Fixes:
1. Fix for SWMR-88: move the 4 tests for file locking from tfile.c to swmr.c
2. H5SMcache.c: restore a line that got removed from previous merges.
3. H5Dproxy.c: delete; not needed
4. Refactor coding in H5C.c: H5C_make space_in_cache() and H5C_evict_tagged_entries()
Tested on jam, koala, ostrich, platypus, moohan, quail.
[svn-r27439] Merged r27428 from trunk.
Bug fix: DAILYTEST-170
Description:
XL compilers in ostrich (PowerPC64 linux) fail in test/dt_arith
because of the removal of the LLONG_TO_LDOUBLE_CORRECT (removed in
r26625) and LDOUBLE_TO_LLONG_ACCURATE (removed in r26623).
Solution:
Reverse revisions
r26623: bring back LDOUBLE_TO_LLONG_ACCURATE configure macro
r26625: bring back LLONG_TO_LDOUBLE_CORRECT configure macro
r26627: bring back WANT_DATA_ACCURACY configure macro which is...
[svn-r27414] Added platform-independent flock wrappers. When flock is not present on
the system, HDflock becomes a no-op. No Windows implementation of flock
has been included. Wflock is currently an empty function.
Only the Autotools are supported. CMake in this branch still needs a lot
of work.
Also fixed a few implicitly defined function warnings due to missing
headers.
Tested on: jam, emu (Solaris passes make check now)
[svn-r27411] Fixes some issues in the source:
- Renamed a few "h5tri_t extended" variables to "h5tri_t was_extended".
Apparently, "extended" is a typedef name someplace in the headers used
on Solaris, which causes potentially confusing warnings to be emitted.
- Added a missing H5F_get_vfd_handle() declaration to H5Fprivate.h.
- Fixed an error condition value in H5C.c.
Tested on: jam (passes)
emu (behaves correctly, but flock issues prevent building in this
branch)
[svn-r27408] Merge of r27407 from the trunk.
Updated the H5_ATTR_* macros.
- Updated the preprocessor logic that defines the H5_ATTR_* macros. The Solaris
Studio compiler understands __attribute__ but not the specific attributes
that we use. The H5_ATTR_* macros are now defined as blank with that
compiler which significantly cleans up the warnings, making debugging
easier.
- Replaced a few lingering __attribute__ macros with H5_ATTR_* markup.
Tested on: jam (passes)
emu...
[svn-r27370] (1) Fix for test/dsets.c failure when enable-debug=all
(2) Modify message when file failed to open due to file consistency flag
Tested on jam, ostrich, koala, platypus.
[svn-r27356] Fixes for revise_chunks daily test failures:
(1) Clear status flag for th5s.h5
(2) Close dataset/file in test_file_image.c
Tested on jam, koala, ostrich, moohan, quail.
[svn-r27344] Updated the SWMR acceptance tests to use h5_fileaccess() to obtain
a VFD-specific file access property list when running 'make check-vfd'.
Tested on: OS X 10.10.2 (Yosemite) w/ clang 6.1.0 (Xcode 6.3.2)
[svn-r27343] Changed test/test_swmr.c to swmr.c to be more in line with other test
names to and to make it more obvious that this is a unit test and not
run via the testswmr.sh script.
The H5FD.c changes are due to updated trace statements when reconfiguring.
Tested on: jam
[svn-r27338] Bring revision #24747 from revise_chksum_retry branch to revise_chunks.
Plus couple fixes:
1) test/test_swmr.c: fix to use H5Dopen2()
2) test/noencoder.h5: clear status_flags in superblock
