[Darshan-commits] [Darshan] annotated tag, darshan-2.3.1-pre1, created. darshan-2.3.1-pre1
Service Account
git at mcs.anl.gov
Wed Oct 29 13:30:15 CDT 2014
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "".
The annotated tag, darshan-2.3.1-pre1 has been created
at 335af998e28f4ae475da2019b144e52e958743e5 (tag)
tagging 2a24693eaa03d3c02b547aa77e07f58e34c6ce8e (commit)
tagged by Phil Carns
on Wed Oct 29 14:29:30 2014 -0400
- Log -----------------------------------------------------------------
version 2.3.1-pre1
Phil Carns (10):
update release checklist to reflect git repo
add wrappers for mkstemp() functions
bump log format for open timestamp change
simplify zlib m4 check
update Changelog for zlib autoconf fix
attribution in Changelog
update pkgconfig for craype 2.x to group libraries
fix ytic labels for time interval graphs
clarify and improve darshan-parser --perf
update doc to match new darshan-parser output
Rob Latham (1):
fixup: assignment, not equality
carns (612):
starting to reorganize repo
reorganizing into trunk,tag, branch directories
document and install darshan-diff
update all of the fortran compilers to list the darshan libraries twice
added comments to darshan-parser explaining what is included in meta_time
added configure script option to use a specific libz build for mpi
bug fixes and install documentation updates for --with-zlib-for-mpi
Add copyright information patterned after mpich2; also add comment to all
correct previous commit (don't put comments before #! in scripts)
some tweaks to how the first graph is displayed
rought cut at converting gnuplot stuff to use eps then convert to pdf
comment out unused graphs for now
working on reorganizing perl utilities so that they can be installed
renaming tool
renaming tool again
fancy argument parsing, ability to specify an output file, etc.
switch to subfigures for page placement
adding copy of gnuplot tar ball to repository for convenience
adding script to install darshan-job-summary.pl on gadzooks
adding scripts and notes for eureka install
fix bug in copying files
install script for surveyor
fix accidental commit of development hacks
added install script for intrepid
Added ability to disable darshan at link time if common pmpi tools are
updated all compilers to disable darshan when PMPI tools are used; needs
added missing bin path variable from previous commit
fix some link problems in cxx compilers
minor fix; lcxxmpich was accidentally added in a previous commit
early draft of mpicc script generator
incremental progress, need to make this a configure-generated file to
generate darshan-gen-cc.pl at configure time
complete version of darshan-gen-cc.pl, needs testing
convert to new directory/permission setup
reverting previous commit after some problems testing on surveyor
make output files only readable by user by default
added changelog for 2nd release
added upgrade scripts for surveyor and gadzooks
updated install doc with pointer to web site
added _r versions of mpi compilers for bgp
started reworking compiler scripts for V1R4, not done yet
added autoconf variables to V1R4 compilers
filled in all fortran compilers, needs testing
finished filling in darshan modifications to all compilers; needs
fix bug in updated cxx compilers
updated changelog
updated install files for anl machines to use correct path
extra ld flags to make sure that a few symbols always get pulled in even
comment out some infrequently used benchmark routines
revert previous changes to ld flags
updated to inspect symbol names when linking and decide whether to
updated remaining c compilers to inspect symbols
propigate symbol check logic to cxx compilers; untested
propitated symbol detection logic to fortran compiler scripts
updated ibm mpi compiler scripts to use "which" to locate underlying
removed old V1R3 compiler scripts, too much trouble to update at this
adapted test compilers to match the others
removed depricated filter script
updated symbol check logic to work in -shared case as well
soften error message on failure to open log file
removed unused variable from compiler scripts
updated gen-cc.pl script, needs testing
bash instead of sh
fix variable substitution
preliminary support for tracking pnetcdf opens, not finished yet
completed tracking of pnetcdf opens and closes
added basic tracking of hdf5 opens
added tracking for time and size of slowest posix operations
updated changelog
updated installation scripts
updated PMPI usage in the library to only affect MPI-IO, Init, and
update all compilers to match new PMPI detection logic
update changelog and install scripts for next release
patch from Kevin to make darshan work correctly for apps that use
fixed more generic bug of darshan_finalize() causing a segfault if
updated changelog and ALCF install scripts for 1.1.6 release
fix compile warning
updating changelog and install scripts for 1.1.7 release
trivial benchmark for realpath() overhead
add a little error checking to darshan-analyzer and add it to the default
