Bug Fix
Description:
When writing (or reading) the entire dataset to a chunked dataset, there
was a boundary case where the code to generate the description of the
piece of the dataset to read into the buffer for data conversion would
attempt to read off the boundary of the dataset. This was occuring because
the code to detect the edge of the dataset was not propagating the change
up through the remaining dimensions when an edge in a fast changing
dimension was detected.
Solution:
Propagate edge detection up through slower changing dimensions properly.
Platforms tested:
Linux 2.2.18smp (eirene)
Fix on Kludge
Description:
Forgot another chunk of parallel I/O code that needed to change for the
generic property list kludge... :-/
Platforms tested:
Parallel Linux 2.2.18smp (eirene)
Kludge
Description:
Since we're only about halfway through converting the internal use of
property lists from the "old way" to the generic property lists, we turned
off snapshots to avoid exposing lots of API changes to users, until the
APIs settled down.
Getting the snapshots rolling again seems to have become a priority, so
some changes are going to have to be made now that were going to be
postponed until we were completely finished with the conversion. This
requires that the old API functions be able to deal with both the old
and new property lists smoothly.
Solution:
Kludge together the property list code so that they can transparently handle
dealing with both the old and new property lists
Platforms tested:
FreeBSD 4.4 (hawkwind)
Code cleanup for better compatibility with C++ compilers
Description:
C++ compilers are choking on our C code, for various reasons:
we used our UNUSED macro incorrectly when referring to pointer types
we used various C++ keywords as variables, etc.
we incremented enum's with the ++ operator.
Solution:
Changed variables, etc.to avoid C++ keywords (new, class, typename, typeid,
template)
Fixed usage of UNUSED macro from this:
char UNUSED *c
to this:
char * UNUSED c
Switched the enums from x++ to x=x+1
Platforms tested:
FreeBSD 4.4 (hawkwind)
Feature
Description:
The H5Pset_fapl_split() used to assume both meta and raw files
have the same prefix name. This prohibited the option to specify
the two files reside on different file systems. E.g., having
meta-file as /user/home/data1.meta and raw-file as /pfs/data1.raw.
Solution:
Added feature that if the raw or meta extension string contains
a "%s", it will be substituted by the filename given for H5Fopen
or H5Fcreate. This is same as the multi-file syntax. If no %s
is found, one is inserted at the beginning. This is the previous
behavior.
Platforms tested:
Modi4 (parallel, serial) and eirene (parallel).
Adding Feature
Description:
Added new H5Pcopy_prop function to copy a property (value) from one property
list or class to another.
Platforms tested:
FreeBSD 4.4 (hawkwind)
Thread-safety Bug Fixes
Description:
FreeBSD must use a different layout or mechanism for allocating objects on
the stack when compiling for thread-safe operation. Unused bits in 'long
double' variable are never used ('long double's on FreeBSD are stored in
12 bytes, but only use 10 bytes for computations), but set to different
values, causing the endianness permutation testing to generate incorrect
results and eventually fail an assertion.
Solution:
Clear the temporary variables used for determining the endianness permutation
with memset (affecting all 12 bytes) instead of just assigning a 0.0 to them
(affecting only 10 bytes)
Platforms tested:
FreeBSD 4.4 (hawkwind)
Purpose:
Bug Fix
Description:
Was getting warnings about converting pointers to integers
without a cast.
Solution:
Changed some of the RETURN ERRORs from returning NULL to
returning FAIL instead.
Platforms tested:
Linux
Bug fix
Description:
SAF test code exposed an error in the HDF5_MPI_OPT_TYPES
code. The SAF code was doing collective write to chunked
storage dataset. Some processes wanted to flush some chunk
while some other processes were doing something else but the
HDF5_MPI_OPT_TYPES code thought the chunk flushing were collective
calls since it only looked at the condition when H5Dwrite
was called. So, it hanged when doing MPI_File_setview.
For now, turned off the HDF5_MPI_OPT_TYPES code so that the
SAF code would work. More long term fix later.
Platforms tested:
modi4(pp)
bug fix
Description:
H5FD_mpio_flush() would try to file seek negative if the
file->eoa is 0 (e.g., doing mpio on the raw-file of the
split file driver). Put in a code to catch this case
by returning succeed immediately.
Platforms tested:
modi4(pp)
Bug Fix
Description:
Byte swapping routine was core dumping when attempting to convert 0
elements.
Solution:
Don't try to byte swap when there is nothing to do... :-)
Apologies to Kent for my comments on Friday about testing his checkins,
especially when it was I who needed to do moer testing. :-/
Platforms tested:
Solaris 2.7 (arabica)
Code improvement
Description:
The byte swapping routine for data conversion was inefficient.
Solution:
Applied a number of optimizations which should yield around a 2-3 times
faster algorithm.
Platforms tested:
Solaris 2.6 (baldric)
Code cleanup (sorta)
Description:
When the first versions of the HDF5 library were designed, I remembered
vividly the difficulties of porting code from a 32-bit platform to a 16-bit
platform and asked that people use intn & uintn instead of int & unsigned
int, respectively. However, in hindsight, this was overkill and
unnecessary since we weren't going to be porting the HDF5 library to
16-bit architectures.
Currently, the extra uintn & intn typedefs are causing problems for users
who'd like to include both the HDF5 and HDF4 header files in one source
module (like Kent's h4toh5 library).
Solution:
Changed the uintn & intn's to unsigned and int's respectively.
Platforms tested:
FreeBSD 4.4 (hawkwind)
Bug Fix
Description:
The datatype conversion buffer size was only halfway converted from hsize_t
to size_t and was causing problems. Also, the H5P_peek_<foo> routines have
a bunch of identical code.
Solution:
Finished converting the datatype conversion buffer size to size_t and
cleaned up the implementation of the H5P_peek_<foo> routines to call a
common routine instead of reimplementing a bunch of code.
Platforms tested:
Solaris 2.7 (arabica)
More code cleanups
Description:
Wrap up the code cleanups for changing the dataset transfer property lists
over to using the generic property list code.
Platforms tested:
IRIX64 6.5 (modi4)
Code cleanups, mostly..
Description:
Work on pacifying the SGI compiler to get the generic properties working
correctly with --enable-parallel and --enable-fortran. It's not quite
fixed yet, but I need to head home and these patches help... :-/
Platforms tested:
IRIX64 6.5 (modi4)
New Features!
Description:
Start migrating the internal use of property lists in the library from the
older implementation to the new generic property lists.
Currently, only the dataset transfer property lists are migrated to the
new architecture, all the rest of the property list types are still using
the older architecture.
Also, the backward compatibility features are not implemented yet, so
applications which use dataset transfer properties may need to make the
following changes:
H5Pcreate(H5P_DATASET_XFER) -> H5Pcreate_list(H5P_DATASET_XFER_NEW)
and
H5Pclose(<a dataset transfer property list>) -> H5Pclose_list(id)
This still may have some bugs in it, especially with Fortran, but I should
be wrapping up those later today.
Platforms tested:
FreeBSD 4.4 (hawkwind)
Feature shift
Description:
Take out the v1.2.x compatibility stubs and put in the hooks for v1.4.x
compatibility when needed.
Platforms tested:
FreeBSD 4.3 (hawkwind)
Bug Fix
Description:
H5FD_fapl_copy is being called in H5Pset_driver when copying a dataset
transfer property list instead of H5FD_dxpl_copy and could potentially
cause problems if the file driver information was different.
Solution:
Changed call to H5FD_dxpl_copy() until the generic property code gets
merged in.
Platforms tested:
FreeBSD 4.3 (hawkwind)
Purpose:
Bug Fix
Description:
The way we were generating Dependencies and .depend files was broken.
If the $srcdir or other macros began with a ".", then it would match
anything and cause problems since it would then overwrite the
beginning of the header file's path.
Solution:
Wrote a Perl script which can handle this type of weirdness better.
It's only used when the environment is a GNU one with a GCC
compiler...
Platforms tested:
Linux
Bug Fix
Description:
H5Pset_driver had a resource leak which was dropping dataset transfer IDs
when switching drivers.
Solution:
Decrement dataset transfer ID reference count properly
Platforms tested:
IRIX64 6.5 (modi4)
Bug Fix
Description:
Header file is using incorrect macro for detecting parallel I/O when
backward compability is turned on.
Solution:
Fix header files to define prototypes for parallel I/O property list
functions when compiled with v1.2 backward compatibility turned on.
Platforms tested:
Eyeballed... (Elena will be testing shortly)
Bug Fix
Description:
In certain circumstances, raw data was inadvertantly attempted to be read
from the metadata cache. This was caught with an assertion failure (i.e.
core dump) in the development branch or an eventual sequence of errors in
the release branch.
Solution:
Corrected off-by-one error in metadata caching code.
Platforms tested:
FreeBSD 4.3 (hawkwind)
Purpose:
Bug Fix
Description:
Forward port of the bug fix to the 1.4 branch reguarding the C++
compile error.
Solution:
Need to put
#ifdef __cplusplus
extern "C" {
#endif
/* ... */
#ifdef __cplusplus
}
#endif
around function declarations.
Code Cleanup/Bug Fixes
Description:
Raymond's last fix corrected the problem with eirene, but there were still
a number of warnings, etc. during compiles.
