Bug 2284431

Summary: 6.2.2 regression: pmResult.numpmid has bogus values on i686
Product: [Fedora] Fedora Reporter: Martin Pitt <mpitt>
Component: pcpAssignee: Nathan Scott <nathans>
Status: CLOSED CURRENTRELEASE QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: unspecified    
Version: 39CC: agerstmayr, jkurik, lchilton, nathans, sfeifer
Target Milestone: ---Keywords: Regression
Target Release: ---   
Hardware: Unspecified   
OS: Linux   
URL: https://kojipkgs.fedoraproject.org//work/tasks/6753/118496753/build.log
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2024-07-30 06:37:08 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Martin Pitt 2024-06-03 09:56:49 UTC
The recent Cockpit update in Fedora 39 failed to build due to unit test regressions on i686 [1]. This started with pcp-6.2.2-1.fc39 but *only* on i686. Later Fedora version don't build pcp on i686 any more, so the status there is unknown (and uninteresting).

I can reproduce this with `fedpkg mockbuild -N --mock-config fedora-39-i386`, and --shell.

When I downgrade to 6.2.1, doing a pmFetch() and iterating over pmResult.vset[i] for i between 0 and pmResult.numpmid works fine. In our tests we usually expect small values like 1 or 2.

But with 6.2.2, the test started give completely bogus results:

(gdb) p *result
$2 = {timestamp = {tv_sec = 1717407857, tv_usec = 0}, numpmid = 998643, vset = {0x0}}

or in another run

(gdb) p *result
$2 = {timestamp = {tv_sec = 1717407619, tv_usec = 0}, numpmid = 233042, vset = {0x0}}


This is unreasonably high and looks rather random (corrupted memory), and vset is NULL.

I don't have a standalone reproducer yet, just `./configure; make check` in Cockpit.

[1] https://koji.fedoraproject.org/koji/taskinfo?taskID=118496753 .
[1] https://bodhi.fedoraproject.org/updates/FEDORA-2024-729b5ca69e

Reproducible: Always

Comment 1 Martin Pitt 2024-06-04 11:09:31 UTC
This also affects C9S now, see https://kojihub.stream.rdu2.redhat.com/koji/taskinfo?taskID=4160964

Comment 2 Nathan Scott 2024-06-05 04:46:46 UTC
Hi Martin,

Thanks - this is some known fallout from time64_t changes in the last upstream PCP release.
It's being reworked for pcp-6.3.0, I'll see if I can get a back-port/revert build to sort out 6.2.2 in the meantime.

Comment 3 Martin Pitt 2024-06-05 05:04:12 UTC
Thanks Nathan! Let me know if you want me to verify a test build/copr or similar.

Comment 4 Nathan Scott 2024-07-30 06:37:08 UTC
Resolved some time back by reverting the change.