port darshan-parser to use logutils library, also some changes to semantics
rename some of the logutils functions
added simple utility to print log format parameters
noted differences in 1.21 and 1.22 log format versions, added utility
added logutils functionality to translate 1.21 format log files
fixed bug in darshan name table
updated analyzer utility to use logutils api
move print macros and use name arrays instead of enum name
get rid of explicit naming of every counter
logic to scan mount points at finalize time; still need to store the
storing mount point information in trailing data for job
add gzseek calls where appropriate to logutils, so that it is a little more
minor bug fix to eof detection after log_getexe()
make getexe() ignore any mount information in the trailing job data
bug fix: print version of log, not necessarily current version any more
added collection of device id for each open file, bumped log version,
updated parser to map files to mount points, other misc bug fixes
fixed typo bug
another bug fix to gzseek usage
update reduction and condense functions to account for device record
remove largefile cflags from utility builds, working aroung a libz bug on
fixed buggy access size counting
updated comments in darshan parser output to explain new fields
updating installation scripts for next release
updated changelog for 1.1.8 release
- added CP_SIZE_AT_OPEN metric (records size of file when opened)
changed title of access size chart
updated changelog for 1.1.8
patch from Rob Ross to get some of the tools to build on Darwin
added exclusive flag when opening output file to make sure there is no
stub for utility to merge fsstats csv files
merge first few summary lines
pulled in histogram module from fsstats
initialize histograms, and stub out routine to find and parse specific
incremental progress on aggregating histograms, needs more work
added remaining histograms, but it doesn't quite work yet
bug fix to histogram search
completed first cut at fsstats-merge.pl; now accumulates absolute and
added some utilities for running fsstats in parallel using ssh
added "fast" mpi compilers, also fixed minor problems in new off64_t
update Changelog for 1.1.9
updated install/upgrade scripts for 1.1.9
fix bug reported by RobL when configuring pnetcdf: add linker flags to
updated Changelog to reflect latest bug fix
fix bug in CFLAGS that accidentally triggered a zlib bug again in
Added sanity test in darshan parser to throw an error if the log file
make sure that stat() and its variants correctly initialize the file record
update changelog for 1.1.10 release
updated install/upgrade scripts for 1.1.10
adjust gen-cc script to work on other systems besides bg/p
more updates to gen-cc to match new PMPI tests in stock compiler wrappers
added gen-cxx utility
update comments and usage information for gen-cxx
bug fix to cxx wrappers
fix pointer arithmetic for x86_64
bug fix in cxx gen
tentative fix for lseek bug in applications that are compiled without
added seek test program
bug fix in lseek64 wrapper
updated changelog
updated install scripts for .11 release
experimental fix for mismatched mount ids across nodes
fix logic from previous commit
bug fix in logutils version handling
bumped version number to reflect improved mount point mapping
started on graphs to show file access durations
tweaking file access graphs and making a seperate one for shared files
accumulate some seperate shared and independent values
new graphs in summary showing I/O durations
incremental progress in summarizing, needs to be put into a table
new table, not formatted properly yet
tweaking table and layout on page 2
cleanup of file access table
bug fixes and wording fixes
more formatting tweaks
updating Changelog for .12 release
update install scripts for .12 release
updated compiler scripts to fix link order in cases like this:
updating changelog and install scripts for .13 release
fixed bug causing erroneous incompatibility warning from darshan-parser
updated changelog and installation scripts for 1.1.14 release
modified fsstats-merge.pl to be able to operate on files that have already
fixe regex to match version number in fsstats-merge.pl
removed debugging printf accidentally added in last commit
clean up warnings on 64 bit systems
fix here document syntax in compiler gen scripts
fix previous commit
more bug fixes to gen scripts
more bug fixes to gen scripts, now works on gadzooks
fix more build warnings
update configure so that --with-mpi argument overrides whatever mpicc is
fix bumbled printf warnings for 32 and 64 bit machines
experimental darshan-gen-fortran.pl
add gen-fortran to installation list
converting to fixed sized types with 64 bit alignment
add magic number, hack up the logutils to handle 2.00 file format
some infrastructure changes to prepare for byte swapping
minor update to prepare for new analysis
incremental progress on gathering unique file statistics
