Figured out a way to phase in individual message class support for
the new shared message "interface", so check that in before working on the
message classes themselves.
Tested on:
Mac OS X/32 10.4.8 (amazon)
FreeBSD/32 6.2 (duty)
Comment out adding "old" fill values to "new" fill values until we can
figure out why the tests fail on 64-bit machines.
Tested on:
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
Checkpoint more progress toward implementing shared message "interface"
methods for message classes that can have messages shared.
Tested on:
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
Add "old" fill value messages to "new" style fill value messages
when "FILL" flag is enabled for shared messages.
Minor code cleanups
Tested on:
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
bug fix.
Problem:
1. The last fix using "svn --version" did not work since the source code
is shared between different machines which have different svn versions.
Changed the code to inspect the .svn/entries content to guess what version
of svn uses. From that to decide which GETSVNENTRIES to use.
2. GETSVNENTRIES_14 had two errors:
2.a., % as 1,$ is actually not recognized by ed. Changed that to "1,$s/..."
but that got mixed up with $ being a shell meta-character. Changed to use
g command instead. (Could have use \ to escape $ but g is cleaner.)
2.b., Some directories have no files but sub-directories. The ed script
might end up with an empty file in some steps and ed does not like to
run g command with an empty file. Fixed it by adding a dummy blank line.
Platforms tested:
Osage and copper.
Get rid of two different types of fill value struct (merge H5O_fill_t
and H5O_fill_new_t) and clean up & simplify dataset initialization code.
(In preparation for shared object header message method call
refactoring).
Tested on:
FreeBSD/32 6.2 (duty)
Mac OS X/32 10.4.8 (amazon)
would become shared NULL messages (oops!).
Added a test for this case.
Also fixed the #define that caused errors on Windows (and probably other
platforms). This may fix the Daily Test error.
Tested on Windows, smirom, kagiso, and juniper.
Upgrade.
Description:
svn version 1.4.x uses a different format of .svn/entries files as older
versions like 1.3.x. chkmanifest could not parse the file correctly.
Solution:
Created different versions of entries file parsing according to svn version.
There is also a bug in grep that matches ./configure with ./fortran/configure.
Don't know how to or even possible to nullify the wildcard character effect
of '.'. Added a ^ to mark beginning of line. At least it makes it the same
length though it would still match with "X/configure". Hopefully this sort
of error are easier to detect by human
Add "set_crt_index" and "get_crt_index" methods for the object header
message class.
Unify fractal heap definitions for shared messages and attributes, under
"object header" fractal heap definitions.
Initial code for adding creation order index to object header messages.
Tested on:
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
superblock).
Tried to standardize sizes and added #defines to H5Fpkg.h so that hopefully
the code is a little easier to read.
Of course this is a file format change.
Tested on Windows, juniper, smirom, kagiso.
Add anotherdense/shared attribute test.
Clean up minor attribute/object header codelets, in preparation for
adding creation index to attributes.
Tested on:
Mac OS X/32 10.4.8 (amazon)
FreeBSD/32 6.2 (duty)
Added can_share callback for OH messages. This determines whether the
message is allowed to be shared in the heap (committed and immutable datatypes
can't be).
Fixed a bug in the dense attribute storage that tried to open the shared message
heap when it hadn't been created yet.
Made the test to extend shared dataspace messages a bit more robust.
Refactored the code the searches a shared message list index to be a little
more efficient.
A few other minor changes.
Tested on smirom, kagiso, Windows, and juniper.
Tested 1.6 compatibility feature: work for vs studio 6.0
Corrected several small bugs inside tool batch files
Add a few project files to test error APIs
Auto-detect the support for the posix_memalign routine and disable
the direct I/O VFD if posix_memalign is not supported. This should allow
direct I/O to be auto-detected and disabled on FreeBSD. (Direct I/O itself
works fine, but doesn't have the alignment requirements on FreeBSD that it
has on Linux, so the code would have to be re-worked to not use posix_memalign
and we don't need to support that right now).
Tested on:
FreeBSD/32 6.1 (duty)
Linux/32 2.4 (kagiso)
Moved SOHM table version out of table encoding and completely into superblock.
This is a file format change.
Added test that extends shared dataspaces. Dynamically allocate arrays in
shared message cache code. Clean up comments.
Tested on windows, kagiso, smirom.
New version of the function h5tools_dump_simple_subset, to display subsetting. The new algorithm is:
Introduced an outer loop for cases where dimensionality is greater
than 2D. In each iteration a 2D block is displayed by rows in a inner
loop. The remainning slower dimensions above the first 2 are incremented
one at a time in the outer loop
Note: when blocks are introduced, the display is not correct. This is a bug that requires an improvement of the algorithm.
modified the format for the printing of reference information for a string, related to the same change in h5dump to display the name of the referenced dataset
Fixed#720 h5dump: improve how region references are displayed. h5dump now uses the new API function H5Rget_name to display the name of the dataset referenced instead of its ID. Added a case to the script test file
Further tests & bugfixes for shared/committed components of shared/un-
shared attributes using dense/compact storage.
Tested on:
Mac OS X/32 10.4.8 (amazon)
FreeBSD/32 6.1 (duty)
More progress on address bugs in combinations of shared/committed/unshared
components of shared/unshared attributes in compact/dense storage. (Yes, there
are a lot of combinations. :-)
Tested on:
Linux/32 2.6 (chicago)
Linux/64 2.6 (chicago2)
