This is a log of changes to pounder21. pounder30-2011-08-09 - Created new documentation CONFIGURATION and moved it and SCHEDULER into a newly created doc/ directory - Deleted the test-all test scheduler - Created /schedulers directory and moved the remaining test schedulers there - Removed option to specify "NONE" when asked to unpack test scheduler during build - Removed check for existing kernel directory in /tmp in test_scripts/build_kernel since it appears that some files get lost after running build_kernel once; Instead just untar the kernel each time we run the test script to be on the safe side - ltp test script would pass even if it didn't build currently, fixed this in test_scripts/ltp - changed ltp build_script to install ltp to $POUNDER_TMPDIR - removed QUICKSTART and included it in README instead - removed trailing "/" from POUNDER_LOGLOCAL export in libpounder.sh - Added functionality for automatic skipping of subtests (see README) - Created xterm_stress build script and merged 00xbonkers with it - Created ide_cdrom_copy build script and merged 00check_cdrom_presence with it - Merged nasm and schedutils build scripts with the lame build script - Merged time_test build script with the time_consistency and time_drift build scripts - Created test_repo/ directory - Uncommented a piece of code in time_drift that allowed it to always pass - Added pounder -c option for creating new test schedulers - Modified POUNDER_VERSION in libpounder.sh pounder30-2011-07-21 - Updated bonnie++, ipmitool, kernel (used in build_kernel), and memtest script to latest versions - Updated memtest build scripts and $POUNDER_HOME/src/memtest.patch - Added functionality for skipping of subtests - Added functionality for automating the skipping of subtests (see README) - Removed unnecessary 00checklatest test - Moved checking for system requirements from test run to build phase - Affects bonnie++, memtest, cpufreq, and ide_cdrom_copy - Added environment variable MAX_FAILURES that, if defined, sets an upper bound on the number of failures a looped test will allow before aborting the test altogether (see SCHEDULER) - Added functionality for removing and re-adding subtests to the test scheduler (see SCHEDULER) - Updated README, SCHEDULER, and config files pounder21-2011-04-08: - LTP: Updated to LTP 20101031 release. - Build kernel testcase - Updated kernel from 2.6.18 to 2.6.38. - Updated 2.6.38 kernel source tar in pounder cache. - Did corresponding kernel changes i.e for 2.6.38 in "memtest" testcase too. - Files modified are:- -$POUNDER_HOME/test_scripts/memtest. -$POUNDER_HOME/test_scripts/build_kernel . -$POUNDER_HOME/build_scripts/memtest. -$POUNDER_HOME/build_scripts/build_kernel. -$POUNDER_HOME/opt/memtest.sh. [Actually this file need to get changed in tux1 cache]. pounder21-2011-04-12 -Integrated bash-memory testcase in pounder21 -Files added/modified are:- -Copied bash-memory test case tar to pounder cache. -Added file $POUNDER_HOME/build_scripts/bash-memory -Added file $POUNDER_HOME/test_scripts/bash_memory -Added file $POUNDER_HOME/tests/T10single/T03bashmemory pounder21-2006-11-07: - Fix a bug in randasys on x86-64 where we had insufficient random bits and would truncate whatever we got, leading to all 0 arguments by simply generating enough random bytes to fill an unsigned long. Also add a -z switch to enable this zero-only mode because it found some bugs. :) pounder21-2006-10-12: - Include /sysfs contents in the hw inventory - Put 'lspci' into the PCI inventory for easy finding. pounder21-2006-09-23: - Various time test fixes from jstultz. - IPMI tests from James Simshaw/Carol Hebert. - Update kernel from 2.6.15 to 2.6.18. pounder21-2006-01-24: - Establish all files in the tarball as originating from IBM LTC. - License all files under the GNU Public License (GPL) v2. - Install: Make USE_CACHE=0 if the pounder cache dir is not set. - Makefile: distclean -> mrclean for more GNUness. - nfs_logging: Clarify some of the documentation, and use TCP NFS mounts. - 00checklatest: Handle leading zeroes in date components. - build_kernel: Update to kernel 2.6.15. - ltp: Update to 20060105 release. - memtest: Update to kernel 2.6.15. - cpufreq: Report the return code in the log, and kill cpufreq programs when it's time to terminate. - randasyscall: Add -Wall to CFLAGS and fix warnings. Also add -d option to log every call made. NOTE: That may affect race conditions exposed by this program! - drift_test: Only set system time from NTP source if -s option is passed (John Stultz) - inconsistency_check: Use CLOCK_MONOTONIC if available (John Stultz) - nfs test: Use TCP NFS connections. - copytree: Diff source and dest before deletion. pounder21-2005-12-02: - Make CHANGELOG reflect the CVS history. - pounder: Make the -d switch (total test duration) work again. - check_cdrom_presence: Remove unnecessary variable. - build_kernel: Update kernel to 2.6.14.3. - libidecd.sh: Grep for "^$DEV[<space><tab>]" so that we ONLY look for full device names in the first columns. Also update documentation to reflect the code intent better. - memtest: Update kernel to 2.6.14.3. - time_tests: New code from John: o Use CLOCK_MONOTONIC (calls to settimeofday won't affect this) o Takes an argument to run for a set number of seconds(nothing or -1 == run forever) o Exits if any time inconsistency is found. o Outputs the delta size of the inconsistency found - mem_alloc: Only run every 15min. Constantly being OOM is