basic table with file count
added human-readable number parsing to job summary
filled in file count table and added metadata to final table
merge install script updates from 1.1 branch
merge mntid type patch from 1.1 branch to trunk
add byte swapping of remaining logutils functions for new darshan file
started stubbing in backwards compatibility hooks
add configure check for endian-ness
backwards compatibility with old job struct format
fix typo and a few warnings
backwards compatibility with 1.24 and 1.23 file formats
backwards compability for 1.22 and 1.21 file formats
minor adjustment to comments
modify darshan-job-summary.pl to use darshan-parser directly
tentative changelog for 2.0.0 release
adding example output files from previous versions of darshan
fix various bugs in backwards compatibility
bug fix to exe parsing for legacy log format
new example output with jobid changes
added example output from 32 bit x86 machine
add x86_64 example
backwards compatibility and endian-ness handling for new jobid field
configure help for systems that don't have a jobid env variable
Fix for bug reported by Noah Watkins: rename hash function to avoid
backwards compatibility for new min/max fields
warnings about missing fields in old versions
bug fix to backwards compatibility on 1.23 and 1.24 file formats
add proper Darshan 2.0.0 output file example
update 2.0.0 example output for ppc32
crude test script to exercise compiler wrapper combinations
final example output for darshan 2.0.0 on x86_64
updating draft changelog for 2.0.0 release
correct off by one bug in setting up buffers for compression
experimental patch in case we want to try using a single aggregator for
fix typo in divide-by-zero fix
don't forget to count fopens as posix opens too
fix typo in previous commit; darshan-job-summary.pl now correctly handles
trivial wording change
update install scripts for 2.0.0
remove example output file from 2.0.0 pre1 tag
#20 Add an environment variable that enables timing of shutdown routines at
updating changelog for 2.0.1
update installation scripts for 2.0.1 release
Update compile test script to handle xl compilers. Use subroutines for
bug fix to darshan-gen-fortran.pl; the -lfmpich substitution has to be
typo in test program
more test script bugs
propigate darshan-gen-fortran bug fix to darshan-gen-cxx for runtime
update stale INSTALL notes
incremental step in performance calculation: find the time of the slowest
2nd step to calculating performance: tabulate the time spent in shared files
calculate total I/O time and (sometimes broken) total bytes written
use gnuplot from system path if there is not one installed with darshan
updating Changelog for 2.0.2 release
bump version numbers in alcf maint scripts
notes about how to run compile tests
added 2.0.2 output example from ppc32
added x86 example output for 2.0.2
added 2.0.2 x86_64 example output
compile time error if darshan is configured with a bad memory alignment
set default mem alignment to 1; silently ignore if env variable for
add -lz as a dependency for libdarshan.so
cleanup collective operators (new mpich2 releases generate warnings about
be more careful about freeing datatypes
remove extra locking around darshan_initalize(). This causes deadlocks in
perl script to summarize a few statistics for each log file in a specified
partial fix for DESTDIR problem reported by Bill Barth. Use a more
2nd half of fix for DESTDIR problem reported by Bill Barth: honor the
configure test for MPI-IO support
break compilation with a more helpful error at least if MPI 1.x is detected
use recursive mutexes by default to account for system libraries that make
switch to compile-time test for how to print int64
replace non-portable MPIO_Request with MPI_Request in prototypes
get rid of c++ style comments
experimental MPI 1.x support
more MPI 1.x support (should have been in last commit)
configure tests for pdflatex and pdflatex -halt-on-error
put get_envelope() calls back in for MPI 1.x and define any missing combiner types
clean up MPI_Request configure test
experimental performance calculation fix: if files are partially shared,
fix bug in perf calc for both darshan-parser and darshan-job-summary.pl;
- overhaul int64 print handling (use PRId64 and friends with appropriate
fix previous commit
rebuild configure script to get datarootdir
quick hack scripts to gather file count data from log files
bug fix, missing newline in output
fix for bug during shutdown on openmpi
fix bugs in previous datatype workaround for openmpi
fix broken fallback definitions of PRId64 and PRIu64
added --disable-ld-preload option to configure
bug fix to previous commit
update install scripts for bg/p to disable ld preload
drop gnuplot and zlib from svn tree
only print log file open warning on rank 0
gather more helpful error information when we fail to open a log file
Fix bug reported by Rob Latham: the darshan-gen*.pl scripts should detect
new (optional) configure argument --with-logpath-by-env
