Modify semantics for constant messages slightly, to allow them to be
deleted, but not modified. This works around the issue where "old" style
groups are being created with the symbol table message marked constant in
the 1.6 (and prior) branch, but the 1.8 branch needs to remove the symbol table
message when the group has been changed to the "new" style (upon insertion
of a link that the "old" style can't handle).
Tested on:
Mac OS X/32 10.4.10 (amazon)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
Fix problem with VL-datatype fill values when shrinking dataset's dimensions
Tested on:
Mac OS X/32 10.4.10 (amazon)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
Allow H5Dset_extent() to reduce the size of dimensions that aren't unlimited
Tested on:
Mac OS X/32 10.4.10 (amazon)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
Another minor code cleanup to chunked storage "init by extent" routine.
Tested on:
Mac OS X/32 10.4.10 (amazon)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
Finish cleaning up "initialize by extent" routine (used when shrinking a
dataset's dimensions) before addressing VL-datatype fill value bug.
Tested on:
Mac OS X/32 10.4.10 (amazon)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
2) Removing configuration of c++/src/H5cxx_pubconf.h file.
Reasoning: the additional pubconf file caused compilation complications, this is cleaner.
Tested: kagiso, smirom
Clean up the chunked storage "initialize by extent" routine, in preparation
for fixing another VL-datatype fill value bug.
Tested on:
Mac OS X/32 10.4.10 (amazon)
Add small interface to "wrap" a static buffer (usually on the stack), but
still allow for buffers larger than the static buffer to be allocated. This
can eliminate _many_ short-lived buffer allocations in situations where the
buffer is a predictable size (or at least a "very likely" size).
Also, some minor code cleanups, particularly in the SOHM caching code.
Tested on:
Mac OS X/32 10.4.10 (amazon)
Description: added new configuration to generate a pubconf file
in the c++/src directory, H5cxx_pubconf.h. When C++
compiler recognizes 'offsetof', the macro
H5_CXX_HAVE_OFFSETOF is defined in the new
pubconf file.
tested: kagiso, smirom
Refactor fill value buffer code into one location, for better long-term
maintenance.
Tested on:
Mac OS X/32 10.4.10 (amazon)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
Description:
Windows does not accept the buffer size parameter of setvbuf to be less than 2. Previously, we had our fix split between two files, H5private.h and the Windows H5pubconf.h. This was a problem for MinGW, because it generates its own H5private.h. This moves the fix out of H5pubconf.h and into H5private.h.
This shouldn't affect other platforms.
Tested:
MinGW on WinXP
Clean up SOHM code, to copy message in H5HF_op call instead of separate
calls to H5HF_get_obj_len() and H5HF_read().
Tested on:
Mac OS X/32 10.4.9 (amazon)
Linux/32 2.6 (chicago)
company. The I/O is optimized when the source and destination
members are a subset of each other one way or another, and
the order is the same, and no conversion is needed. For example:
struct source { struct destination {
TYPE1 A; --> TYPE1 A;
TYPE2 B; --> TYPE2 B;
TYPE3 C; --> TYPE3 C;
}; TYPE4 D;
TYPE5 E;
};
or
struct destination { struct source {
TYPE1 A; --> TYPE1 A;
TYPE2 B; --> TYPE2 B;
TYPE3 C; --> TYPE3 C;
}; TYPE4 D;
TYPE5 E;
};
The optimization is simply moving data from the source to the
appropriate places in the buffer and bypass the reading of
the background data and data conversion.
Tested on smirom, liberty, sol, and copper.
Description:
Changed a macro in H5FDwindows.c from IO_BUF_SIZE to a more specific WINDOWS_MAX_BUF, and moved it to our H5pubconf.h, where users can customize it to their own preference.
This value is needed because of a bug in the Windows _write function, it is unsafe to write buffers larger than 2GB-1 bytes in one write.
Tested:
VS2005 on WinXP.
Fix writing variable-length datatype fill values for contiguous dataset
storage
Tested on:
Mac OS X/32 10.4.9 (amazon)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
Description: When the compiler has the __offsetof__ extension, the
H5_HAVE_OFFSETOF macro will be defined. When the compiler
does not have the extension, the macro will remain undefined.
