- HDFFV-5928 - GMQS: h5diff problem and improvement on comparsing the same objects
Description:
Fixed:
1) adding h5tools_is_obj_same() function to check if two given IDs or paths point to the same object. This function can be very useful for other tools and applications.
2) using h5tools_is_obj_same() at h5diff() and diff() in h5diff.c. If two paths point to the same object, there is no need to check the details of the object since we know there is no difference. The fix will increase the performance by skipping the content comparison. It also fixed the problem of reporting difference for some cases of comparing the same file, e.g. empty files or files with incomparable objects the same file.
Test update:
Updat prvious test cases (171, 172, 530) affected by this fix, so they
still perfrom originally intended testing without bypassing.
Tested:
jam (linux32-LE), koala (linux64-LE), heiwa (linuxppc64-BE), tejeda (mac32-LE), linew (solaris-BE), cmake
Fixed Bug 2184 - GMQS: h5diff - incorrect calculation code for
--use-system-epsilon option
Description:
Merged from HDF5 1.8 branch r20369.
Fixed h5diff for --use-system-epsilon option: the calculation changed
from ( |a - b| / b ) to ( |a - b| ). This was decided for better
performance and was corrected only in HDF5 trunk, so 1.8 got updated.
Also comments for equal_XXX() function were updated correctly.
Also help page and RM got updated correctly.
Also add test cases for testing the differences w/wo the option.
Tested:
jam (linux32-LE), amani (linux64-LE), heiwa (linuxppc64-BE), tejeda (mac32-LE), linew (solaris-BE), Windows (32-LE)
Fixed CHICAGO: Bug 2121 - h5diff - incorrect and lack of output for the
different set of attributes (different number and names)
Description:
Previously h5diff compared attributes correctly only when two objects have
the same number of attributes and attribute names are identical.
This fix covers all other cases.
Also didn't display useful information about attribute difference.
This fixes both issues.
Tested:
jam (linux32-LE), amani (linux64-LE), heiwa (linuxppc64-BE), tejeda (mac32-LE), linew (solaris-BE)
Bug 2089 - GMQS: h5diff segfault on a compound dataset with fixed length +
vlen string type order
Description:
This is fix for the segfault when h5diff compares a compound dataset with
combination of fixed length string types and vlen string types in certain
orders. Optimized vlariable length string handling codes. The fix is
referred from h5dump handling vlen strings.
For testing, several compound datasets were added with various combinations.
Previous failed cases:
- Vlen string, Fixed length string, Vlen string, Fixed length string
- Fixed length string, Fixed length string, Vlen string, Vlen string
- Fixed length string, Vlen string, Fixed length string, Vlen string
Tested:
jam (linux32-LE), amani (linux64-LE), heiwa (linuxppc64-BE), tejeda (mac32-LE), linew (solaris-BE)
Fix for group comparison with exclude-path case. Use relative path.
Description:
Related to "1890: h5diff excluding object for file comparison via command line" checkins. (r19406)
Tested:
jam, amani and heiwa
Add extra test cases for bug1975 h5diff - support recursive comparison on group when specified as an object.
Description:
Additional tests for combination of group recursive and --follow-symlinks
with multi-linked external links with several files (same name/strucure).
Tested:
jam, amani, heiwa
Add --exclude-path option
Description:
Specified path to an object will be excluded from comparing the two files or two groups. If group is specified all the member objects will be excluded.
Related to "1890: h5diff excluding object for file comparison via command line"
Tested:
jam, amani and heiwa
Fix for Bug1975 h5diff - support recursive comparison on group when specified as an object
Description:
Compare member objects and groups recursively when two files or groups are specified to be compared. Support parallel diff and handling symbolic links accordingly.
Tested:
jam, amani, heiwa
Problem: some launching command like yod (and sometimes mpirun), do not pass
back the exit code of the executables back to the shell. It always returns 0
if it can launch the executable. This cause tests that expect non-zero exit
code to fail incorrectly.
Short term Solution:
Added the feature that if $RUNSERIAL_NOEXITCODE is defined, do not
test conditions that rely on non-zero exit code.
Tests:
dove-linux using RUNSERIAL=thgyod to see the failures and using
$RUNSERIAL_NOEXITCODE=yes to bypass the errors.
Also h5committested.
Rename '--follow-links' to '--follow-symlinks'
Description:
The '--follow-links' option is to follow symbolic links (soft and external).
Make the name more intuitive and specific to the feature.
Tested:
jam, amani and linew
bugzilla 1754: h5diff: support comparing through links.
Description:
Fix incorrect (or hanging) behavior in parallel mode when specifying
invalid combination of command options.
(ex: -v and -q , --no-dangling-links without --follow-links)
Add relate test case
Update h5diffgentest.c due to add test files this and previous time
Note:
svn #18266 (prior to this)
svn #18164 (original check-in)
Tested:
h5committest (jam, amani and linew)
bugzilla 1754: h5diff: support comparing through links.