commit bug fix to fortran wrapper script pointed out by Sandra Schröder
add mode argument to logutils open function (preparing to support writing
starting point for darshan-convert utility (intended to read from one darshan
autoconf tests for bz2 library
ability to detect, open, and close bz2 files
bzip2 read support in darshan-logutils
bzip2 write support, slightly smarter seeking, misc bug fixes
update api usage in darshan-diff
remove extraneous call to MPI_File_set_size()
crude patch to allow testing of multiple collective open/write
Added --with-log-hints option to autoconf, with default value of
environment variable override for --with-log-hints value
first cut at passing configure time / env variable hints into open of
bug fix: add 2.01 log format to function that prints warnings about
record mpi-io hints used on log file in the log file's header
new #define for size of metadata field in header; should have been included
warning fix: avoid using floating point computation for size of static array
added another sanity check for hint format
Changelog updates
- change delimiter for hints to ;
fix datatype leak
fix changelog, next release is 2.1.1 not 2.1.2
update maint scripts to v 2.1.1
revert to using no MPI-IO hints by default
fix warning
x86 example of 2.1.1 output
added x86_64 example to example output
ppc32 output example for 2.1.1 release
proper cleanup after failure to write log file:
updating changelog and maintenance scripts for 2.1.2 release
first cut at splitting darshan tree into separate runtime and util packages
fix install targets for darshan-util
move patches to a dedicated subdirectory
added brief readme file explaining directory structure
note about tests dir in README
clean out INSTALL file, stop looking for bz2 lib in runtime build
removed duplicate test from configure.in
improve configure error message
fix zlib handling in darshan-runtime configure script
fix ldflags bug
moved file
fix 32bit overflow bug on some systems for MAX_BYTE tracking
starting point for script to parse darshan log files and generate I/O
incremental progress
finished interval script
add start time to jobid in interval file name
fixed pwrite offset tracking
added pread fixes to match pwrite fixes from previous commit
make it possible to manually disable bz2 support
updated darshan-util to give better error messages if the user attempts to
move command pipe closure to earlier in darshan-job-summary script so that
first step in support -cc in darshan-gen-cc.pl; detect the CC portion of the
add -cc support to gen-cc output
support -fc, -f90, and -f77 in gen-fortran script
add -cxx support to gen-cxx script
experimental detection of gnuplot version
allow logpath-by-env without logpath option
produce an error at configure if obsolete args are used
started updating changelog for next release
added example output for 2.2.0 release, x86_64
added example output from surveyor
stub in asciidoc installation docs for darshan-runtime and darshan-util
whitespace change for testing
start filling in darshan-util documentation
start filling in tables in darshan-util documentation
generate html, minor intro tweaks
pull in remaining darshan-parser output tables from wiki
finish draft of darshan-util docs
start stubbing in the darshan-runtime documentation
fill in more generic docs for darshan-runtime
fill in recipes for different systems
changelog update and minor doc tweak
minor fix to subsection hierarchy
proof reading
proof reading
spell check
switch to docbook toolchain for html generation, add link back to web page
commit tweaked css for docbook (fixes table row height)
simplified author information in copyright
test program for cuserid function
add getenv(LOGNAME) fallback for finding user name
add --disable-cuserid configure argument, fixes #72
update Cray documentation
added hugetlbfs to list of filtered file systems
added darshan-config utility
updated darshan-gen-* scripts to use darshan-config for ld flags
adjusting post-ld-flags to handle more Cray test cases
filter out "cgroup" fs type
bug fix: mark open and close timestamps in pnetcdf
bug fix: track open and close timestamps in hdf5
bug fix: in runtime library, make sure that the file record reduction takes
bug fix: make darshan parser correctly handle partial file records
experimental workaround for stat problem on Hopper
debugging function for mount point table
bug fix to stat workaround
fix previous st dev workaround to be configurable, and to still use normal
stubs for new file tracking api
fix some warnings
new framework for tracking multiple concurrent open references per MPI rank
bug fix: don't double-count overlapping read/write/meta operation times
updated log format
rough Cray documentation
bug fix
added link to example of darshan-job-summary.pl output
updated changelog
fix bug reported by Mouhamed Gueye: file system names were not being
update Changelog for 2.2.0
Preliminary patch to linux-cc to enable Darshan on Cray XE6.
