00:07 <fche> not dead yet

00:07 <kerneltoast> and how are you doing

00:07 <fche> not dead yet

00:07 <kerneltoast> excellent

00:07 <fche> though my older boy tried to get us stuck in a mud rutty road during his driving lessons

00:08 <fche> so there was that :)

00:08 <kerneltoast> ah being a new driver and thinking the car is invincible

00:08 <kerneltoast> just get bigger tires

00:09 <kerneltoast> no such thing as bad terrain, only bad tires

00:09 <fche> will consider that :)

00:09 <fche> I think he was spoiled by doing a jeep-hike-on-mountain ride a few years ago

00:10 <fche> but we don't have such a vehicle ourselves, alas

00:10 <fche> ANYWYAYYAYYAYAAAAYYYYY

00:11 <fche> so yeah the rawhide buildbot is still chugging on your second-last commit ("disable preempt")

00:11 <fche> by tomorrow morning that one and the next one should get a verdict

00:12 <kerneltoast> if it still dies, i suspect it'll be due to this: https://gist.github.com/kerneltoast/8ee468c43b800888639a1a406b5cbc22

00:12 <kerneltoast> i need to get onto the bulkmode stuff but i've been trying to finish up the print patch

00:12 <fche> aha, so next step should be to get rid of that silly business with the stp-bulkmode work

00:12 <fche> we can split that

00:12 <kerneltoast> yeah... though i wanted to get the print thingy pushed first

00:13 <kerneltoast> because bulkmode is semi kinda slightly dependent on it

00:14 <kerneltoast> it's the prints in random areas like the module refresher and systemtap_module_exit that cause this sorta issue

00:14 <kerneltoast> bulkmode needs log modifications to be done with at least preempt disabled

00:14 <kerneltoast> holding the context does that, so probes are covered

00:14 <kerneltoast> but other random prints aren't covered as of right now

00:15 <kerneltoast> they will be covered by a local_irq_save with my print patch though

00:15 <kerneltoast> which will wrap things up nicely to go straight to bulkpainmode

00:16 <kerneltoast> current state of me print patch: https://gist.github.com/kerneltoast/7fdf4a1c3edbb4a7a0709bb9559aacf7

00:17 <fche> I'm a believer in bulkpain in one's adversary

00:27 <kerneltoast> have you tried replacing your car's motor with a utrace_engine?

00:27 <kerneltoast> i've heard they're quite powerful

00:27 <fche> not helping

00:30 <kerneltoast> i saw the pun and i went for it

00:43 <fche> I respect the pun hustle

02:00 <kerneltoast> fche yo

02:02 <fche> no news is no news

02:02 <kerneltoast> i was gonna ask something else

02:02 <fche> ok use small words please

02:02 <kerneltoast> i want to push small change to make mutex_trylock deadlock go away until bulkmode comes

02:02 <kerneltoast> so your buildbots can actually test stuff

02:03 <kerneltoast> it would be a temporary change

02:03 <kerneltoast> REVERTME: make trylock deadlocks begone

02:04 <fche> why would it be reverted as opposed to nuked becuase the mutex/inode goo would go away entirely?

02:04 <kerneltoast> fche, https://gist.github.com/kerneltoast/64834e272251a4eab228e3287939c436

02:05 <fche> ah that one

02:05 <fche> sure

02:06 <kerneltoast> koolbeans

02:06 <fche> yes I am

02:06 <fche> beans that are cool

02:11 <irker334> systemtap: sultan systemtap.git:master * release-4.4-17-g6a27888b1 / tapset-timers.cxx: REVERTME: tapset-timers: work around on-the-fly deadlocks caused by mutex_trylock

02:11 <kerneltoast> there we go

02:12 <fche> ok

02:12 <kerneltoast> buildbots should survive now

02:12 <fche> but you

02:12 <kerneltoast> but i

02:12 <fche> stop writing patches faster than the buildbots can test them

02:12 <fche> please

02:12 <fche> take a cocktail

02:12 <fche> have a nap

02:12 <kerneltoast> hah

02:12 <fche> RELAX A DAY

02:12 <kerneltoast> i feel like i'm slacking actually

02:12 <kerneltoast> waiting for this darn testsuite to finish

02:12 <kerneltoast> and then seeing it die

02:12 <kerneltoast> so then i write more patch poo

02:14 <kerneltoast> what do you usually do if not write stap patches?