Solution:
Clean up the warnings, neaten up the code a little and change a few hsize_t's
into size_t's that were missed in my previous sweep through the code.
Platforms tested:
Linux 2.2 (eirene)
Purpose:
Bug fix.
Description:
Daily test fails in parallel collective write testing on eirene.
Solution:
In function H5S_mpio_spaces_xfer in H5Smpio.c, changed mpi_count
from hsize_t to size_t.
Platforms tested:
MPICH(Linux, SunOS, IRIX64), SGI MPI.
Bug fix.
Description:
On some systems (linux when not using gcc) 'dev_t' is not actually
a scalar variable. This causes the code which compares dev_t's in
the file drivers to not compile. Also the H5_inline flag was not being
set correctly in the H5private.h file.
Solution:
Set the H5_inline flag to '' (i.e. define it, but don't assign it a value)
if it is not currently defined.
Use DEV_T_IS_SCALAR flag from configure to correctly compare dev_t's using
memcmp instead of a scalar flag.
Platforms tested:
FreeBSD 4.3 (hawkwind), Linux 2.4.2 (chiba city cluster at Argonne)
Bug fix... (sorta)
Description:
Added --enable-linux-lfs configure flag to have better control over whether
the enable large file support on Linux machines. Also removed the
-malign-double flag for gcc since it can potentially cause errors which
are difficult to detect.
Platforms tested:
Linix 2.2 & 2.4 (eirene and dangermouse)
Bug Fix
Description:
Using a 'long long' type (i.e. hsize_t) in a switch statement is apparently
not allowed by the HP/UX 10.20 system compiler and causes the compile to
fail.
Solution:
Cast the variable to a size_t for the switch.
Platforms tested:
HP/UX 10.20 (sangamon)
Bug Fix, Code Cleanup, Code Optimization, etc.
Description:
Fold in the hyperslab speedups, clean up compile warnings and change a
few things from using 'unsigned' or 'hsize_t' to use 'size_t' instead.
Platforms tested:
FreeBSD 4.3 (hawkwind), Solaris 2.7 (arabica), Irix64 6.5 (modi4)
Documentation
Description:
H5Tget_member_type in v1.2.x returns the base type of array fields in
compound datatypes, not an array type itself.
Solution:
Changed to emulate this behavior in v1.4.x when the --enable-hdf5v1_2
flag is given during configuration.
Platforms tested:
FreeBSD 4.3 (hawkwind)
Code cleanup
Description:
File driver macros were defined in this file, instead of the file driver
source file.
Solution:
Removed redundant file driver information, which is already in each file
driver source file.
Platforms tested:
FreeBSD 4.3 (hawkwind)
Purpose:
Code Cleanup
Description:
Updating CodeWarrior Port
Solution:
Move around some Windows and Metroworks ifdefs, etc.
Platforms tested:
Solaris 2.7 (arabica) Linux 2.2 (eirene)
Purpose:
Implemented port hunting
Really catch SIGPIPE signals
Description:
Port hunting allows to try and bind to a successive port number
if the port number given in the filename is already used.
Because of a stupid typo in the code, SIGPIPE wasn't catched
which might have caused the sending side to hang.
Solution:
The Stream VFD's file access property list was extended by two
elements:
- maxhunt: how many successive ports to try if the one
given in the filename is already in use
Default is not to hunt for additional ports.
- port: port number which is finally used to bind a socket
This might be different to the port number
as given in the filename if port hunting is enabled.
The H5Pget_fapl_stream() can be used to obtain
this port number.
Platforms tested:
x86 Linux, Irix 32/64 bit, Dec Alpha, Unicos on T3E, AIX on SP2
Hitachi SR8000
Code cleanup
Description:
Recent CodeWarrior patches have broken the Unix builds and moved code
around in non-portable ways.
Solution:
Patched things back up to try to accomodate CodeWarrior and still let the
Unix builds work correctly.
Platforms tested:
FreeBSD 4.3 (hawkwind)
code warrior support
and some clean up
the macros file_seek and file_offset_t that were repeated over sevral files were put only in
H5private.h
H5private .h was updated for win32
vthe
Description:
Solution:
Platforms tested:
Purpose:
bug fix for windows MVS
Description:
adding various windows macros to make it work on windows platform
Solution:
see above
Platforms tested:
test on windows 2000, confirmed on linux.
Code clean/bug fix
Description:
H5FL (free-list manager) code currently is taking an hsize_t as the size
of a memory block to allocate. On many machines, the size of an hsize_t
is greater than the size of a size_t, potentially leading to incorrect
memory allocations in rare circumstances.
Solution:
Changed hsize_t parameters and variables to size_t.
Platforms tested:
FreeBSD 4.3 (hawkwind)