update patch to make it easier to apply
update cray patch to cover intel and pathscale compilers
add preliminary craycc patch
patches for cray programming environment (all compilers)
remove accidental fprintf
bug fix to error handling when using log-path-by-env. This fixes a problem
change log for 2.2.2 bug fix release
add module file for Cray installations
move Cray compiler script patch to darshan-runtime tree
draft Cray installation instructions
wording fix
editing
editing
fix typo
fix patch commands
update change log
minor tweak to headings
use MPI_Wtime() value in file name hash calculation to avoid collisions
add barriers between shutdown_bench test scenarios
update gen-cxx script to accomodate bg/q library names
update darshan module with feedback and examples from Yushu Yao
updated documentation
update version numbers in configure
update change log
update Changelog again
add a utility to test stat performance
add an lseek variation as well
print if lseek was used
track working directory
add ability to test realpath performance
revert previous commit
disable stat at open by default
should have been in last commit; configure script changes to support
first cut at fn to construct absolute paths
grab mount information at startup rather than shutdown
ability to detect mount point by path comparison
add disclaimer to file size table if files were not stat'ed at open time
update runtime docs to reflect --enable-stat-at-open option
update darshan-util docs to note possibility of -1 in statistics values
add note to darshan-parser output about -1 values
update changelog for 2.2.4-pre1
update configure scripts to 2.2.4-pre1
fix memory problem triggered by cp-shutdown-bench with latest trunk library
update version numbers to 2.2.4-pre2
add -lz to post ld flags to address link bug reported by Yushu Yao
update changelog and version numbers for 2.2.4-pre3
recommend --enable-stat-at-open option for BG installation
bump version numbers to 2.2.4-pre4
fix bug reported by Yushu Yao; compiler options being set incorrectly
update changelog and version numbers for 2.2.4-pre5
added warning/reminder to changelog
experimental fix for mvapich2 on ranger/tacc
propagate mvapich2/ranger fixes to cxx and fortran scripts
bump version numbers and update changelog for 2.2.4-pre6
add example x86_64 output for 2.2.4
added ppc32 example output for 2.2.4 release
added ppc64 (bgq) example output
bump version numbers to 2.2.4
split mount point information collection loop into two pieces to simplify
sort mount points so that we correctly match long mount point strings first
bump version number in trunk to 2.2.5-pre1
update documentation to no longer recommend --enable-stat-at-open anywhere
Bug fix to cray linking of non-mpi programs. When Darshan is enabled we
use fstat() in place of stat() when stat'ing files at open time
use statfs() on mount point to detect default block size for mount points
add --enable-group-readable-logs option
bump version number to 2.2.5-pre3
added example output from ppc64 (BG/Q)
fix warning
add patch for Cray xt-asyncpe 5.12 or higher, contributed by Yushu Yao
update documentation for Cray xt-asyncpe 5.12
update changelog and versions for 2.2.5
adding ppc32 example log
adding x86_64 example log
bug fix to DARSHAN_LOGHINTS handling
add init time instrumentation
switch unecessary allreduce calls to reduce calls
update changelog
expose timing hooks so that darshan-shutdown-bench can be executed without
rename cp-shutdown-bench to darshan-shutdown-bench
remove unused "iterations" argument to benchmark
add a 1 second delay between test scenarios in shutdown bench
update changelog
clarify shutdown "all" category
change default MPI-IO hints for writing log file
refactor mount point tracking
fix warning
implement broadcast of mount data from rank 0
fix minor memory leak
remove old logic for collecting device id via stat
bump to 2.2.6-pre1
not sure if we will keep this feature or not, but added --red-read option to
add some summary information for redundant reads
add diff in redundant read summary
add experimental --meta-ratio option
experimental option to find shared files with no collective I/O and writes