(original check-in svn revision #18164)
Description:
fix the hang issue in parallel mode when compare external-link.
add --no-dangling-links option.
add test cases (#450-#459) relate to the new option.
improve test script to check exit code.
update --help relate to the new options.
correct some indentations.
Tested:
h5committest (jam, amani and linew)
They are external-link test cases with following link option.
The related code was checked in as svn #18164 and #18165 to support follwing
links for h5diff.
The test cases will be added back when find solution for parallel mode.
Tested on Jam
Add test cases for h5diff check-in (svn revision #18164).
Description:
This test cases are for testing comparing through various link combinations
relate to bugzilla report 1754.
Also improved the test script to handle external link cases and
simplified to use single line instead of double for each test cases.
Tested on Jam
The ph5diff tests printout incorrectly reporting they are running
h5diff when it actually is running ph5diff.
Fixed.
Tested: Jam (parallel), linew (serial)
No H5committest since it is a shell script and Jam tests both
serial and parallel h5diff for Linux while Linew tests Big
endian platform.
" Use "--use-system-epsilon" for system EPSILON
" Use "-p" or "-d" for whatever user's choice of epsilon
" Use "-p 0" or "-d 0" for strict equality (same as default)
Solution: for compound types, recursively apply that check
Two new cases are added
1) the compound type has a different number of members. Message printed is
<obj1> has X members <obj2> has Y members
Where X and Y are the number of members of each compound type being compared
2) the compound type has not comparable types (for example a double and an int at the same index)
In this case the message
Comparison not possible: object1 is of class1 and object2 is of class2
Is replaced with
Comparison not possible: object1 has a class1 and object2 has a class2
Modified the test generator program to have these 2 cases
Added a shell run for these 2 cases
Tested: windows, h5committest
Summary: when using h5diff to compare the results of h5repack (or other tools that copy one HDF5 file to another), a new option is needed to allow h5diff to make an "absolute" comparison of the 2 files. This is the "contents" mode explained in the usage below.
If this mode is present, objects in both files must match (must be exactly the same). If this does not happen, the tool returns an error code of 1 (instead of the success code of 0)
Changes to the h5repack test script: the call to h5diff was changed to include -c (maintaining the previous -q).
tested: windows, linux, solaris
The name of the files are now given by its full name relative to $srcdir
To avoid the printing of the complete full path of the test file, that hides
all the other parameters for long paths, the printing of the command line
is done first in TESTING with the name only of the test file, not its full path
the printing in the expected output that had the file name was removed as well as 3 tests that tested error messages in which the file name was present
tested: linux (in 2 different build directories relative to $srcdir), solaris
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
Bug fix: the macro used for percentage in the unsigned types needed a cast to signed (the difference can be negative)
Unlike the 1.6 branch the unsigned long long conversion to float is supported by H5Tconvert , so the test commented for 1.6 is run (tools/testfiles/h5diff_16_2.txt)
Added a relative error formula to deal with floating point uncertainty
in the comparison of floats and double types.
Added new tests for this feature to the file generator program and to
the shell script
h5diff: print a message of "not comparable" in a case where the relative error
compare is not possible, due to the denominator being zero. Modified
the test file generator program to include a example for this and a new
test on the shell script
1) added a new parameter to the h5diff function diff_array that contains
the beginning position of the hyperslab, so that the total position in
the array is printed correctly when reading by hyperslabs.
2) added a new test to h5diff that reads and diffs by hyperslabs. The
test reads a 1GB dataset, from which a 1KB hyperslab was written with
differences .
3) added the generation of 2 files to the generator program to test the
h5diff hyperslab read.
4) changed the h5diff binary pre-generated file names to be more
descriptive (e.g, instead of file1.h5, made it h5diff_basic1.h5)
5) changed the name of the h5repack options text file to info.h5repack
new feature
Description:
added support for the printout of dataset region references differences
added a new test for this
merged the h5diff generator of test files into a single file
Solution:
Platforms tested:
linux 32, 64
solaris
Misc. update:
bug fix
Description:
percent relative error was done using integer arythmetic; use floating point instead
added the case for unsigned long long integer to float conversion
Solution:
Platforms tested:
linux (32,64)
AIX
solaris
Misc. update:
bug fix
Description:
1) added a more explainative usage message
2) the percent relative error for the integer type (division) was being done using integer arythmetic; use floating point arythmetic instead
3) added a new test for integer percent
Solution:
Platforms tested:
linux (32,64)
AIX
solaris
Misc. update:
bug fixes
Description:
1) the option logic for the print of long types was done incorrectly (extra lines of code that were not supposed to be there)
2) the print of loong long types was incorrect
Solution:
1) removed the incorrect code
2) made a long long cast on the printf call
Platforms tested:
linux (32 and 64)
solaris
Misc. update:
