commit cfa48f508fe83be9c9a63010ddac44fe650c2d98
Merge: 7c3b63fd 784b4d8b
Author: Ken McDonell <kenj@kenj.id.au>
Date: Fri Apr 10 14:42:26 2026 +1000
Merge branch 'main' of https://github.com/performancecopilot/pcp
commit 7c3b63fd2c5c28ed7cc806f7183e19ca4388d67c
Author: Ken McDonell <kenj@kenj.id.au>
Date: Fri Apr 10 14:33:14 2026 +1000
COOKBOOK.md: reworked <a ...>...</a> index entries
Alternate format (anchor does not span any text, index {entry} is first
word after the anchor) is understood (now) by fix-md and seems to work
better in terms of rendering and link navigation for github, okular
and retext.
commit 7c533b7ad51988c1c7c5ac3d294627f1b6a22b7b
Author: Ken McDonell <kenj@kenj.id.au>
Date: Fri Apr 10 14:31:02 2026 +1000
scripts/fix-md: rework emnbedded index syntax
Citations used to be of the form
<a id="idx+{ident}">{entry}</a>... (all on the same line)
but now I also support the form
<a id="idx+{ident}"></a>{entry}... (all on the same line)
The first form is lexically better because the index term is
defined by the span of the anchor.
But the second form is better for some MD renderers where the
anchor does not span any text (and in particular any font changes)
... in this case {entry} ends at the first non-word character
in the set [ ,.<:|)]
commit 12d2b297826ed12af70a3c1214b25a1f803ebdb5
Author: Ken McDonell <kenj@kenj.id.au>
Date: Fri Apr 10 07:17:29 2026 +1000
COOKBOOK.md: experiment with anchor span for github renderer
commit ac2af542a9c601761df2b33446ca618d0766ec66
Author: Ken McDonell <kenj@kenj.id.au>
Date: Fri Apr 10 07:12:08 2026 +1000
qa/COOKBOOK.md: some font change tweaks and _ escaping
commit 8e1d8d82b5a06bf95cd75ddac68d0a2cde6e1c5e
Author: Ken McDonell <kenj@kenj.id.au>
Date: Fri Apr 10 06:22:29 2026 +1000
qa/COOKBOOK.md: small tweaks
commit 3b4a3a82bdd18f2b89d6333277fb11e05f7bb839
Author: Ken McDonell <kenj@kenj.id.au>
Date: Fri Apr 10 06:21:44 2026 +1000
scripts/fix-md: also remove ":" from charset for anchors
commit 3c5a221d2616d3294fec436ec578c43dbc575292
Author: Ken McDonell <kenj@kenj.id.au>
Date: Fri Apr 10 06:19:45 2026 +1000
qa/common.check: _private_pmcd() improvement
From COOKBOOK.md authoring and review, it seems sensible to add an
optional pmcd config file as an argument.
commit bc2960ae6543b8d594f2b0bffd0204be7d45750d
Author: Ken McDonell <kenj@kenj.id.au>
Date: Thu Apr 9 17:30:29 2026 +1000
COOKBOOK.md: move a section
commit 7b486508e7268f59cde55a60ee9b50b817a92ca0
Author: Ken McDonell <kenj@kenj.id.au>
Date: Thu Apr 9 17:15:41 2026 +1000
qa/COOKBOOK.md: close to version 1.0
commit 8014b595280d6708abf8c7d31ebb046068d5ea2b
Author: Ken McDonell <kenj@kenj.id.au>
Date: Wed Apr 8 10:40:37 2026 +1000
qa/COOKBOOK.md: initial version almost done
Also ome small changes in qa/common.check and qa/common.filter
that were noted in passing.
commit 4942b0fc0f09e18740d22fad7e396f039cbf8c4d
Author: Ken McDonell <kenj@kenj.id.au>
Date: Wed Apr 8 10:39:50 2026 +1000
scripts/fix-md: (#) cannot be a link ...
commit 816c066e5a628cec23eee40836e7e2e172f29634
Author: Ken McDonell <kenj@kenj.id.au>
Date: Wed Apr 8 10:38:07 2026 +1000
scripts/man-spell: better hqandling of markdown documents
For .md files ...
- less troff hacks
- more anchor-and-link hacks
For troff ... no changes.
commit 784b4d8b8f9378b0a82c81ecee0099efb55656f0
Author: kurik <jkurik@redhat.com>
Date: Tue Apr 7 09:01:56 2026 +0200
Fix of a bogus date in redhat.spec
May 30 2026 is a date in the future causing rpm-build to fail. This commit fixes the typo and sets the real release date to Mar 30 2026.
commit 609e66b6a2f851722efc8ae3a4dc8dade4279f78
Author: Ken McDonell <kenj@kenj.id.au>
Date: Mon Apr 6 11:43:24 2026 +1000
COOKBOOK.md: completed up to the end of common.check
Also some small changes to common.check and tweaks to scripts/fix-md.
commit 093997435c0854724459157ea53761d09981cb6d
Author: Jan Kurik <jkurik@redhat.com>
Date: Tue Mar 31 15:39:11 2026 +0200
Fix heap corruption crash in pmproxy
Fix heap corruption crash (SIGSEGV) caused by unsafe cross-thread
libuv timer operations by deferring all timer handle manipulation
to the event loop thread via uv_async_t.
commit 35e784df741b90896212c191729343150f3fd559
Author: Ken McDonell <kenj@kenj.id.au>
Date: Thu Apr 2 07:44:19 2026 +1100
qa: retire some old Irix warts and wrinkles
It is coming up for 20 years the last Irix release (6.5.30 in Aug 2006)
and so it is probably time to retire some more Irix-isms, especially in
the QA suite.
- no one has used Purify since then (cull *purify* functions from
common.check and COOKBOOK.md, cull purify from group file)
- the Irix version of netstat(1) is no longer interesting
- and we haven't had a 64-bit Irix platform in the QA mix since
then (qa/272)
commit 914a470424cef3766ec820d48b9dd548e6312415
Author: Ken McDonell <kenj@kenj.id.au>
Date: Wed Apr 1 13:47:20 2026 +1100
qa/COOKBOOK.md: interim checkin
commit c12170937bd6c1798867675571a31105b280abb8
Merge: 94a9f2b9 c09974f4
Author: Ken McDonell <kenj@kenj.id.au>
Date: Wed Apr 1 08:10:18 2026 +1100
Merge branch 'performancecopilot:main' into main
commit 94a9f2b9f018f491de0396f7b8ad711a109e0047
Author: Ken McDonell <kenj@kenj.id.au>
Date: Wed Apr 1 07:29:40 2026 +1100
qa: audit and fix _stop_auto_restart() and _restore_auto_restart() usage
Across multiple QA tests, ensure ...
1. _restore_auto_restart() is only called for a service if there has
been an earlier call to _stop_auto_restart() for the same service
2. in _cleanup() in particular, make sure _restore_auto_restart()
comes *before* _service start|restart for the same service
In the QA Farm there has been instability for multiple tests over a
long time where services failed to start|restart typically after an
otherwise successful QA test execution. Enforcing the protocols above
seems to have improved the situation.
commit c09974f432b5ebb45d9665d6ebe418adf31651f5
Author: kurik <jkurik@redhat.com>
Date: Tue Mar 31 09:00:12 2026 +0200
Fix of broken redhat.spec file
Fix of a syntax issue - forgotten `%endif` after the `%files` list extension.
commit 67a67be9687e088d2697059de77a19a614285424
Author: Ken McDonell <kenj@kenj.id.au>
Date: Tue Mar 31 07:28:12 2026 +1100
qa/group: rename group BAD to not_in_qa_farm
Similar to not_in_ci, these tests are so badly behaved in the QA Farm
that they either hang QA runs or produce bad side-effects for subsequent
tests. They are apparently immune to triage.
Initial tests in this category are
1703: Exercise the bpftrace PMDA - run unittests
1948: pmdaopentelemetry scalability testing
Also change pcp-daily to use -x not_in_qa_farm with check.
commit 06814407bd25378f2e8c47339148d04215c25284
Author: Ken McDonell <kenj@kenj.id.au>
Date: Mon Mar 30 13:35:11 2026 +1100
qa/triaged: tag 1222 for 32-bit platforms
Problem for linux proc PMDA and libnl is subject to on-going triage
and diagnosis.
commit 317e9601b6b882b62eb7c42bf5b6aff71689064e
Author: Ken McDonell <kenj@kenj.id.au>
Date: Mon Mar 30 13:31:57 2026 +1100
Revert "qa/1222: add valgrind suppressions for libnl botch on 32-bit systems"
This reverts commit 033abaca07f7bd7f9c90ba4d909b71f88585f988.
Change of plan, deal with this via triaged.
commit a08ce161db7dd6b5def45387f3479248d61dbe6b
Merge: 272107ad 033abaca
Author: Ken McDonell <kenj@kenj.id.au>
Date: Mon Mar 30 13:31:21 2026 +1100
Merge branch 'nextrel'
commit 272107ad15844b1e369516553a579247b991c2ba
Merge: 4102a620 e6fc8547
Author: Nathan Scott <nathans@redhat.com>
Date: Mon Mar 30 13:15:33 2026 +1100
Merge branch 'orasagar-pcp_mpstat_enhancement'
commit e6fc85476fedd2cd9d65f005d8421b8846771673
Merge: 4102a620 7e32eb2c
Author: Nathan Scott <nathans@redhat.com>
Date: Mon Mar 30 13:15:16 2026 +1100
Merge branch 'pcp_mpstat_enhancement' of github.com:orasagar/pcp into orasagar-pcp_mpstat_enhancement
commit 4102a620334daaa4af52906e1e710ecb2c7f66d6
Author: Nathan Scott <nathans@redhat.com>
Date: Mon Mar 30 13:13:26 2026 +1100
docs: bump changelogs and version for next planned release
commit 033abaca07f7bd7f9c90ba4d909b71f88585f988
Author: Ken McDonell <kenj@kenj.id.au>
Date: Mon Mar 30 08:05:25 2026 +1100
qa/1222: add valgrind suppressions for libnl botch on 32-bit systems
Observed on Debian-based systems.
Also some cleanup of the filtering and remove the proc.numa_maps
metrics that are not available on (at least some) of these systems.
commit ed88ddcaa27ecd4c440880b31d14441abacf5fb9
Author: Ken McDonell <kenj@kenj.id.au>
Date: Sun Mar 29 15:45:50 2026 +1100
rc scripts: rework "enabled" tests
This is for FreeBSD in particular where any script in /usr/local/etc/rc.d
will be called and each is responsible for consulting /etc/rc.conf to
see if they are "enabled".
More generally, if our wrapper scripts are installed in the place where
init (or similar) runs things from, we may need to do platform-specific
tests to see if a particular PCP service is indeed enabled.
This is all immaterial for systemd systems where the infrastructure
deals with the "enabled" state and does no call scripts for disabled
services.
- add a new is_enabled() function into src/pmcd/rc-proc.sh ...
today, it returns true for all platforms other than freebsd,
and for freebsd interrogates /etc/rc.conf for most start actions
(forcestart is a special action for FreeBSD that explicity
by-passes the /etc/rc.conf enabling controls)
- call is_enabled() from the rc_wrapper scripts for all actions
except status ... pmie, pmlogger and pmproxy
- pmcd does not have a wrapper script, so is_enabled() needs to
be called from src/pmcd/rc_pmcd
commit b46a82a193dda262dcb623b5601b352d916a8629
Author: Ken McDonell <kenj@kenj.id.au>
Date: Sun Mar 29 07:31:29 2026 +1100
configure.ac: include early explicit tests for headers autoconf will use
Specifically:
stdio.h stdlib.h string.h inttypes.h stdint.h unistd.h strings.h
sys/stat.h sys/types.h
This bunch need to be in confdefs.h so that later fragments of
conftest.c will compile properly ... on OpenBSD I saw the netinet/tcp.h
test fail incorrectly because <sys/types.h> was conditional included
in conftest.c but HAVE_SYS_TYPES_H had not (yet) been set.
commit 7e32eb2cd38ad8d3a5b2c78462cb9c81d8a8b7eb
Author: Sagar Sagar <sagar.sagar@oracle.com>
Date: Fri Mar 27 11:44:19 2026 +0000
pcp-mpstat.py: fixed build issues
commit 1fe757f64876cf4c8f8e85ab21d6a952fbb24dc3
Author: Sagar Sagar <sagar.sagar@oracle.com>
Date: Fri Mar 27 09:22:29 2026 +0000
qa: added pcp-mpstat4 archive coverage for interrupts
Introduce QA test 1782 with expected output to exercise pcp-mpstat
reports against the pcp-mpstat4 archive, including interrupt and
softirq views for multiple CPU selections.
changed the output in 883.out where it was just showing the cpu number instead of output for softirq and inq.
Now it clearly throws warning for user
commit 0ea63672ad9d68e155144fa56b61984bcbda97f8
Author: Sagar Sagar <sagar.sagar@oracle.com>
Date: Thu Mar 26 13:09:09 2026 +0000
pcp-mpstat: interrupt ordering for -I option
Added ordering strategies for hard/soft IRQ columns, prioritize
numeric IRQs and canonical softirq sequence, and skip tables with
clear messages when per-CPU metrics are absent; document behavior in
the man page.
this change has been done to match the mpstat output
There are 4 open security issues in bookworm.
You can find information about how to handle these issues in the security team's documentation.
![[bug history graph]](https://qa.debian.org/data/bts/graphs/p/pcp.png){kind=link}