change algorithm for calculating meta ratio
revert experimental darshan-parser changes from trunk; see
Count freads and frwrites in read and write counts for
fix segfault on invalid arguments to darshan-parser; fixes #103
add notes on Open MPI compatibility
update documentation for CP_DEVICE field
update documenation re: Lustre file alignment
added more environment variable documentation
integrate contribution from Matthieu Dorier: installation process for
update changelog for 2.2.6-pre2
bump version numbers for 2.2.6-pre2
added example output from x86_64
add example output file from bg/p
example output from bg/q
bump changelog for 2.2.6
bump version numbers for 2.2.6 release
draft release checklist
experimental stub functions for HDF5 wrappers
add dummy stubs for pnetcdf wrappers
update changelog and bump version number for 2.2.7-pre1
revert hll dummy function changes
revise cray module file to use new mechanism for injecting Darshan
change order of module file a little
update the example module again for clarity
updated documentation to reflect new best practices for Cray installation
bug fix: stat() before open() should not set the device id for a file
update changelog
store library version number in log metadata
edit changelog
update version number for 2.2.7-pre1
fix typo
bump version numbers for 2.2.7 release
Support MPICH installations with const-correct prototypes. Also bump
fix bug in MPI_Init_thread() symbol resolution with LD_PRELOAD; reported by
development branch for darshan-parser changes
start stubbing in routine to list statistics per file
print file name suffix
start stubbing in normal and detailed file list
a few more fields in the file list mode
report slowest IO time in file list mode
avg. cumulative IO time in file list
print timestamps in detailed file list view
show open counts
read/write access size histograms
fix buffer overflow
add darshan-convert option to filter output to a specific file
git-svn-id: https://svn.mcs.anl.gov/repos/darshan/branches/darshan-parser-file-list@753 3b7491f3-a168-0410-bf4b-c445ed680a29
merged darshan-parser-file-list branch to trunk
integrate darshan-summary-per-file.sh utility, written by Rob Latham
minor enhancements to darshan-summary-per-file.sh script
a little more output
stub wrappers for aio functions
stub in tracker structure for aio_write64 example
incremental progress on tracking aio operations
match aio_return() with aio_return64() wrapper
clean up logic to add aio trackers
fill in remaining aio_read and aio_write functions
filled in lio_listio() wrappers
handle aio list in condense procedure
update Changelog for Darshan 2.2.8-pre2
fix 64 bit handling for aio functions
update release numbers in autoconf
adjust post ld flags to reflect that some systems will need to pick up aio
bump version number to 2.2.8-pre3
example output from bgq
add example output from 2.2.8
update cray module for Darshan 2.2.8
updated Changelog for 2.2.8-pre4
bump version numbers to 2.2.8-pre4
updated darshan-util documentation to reflect darshan-summary-per-file.sh
tweak changelog
bump version numbers to 2.2.8
proof read changelog
fix mnt table overflow, reported by David Shrader
update Changelog
quick script to pull compute node and runtime statistics out of an
show nprocs too
updating Changelog
fix const correctness in MPI libraries; partially addresses #126
remainder of fix for #126, disable internal benchmark routines that require
fixed -ldl link order bug that caused undefined dlsym error on some
add reminder to comments in darshan-convert-logs.pl
minor fix to install rule for darshan-util.pc
safety checks to separate metadata entries with newlines
add ability to reset old metadata when converting logs
saving some notes from Florin regarding how to add new counters
added warning to indicate when I/O timespan intervals are out of range, see
add wtime_offset field to job header to detect when MPI_Wtime() is not
minor fix to job-summary logic to handle wtime_offset
revert previous changes to log format and wtime_offset feature
implement new approach to handling TIMESTAMP offsets; the library now
Fixes #114; make darshan-util honor the CC variable for compiler to use.