Tested: kagiso
der /src. So when gmake check-install is run for HDF5 examples, it will fail. Th
e fix is to include a macro #ifdef H5_HAVE_WINDOW around #include H5FDwindows.h
so that H5FDwindows.h will only be included on windows, not on other platforms.
We've fixed with the same way for other source code that includes H5FDwindows.h.
This has been tested with h5committest. However, it takes a few hours to test all tests and except kagiso. Tests on linew and smirom cannot finish. It takes forever for the fheap test.
And for parallel test at kagiso, ph5diff test seems to be hanged. I think all of our changes will only affect windows. So in order not to block other developers to check in their code, we just go ahead to check our changes.
Kent Yang
Hook "default VFD" configure macros into library.
Also, make "default VFD" default to sec2 instead of nothing.
Tested on:
FreeBSD/32 6.2 (duty)
FreeBSD/64 6.2 (liberty)
Mac OS X/32 10.4.9 (amazon)
Description: this flag will define the macro H5_DEFAULT_VFD in src/pubconf.h
and set it to be either H5FD_SEC2 or H5FD_STDIO, depending on the
value set with the configuration flag (either sec2 or stdio).
Tested: kagiso, smirom, linew
Fix compact dataset storage to initialize VL datatype fill values
correctly.
Also, fix bug in global heap code when the fix action on a global
heap is to delete an object in a heap with no free space.
Tested on:
Mac OS X/32 10.4.9 (amazon)
FreeBSD/32 6.2 (duty)
Add a little more information to the debugging routine, for VL datatypes.
Minor code cleanups
Tested on:
Mac OS X/32 10.4.9 (duty)
Too minor to require h5committest
Description: a new configure flag has been created that will now set a macro
in the test/Makefile. The flag is with-default-vfd=. When set to
a virtual file driver (e.g. "--with-default-vfd=sec2"), the macro
DEFAULT_VFD will hold this value.
Tested: kagiso, smirom, linew
file size from MPI_File_get_size. Bypass this problem by replacing it with
stat. Add an option --disable-mpi-size in configure to indicate this function
doesn't work properly. Add a test in testpar/t_mpi.c, too. If it returns wrong
file size, print out a warning.
Tested on kagiso (parallel) because already tested the same change to v1.6 on
several platforms (kagiso, cobalt, copper, and sol).
Close another memory leak for VL fill values, and clean up some
temporary data structures on error.
Tested on:
FreeBSD/32 6.2 (duty)
Mac OS X/32 10.4.9 (amazon)
Incremental step forward in fixing problems with fill values that have
a variable-length daattype. This set of changes fixes problems with the copying
the property list values.
tested Tn:
Mac OS X/32 10.4.9 (amazon)
Fixed a bug at copying shared reg_ref attribute. The ref values of an attribute
was not writen in file. The post_copy_file() at H5Oattr.c only updates the value
in memory.
To fix the problem, we added H5O_shared_post_copy_file(), which delete the attribute
in file (with the incorrect value) and create an new attribute with the correct values.
I added the following functions:
H5Oshared.c:H5O_shared_post_copy_file(H5F_t *f, hid_t dxpl_id, H5O_t *oh, void *mesg);
H5Oshared.h: H5O_SHARED_POST_COPY_FILE
and modified the following files to include shared_post_copy_file()
H5Oattr.c
H5Odtype.c
H5Ofill.c
H5Opline.c
H5Osdspace.c
Change back to clearing memory buffers by default, but give users option
to disable this, for performance reasons (--disable-clear-file-buffers).
Also, add configure option (--enable-using-memchecker) to conveniently
control making the library "play nice" with memory checking tools like valgrind
and purify.
Tested on:
Mac OS X/32 10.4.9 (amazon)
FreeBSD/32 6.2 (duty)
Linux/32 2.6 (kagiso)
Fix possible file corruption when using "new" format object headers and the
size of chunk #0 for an object header transitions between needing 1->2->4->8-
byte encoding for the size and there are "clean" messages in the object header
already. (Usually triggered by flushing the file while adding attributes to
an object)
Tested on:
Mac OS X/32 10.4.9 (amazon)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