annoying. - nfs: Better status printing during cleanup. - random_syscall: Put the program in /tmp again. pounder21-2005-11-18: - Install: Check for g++ and lex. - statslogging: Copy README to the log dir to capture version and relevant notes. - randasys: Use $POUNDER_TMPDIR, not /tmp, and use quotes with variable expansions. Also forbid calling settimeofday(2). - checklatest: Be more intelligent about figuring out the version number--if what we're running is newer than what's on the server, that's fine. We're probably just running a development version. - cpufreq: This is a new test that detects the presence of speedstep, powernow, or whatever other CPU frequency changing mechanisms are in place. If present, it will cycle the CPUs through all power state changes in arbitrary order. It's also clever enough to find and ignore CPUs whose frequency is tied to another processor. - pounder: Fix argument passing so that -l switch works. Added -s switch to disable NFS logging. - nfs_logging: Move variable definitions into libpounder.sh and flatten the function calls into a more procedural script. - ide_cdrom_copy: Total rewrite of this script: The old script did not detect SCSI CD-ROM drives at all unless /dev/cdrom was set up properly (this is not the case if you netinstall RHEL4 on a blade with the media tray disconnected) and it did not deal with multiple optical drives at all. This new one creates a list of potential CD/DVD devices and tries to mount each of them; each drive that has a disc inserted is then queued up for testing. Also, the actual device detection code is in src/libidecd.sh so that the 00check_cdrom script can take advantage of it. - 00check_cdrom: Rewritten to use libidecd.sh; if no discs are found, all drive trays are ejected. - default/fast-tests.tar.gz: Updated to incorporate cpufreq test. - libpounder.sh: Added and rearranged variables as required for NFS logging support. pounder21-2005-11-03: - fast-tests: Run mem_alloc - statslogging: Start the logfile tail. pounder21-2005-11-01: - README: pounder 2005-10-31 - nfs_logging: Put log files in /crash/pounder-logs, not /crash. Also use uname-arch as a key, not just uname. - pounder: Add help() func and -r flag to remount log storage. pounder21-2005-10-28: - Sarunya's NFS logging support. - Sarunya's memory allocation stress test. - NFS test only removes what it copies (it was deleting the per-host directory too). - IDE CDROM copy test only probes the first CDROM device found in fstab... rather than grabbing both entries and breaking. pounder21-2005-10-11: - Check CDROM presence when installing pounder. (patch by Sarunya) - Fix NFS test's diff failure detection. (patch by Darrick) - Clamp ramsnake so that it won't use more than 1/4 of the address space so that i386-bigsmp boxes won't fail. (found by Russ, patch by Darrick) pounder21-2005-09-20: - README: pounder 2005-09-20 release. - get_hw_info: Add missing '-l' to list partitions. - nfs: $SHORTHOST needs to extract from $HOSTNAME, not $SHORTHOST. This was causing nfs failures. - xterm_stress: Remove old log files before starting X server, copy the log file if the X server failed to start, and kill the X server if there's a failure to start. pounder21-2005-09-19: - get_hw_info: Record partition layout. - ramsnake: Kill program after 300s, not thirty! pounder21-2005-09-09: - checklatest: Download tarball to $POUNDER_HOME/../ so that untar goes in the right place. - xterm_stress: Make script use kill -9 on X servers that ignore -4 after 180s and abort the script if it's still not dead after 360. pounder21-2005-08-31: - nfs: Strip domain bits (".ibm.com", ".west.foo.com", etc) out of the $HOSTNAME, put that into $SHORTHOST, and use that on the NFS server so that we cut down on the number of bogus dirs and crud on that filesystem. pounder21-2005-08-26: - xbonkers: Program to make windows in an X session move all over the screen. Add this to xterm_stress so that our poor window placement doesn't much matter anymore. - xterm_stress: We're allowed 1 fatal error at the end because killing X is a fatal error. - time_consistency: We're NOT allowed to excuse any errors. pounder21-2005-08-17: - More fixes to make sure the xterm_stress test dies when it should, an LTP refresh, and a couple of patches designed to make pounder work better with the mainline testing rig. I haven't seen any problems with it, so I'm pushing it out to the group to find more error cases. :) pounder21-2005-08-16: - README: Add warning that changing the first line is a bad idea. - Install: Build helpers after building tests, so that the 00checklatest script runs immediately. - checklatest: Compares this pounder's version (as defined as the sixth column in README) against what's on the server. If there's a mismatch, ask the user if he wants to download the new one. - libpounder.sh: Add a $POUNDER_VERSION variable and location of $POUNDER_HEAD. pounder21-2005-08-15: - NFS test: diff whatever it is we copy. Hope that nobody's changing /usr... - X test: Use `startx' to start X with the appropriate KDE/GNOME session, and start some OpenGL tests too. pounder21-2005-07-13: - Modify build scripts to use a download_from_sourceforge fxn that will retry failed downloads with other SF mirrors. pounder21-2005-04-04: - No more bluebench. pounder21-2005-02-25: - Fix install failure on SLES8. - New test to stress concurrent memory access on shared memory on numa-ish systems. Well, any system with shared memory...