02:14 <fche> think about writing more stap patches?

02:15 <kerneltoast> redhat doesn't throw anything else at you?

02:15 <fche> at some point we should shrink the stap testsuite - eliminate unnecessary stuff

02:15 <fche> am hoping serhei's bunsen gadget will be able to help with that too

02:16 <kerneltoast> i think we should make parallel testing the norm

02:16 <fche> yeah possibly

02:16 <kerneltoast> parallel testing should be good once the probe_lock lockup is fixed

02:17 <fche> maybe - though separate stap sessions should not interfere with each other at all anyway, except perhaps cpu contention

02:17 <fche> they don't share locks etc

02:17 <kerneltoast> yeah i dunno why the lockup happens

02:18 <fche> well

02:18 <kerneltoast> it'll be an xmas miracle if these bugs are solved by the end of the year

02:18 <fche> I heard you have some spare time while you're waiting for testsuite results!! :)

02:18 <kerneltoast> 😊🔫

02:18 <fche> don't do it

02:19 <fche> puns need you

02:19 <kerneltoast> I'LL SQUIRTGUN MYSELF

04:08 <agentzh> fche: i've been pushing kerneltoast. so blame me ;)

04:10 <agentzh> kerneltoast: i really hope we can fix them WAY before xmas :)

05:20 <kerneltoast> fche, testsuite results for the patch in this gist: https://gist.github.com/kerneltoast/7fdf4a1c3edbb4a7a0709bb9559aacf7#file-systemtap-sum-diff

05:23 <agentzh> kerneltoast: are these expected? https://gist.github.com/kerneltoast/7fdf4a1c3edbb4a7a0709bb9559aacf7#file-systemtap-sum-diff-L66

05:23 <agentzh> +FAIL: hrtimer_onthefly - otf_start_disabled_iter_1 (invalid output)

05:36 <kerneltoast> yeah idk what's up with those

11:24 <fche> agentzh, those invalid output messages are I believe related to diagnostic message format changes made some time ago that just weren't updated in the testsuite

11:24 <fche> so no big deal

16:28 <kerneltoast> fche, the test results I'm comparing this to were generated with master at the commit before my workaround patch

16:34 <kerneltoast> so those failures must be caused by my print patch

16:59 <irker912> systemtap: alizhang systemtap.git:master * release-4.4-15-gd1804e051 / : Conscious language initiatives: replaced whitelist->passlist, blacklist->blocklist, master->main/primary. Some occurences of master and slave may not be able to be replaced at this point, eg. name of a terminology or usage of other programs interface.

16:59 <irker912> systemtap: alizhang systemtap.git:master * release-4.4-19-gf02f08bc6 / : Merge branch 'master' of ssh://sourceware.org/git/systemtap into concious_language

18:18 <agentzh> fche: got it.

18:19 <agentzh> kerneltoast: ah, check the detailed systemtap.log for those failed test cases?

18:19 <agentzh> or just run those test files individually?

20:24 <irker912> systemtap: fche systemtap.git:master * release-4.4-20-g963c215e7 / NEWS configure configure.ac doc/SystemTap_Beginners_Guide/en-US/Book_Info.xml systemtap.spec testsuite/configure testsuite/configure.ac: post-release prep: bump version number to 4.5

20:49 <kerneltoast> agentzh, fche, i fixed it

20:49 <kerneltoast> the onthefly failures are now caused by the REVERTME patch, and my print patch is good

20:50 <kerneltoast> this was the problem: https://gist.github.com/kerneltoast/7fdf4a1c3edbb4a7a0709bb9559aacf7#file-0001-runtime-fix-print-races-in-irq-context-and-during-pr-patch-L152

20:50 <kerneltoast> shouldn't have had an extra +1 there

20:51 <fche> an off by one

20:51 <fche> surely you jest

21:00 <irker912> systemtap: sultan systemtap.git:master * release-4.4-21-g7187dcf39 / runtime/linux/io.c runtime/linux/print.c runtime/print.h runtime/print_flush.c runtime/stack.c runtime/transport/transport.c runtime/vsprintf.c translate.cxx: runtime: fix print races in IRQ context and during print cleanup

21:00 <kerneltoast> DONE WITH THE PRINT PATCH

21:01 <kerneltoast> now time to do bulkmode

21:48 <agentzh> kerneltoast: yay