minor bug fix: make sure that LDFLAGS (including possibly library path for zlib) is propagated to rules to build binaries
Fixes #128; document how rpath influences LD_PRELOAD of darshan library.
some reorganization to make it possible to separate craype 1.x and 2.x files
initial step to adding CrayPE 2.x support to tree: generate .pc file
add modulefile for craype 2.x deployments
update documentation to include Cray PE 1.x and 2.x instructions
update Changelog with craype 2.x information
add environment variable to optionally disable shared file reduction
fixes #124, overhauls the output shown with darshan-parser --perf
update darshan util documentation to reflect new darshan-parser --perf
fixes #130
fixes #132
Fixes #130 again; needed to boost metadata len in addition to job record
Fixes #123, track up to 64 file systems at run time, and prioritize storing
bug fix to timestamp normalization
fixes #135
short-term fix for buffer overflow, fixes #140
update changelog
spelling fix
branch for experimental weak symbol development
experimental: weak symbol stubs for pnetcdf and hdf5 functions
update Makefile for stub library install
fix pkgconfig typo
fix symbol names in weak stubs
sync weak-symbol-devel branch up to trunk
reintigrate weak-symbol-devel branch to trunk
bump version number to 2.2.9-pre2
added --enable-shared option to darshan-util configure script
added example x86_64 output file for 2.2.9
minor refresh of BG terminology in runtime doc
example log file for ppc64
bump version number to 2.2.9
things that should have been committed previously (before tagging 2.2.9)
another release checklist reminder
bump version numbers and changelog for 2.3.0
properly set version number for util pkgconfig
change open timestamp behavior
use reductions to account for skew across ranks in start and end time
bump version number
fix darshan-job-summary.pl bug reported by Matthieu Dorier
harms (96):
Early attempt at some code to crawl drashan log dir and print summarized
Darshan utilities for loading log files in to a mysql database.
SQL commands to generate analysis results from darshan data.
Fixed bugs, more robust.
Handle cases where count can be zero. (Need to check if fsstats should
make user/passwd/etc something default
Add 2 more agruments for passing checkpoint and restart files for
Updated to allow specifiying a restart file that will track progress and
Previous file was for mpif77. Updated to correct mpif90 script.
Initial changes for adding configure item for the JobId environment variable.
Added CP_JOBID macro to darshan-mpi-io.c and checked in latest autogenerated files.
Add rule to create preprocessed source files.
Support the jobid in the log file and display in the summary output.
Add table that display amount of data transferred on a per-filesystem basis
Bug fix for data per filesystem computation in darshan-job-summary.pl.
Fix divide-by-zero error in cases where total i/o for either read or write is zero.
Small program to exercise various aspects of darshan
Add min, max and variance to the darshan log that is relevant for shared files only. This is to aid in performance calculations.
Add additional timing info for the MPI_Bcast portion of shutdown.
First cut at displaying variance information.
Add a warning when per-file system data isn't shown for log version previous to 1.24.
Script that parses all the old example logs and generates PDFs. Then forces you to look at them.
New branch to address fix for trac #5
Initial changes to correct quote handling for compiler scripts generated from the generator scripts.
Updated compiler generation scripts to take a '--trim' option which will
Fix so that variance table of darshan-job-summar.pl encode file name in a latex safe way.
Change the ibdir target back to the original but still using the pipe
Merge trac5 to trunk
Fix minimum calculation; code was using the incorrect counter array.
Updates to the mysql loader. Commits transactions in largers sets and
Quick fix to handle cases where there are lots of shared files and the
Updated the parser to calculate derived stats via various command line
Add in open source hash implementation used in the darshan-parser.c
Add a verbose flag to retain the temporary LaTeX files.
