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
Bug fix.
Description:
The "h5dump -o ..." test generates temporay files in the testfiles of the
source code and later on remove them. This could cuase a racing condition
if more than one --srcdir build is using the same copy of the source code.
Since they use the same file name in the testfiles, they may conflict with
each other.
Solution:
Changed to generate the temporary files in the build-dir's own testfiles
directory. Since the build-dir can have different names, the CMP of expected
output now skip the first three lines which are label lines that contains
the location of the temporary.
Also removed the CREATE code since actual files created now cannot be
blindly copied to the expect files. Also, expected files should be
create by explicit action and careful inspection of files generated.
Tested platform:
Done in kagiso, both by --src-dir and in-place build.
On Windows, Mingw interprets all parameters starting with '/' as paths, and replaces the '/' with its home directory, "C:\Windows\msys\". This was a problem in h5diff tests such as:
h5diff h5diff_101.txt $FILE1 $FILE1 /g1/d1 g1/d2 -v
I've removed the leading '/', as h5diff will interpret it the same either way.
Tested:
kagiso, linew, and smirom, via h5committest
mingw on Windows XP
Changed macro names, conflict with new compiler
-- changed DATASET to H5_TOOLS_DATASET
-- changed DATATYPE to H5_TOOLS _DATATYPE
-- changed GROUP to H5_TOOLS_GROUP
In mingw's version of winsock2.h they define:
typedef unsigned int GROUP;
remove STORAGELAYOUT macro, it was not used
tested: linux
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: 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).
Minor tunings to output verbose messages:
1)when there is not a filter request do not print a message saying the filter was not apllied when the dataset was too small
2) avoid printing the message that has a list of objects to modify when there is none
Tested:linux
Modified the current h5dump test script to use h5import/h5diff calls to validate the binary output. At this moment it can only be used with the native test, since h5import does not deal with input endianess.
tested: linux, sunos 5.10
change _int64 to __int64 within windows macro.
Windows can tolerant _int64 but mingw cannot. So we need to use __int64 to make HDF work on Mingw.
Compiled at linux and Mingw.
h5repack and h5diff hyperslab I/O
changed the limit on which hyperslab I/O is done from 1GB to 128MB
h5repack currently tests this feature by defining a dataset with dimensions of 128MB + 1byte (the datum being 1 byte integer), in which a 1Kb hyperslab was written.
tested: linux
there were some errors in running the tests, but these were in the library, due to a recent check-in that will be fixed shortly (at least we hope :-) )
Take further advantage of the readers/writer locking in the metadata cache
and push the locking of the local heap up to the routines which call B-tree
routines, which should drastically lower the number of metadata cache protect/
unprotect calls.
Tested on:
Mac OS X/32 10.4.9 (amazon)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
Rename new error handling API routines from H5E<foo>_stack() to
H5E<foo>2().
Tested on:
Mac OS X/32 10.4.9 (amazon)
FreeBSD/32 6.2 (duty)
FreeBSD/64 6.2 (liberty)
bug fix.
Description:
If a non-existing or empty source file is used with -c option, the
script ended up doing nothing and exit 0--a false success. Changed
code so that it would do compile and end in error.
Tested:
Kagiso.
Change H5[D|G|T]<foo>_expand() "temporary" API routines to
H5[D|G|T]<foo>2() "versioned" routines. Also added
H5[D|G|T](create|commit)_anon() routines to continue to allow "anonymous"
objects to be created in a file.
Tested on:
Mac OS X/32 10.4.9 (amazon)
FreeBSD/32 6.2 (duty)
FreeBSD/64 6.2 (liberty)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
Add version # and flags to external link format (as fields in a single
byte), in order to accomodate future changes/expansions.
Tested on:
Mac OS X/32 10.4.9 (amazon)
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
FreeBSD/32 6.2 (duty)
FreeBSD/64 6.2 (liberty)
Bug fixes
Reset external file list slots name_offset to a state when created (0) in H5P_dcrt_copy
so that it conforms to an assertion in H5D_update_entry_info that assumes the name_offset is 0 at this point
this fixes the problem of h5repack and external files, add a new test and files for an external file
h5diff, check for an error return in H5D_get_storage_size
tested linux 32, 64
Check in changes from Elena and I to get pgcc compiler working again.
Primarily (all?) changes to move from using 'hsize_t' as array index to using
something else ('size_t') mostly.
Tested on:
Linux/32 2.4 kagiso w/pgcc
Eliminate storing # of links in "link info" message, regenerate it
when the object is opened instead.
Tested on:
FreeBSD/32 6.2 (duty)
Mac OS X/32 10.4.8 (amazon)
Move ref. count of # of links to an object out of the object header's
prefix and make it a header message instead (since it's a "rare" occurence),
eliminating some more space for each object in the file.
Inserting this "ref. count" message exposed a flaw in the library's
mechanism for locating a message to promote to another chunk and replace
with a continuation message, which required some additional work to fix.
It's still not completely robust, but it's working for more cases now and
detects failures robustly.
Reduced the minimum size of an object header chunk to just enough to
contain a header message prefix and continuation message.
Tested on:
FreeBSD/32 6.2 (duty)
Move "creation order tracked" flag from "group info" to "link info"
object header message and make the "max. creation order value" optional in the
"link info", if the creation order for links is not tracked.
Also, get rid of unused "index names" flag - names are always indexed
currently.
Tested on:
FreeBSD/32 6.2 (duty)
Eliminate message count from new version of object header prefix -
it can be computed when the header is loaded and the table of messages is
built.
Tested on:
FreeBSD/32 6.2 (duty)
Move attribute tracking information out of object header prefix and
make it into a message that is inserted only when attributes are present on
the object.
Tested on:
FreeBSD/32 6.2 (duty)
Bug fix
Some data was declared as char and saved with H5T_NATIVE_CHAR type. But the data have negative values
Solution : declared data as signed char and used H5T_NATIVE_SCHAR
The main purpose of this checkin was to eliminate the
space used for tracking creation time indices when there is no way they
can be used (i.e. attributes can't be shared in the file and the user hasn't
turned on attribute creation tracking), however there were some other minor
changes which crept in:
- Fix a cache locking deadlock when a shared attribute and one of its
components end up in the same fractal heap direct block.
(This is fixed the "slow" way for right now, until John has time
to add support for readers/writer locking to the cache.
- Optimize attribute copying when a copy will be kept during a v2 B-tree
search.
- When freeing a block on disk, attempt to merge it with the metadata
and "small data" aggregators.
Tested on:
Mac OS X/32 10.4.8 (amazon)
FreeBSD/32 6.2 (duty)