Add warning when darshan log has no file records in it. Exists without
Add checking of exit status from pdflatex calls. If the pdflatex
Changed to support a finding support files via PREFIX variable set via
Changed so that darshan-job-summary.pl.in is created and
Fix for compiler warning.
Minor updates to improve distclean target.
Set the prefix to something close to valid but still requires it to be
Added a "random" value to the filename of the darshan log. This random
Fix out-of-tree builds and installs so they work again. (trac #28)
Update the Eureka and Gadzooks to install the entire Darshan suite since
Add new environment variables to override the configure based values.
Changes to allow creating both a static and dynamic version of darshan.
Rework how the ldpreload code is initialized. Each function checks to
Correct the usage of MPI_Wtime in darshan-posix.c and add a runtime
Add a check to turn off darshan if a compiler is using an option that causes
Set of test cases for checking performance estimation of darshan-parser
Fix the performance estimation in darshan-parser for file-per-process
Change lib/libdarshan.so target to be compiled with $(CC) like other
The PGI compiler doesn't appear to like pointer arithmetic. It was
Add another performance test that does partially shared files.
Fix the by_open performance calculation so the close timestamp must be
Updates for Darshan 2.1.0 release
Update release version in maint. scripts.
Inital log 'put' utilities for creating log.
Update the log file format so there's an extra generic field to put
Add obfuscation feature to darshan-convert.
Allow arbitrary metadata (key/value pairs) be added be convert utility to new log.
Fix reference to comment (now metadata) in darshan_job struct.
Fix copywrite.
Perl script to iterate through a list of logs and convert them, adding annotations.
Minor fixes for log conversion script.
Set the default darshan-job-summary pdf output based on the filename.
Split MPI_Init/MPI_Finalize off into its own object file so that it can be linked separately.
When the code was processing the file_record_hash it was using setting the CP_NAME_SUFFIX to the value for the new record instead of the old record. Switch the order CP_NAME_SUFFIX is updated.
procs was using the pre-incremented value for the total count, switched it to use the correct 'n' value.
Initial changes to maint scripts.
Remaining updates to maint. scripts. fixes 61
Partial fix for setting xrange on access time graphs
fixes 43. Added correct xrange notation based on timefmt.
Initial attempt at automated build and test
Add reporting for setup.sh
Fix automated build steps
Fix execution of magellan/configure.sh in common/configure.sh
Fix xml spec and darshan-util naming
Fix xml spec in setup.sh
Fix xml file creation
Add automated build phase scripts
Change scripts so that send info to standard out
Minor fixes to documentation
Workaround for case where cuserid returns nothing. Uses the users uid instead.
Add more fstype exclusions, none and rpc_pipefs
Setup stat calls to use same exclusion rules as open calls
Add work around for OSX and strndup
Fix for jenkins magellan based slaves. mpicc is OpenMPI and is now in the default path
Add support for CC and MPICH_CC in compiler wrappers
add support for alcf2 systems
fixes for previous commit; correct MPICH_* handling in wrapper generators, uncomments code in test script that should not have been commented out.
Restrict compiler selection to MPICH_{CC,CXX,F77}, allowing CC, et. al risks an infinite recursion.
fixes 106; Add test to see that MPI_Init is weak linked. This solves an issue with people using a fake MPI for various reasons
Update automated testing for new Jenkins instance
CSS changes should reduce padding around table cell contents and eliminate the bug with a line wrapping on top of itself.
fixes 120. fixes 121. This change prevents a crash in getopt on bad user arguments and prevents the add_annotation function from overrunning the job->metadata array if there is not enough space for the user provided string
robl (2):
listing darshan libs twice gets us trace files for testpio
fix 'tmp might be used uninitialized' warning
rross (1):
added check for mntent.h, mount.h (probable replacement), conditional include of mntent.h. this cleans a couple of things up, but reported
ssnyder (4):
added darshan-util pkgconfig file
add out-of-tree build support to darshan-util package config file
Update darshan configure, installed headers, and darshan-util pkg-config file to allow darshan-util lib to be linked with external software
export gzip include/link paths in pkg-config
-----------------------------------------------------------------------
hooks/post-receive
--
More information about the Darshan-commits
mailing list