14:27 <franks2> Hi. Im trying to use systemtap on PHP. However the function__return are not displayed. Ive posted my scripts and info at: https://paste.ubuntu.com/p/XRdwXczHDB/ Can anyone spot whats wrong?

15:04 <fche> franks2, stand by a few mins, will try to help shortly

16:49 <gromero> fche: howdy

16:50 <gromero> I'm preparing to ship fp patch for fp regs in ppc64

16:50 <gromero> on running make check I see a couple of tests failing

16:50 <gromero> I'm using upstream kernel

16:50 <gromero> so I'm wondering if it's expected that some tests will fail anyways

16:51 <gromero> I don't have a big picture on what's current status for the tests on ppc64

16:54 <gromero> I don't think tests failing have any impact for the fp patch, but well, it's good to know : )

17:10 <fche> yeah some failures are normal

17:11 <fche> the main thing to look for in this context are regressions - reproducible changes from version with-vs-without your patch

17:12 <gromero> fche: I see. thanks!

17:13 <fche> franks2, looking at your test case now

17:13 <fche> sorry for the delay

17:13 <fche> thanks for the good packaging of the info, it all looks like it should work (tm)

17:14 <fche> can try stap -t .... to let stap generate a probe hit-count apart from any printing in your handlers

17:14 <fche> can try attaching to the function__return probes via gdb (it speaks sys/sdt.h also)

17:14 <fche> (to see whether php is running across that stuff at all)

17:15 <fche> can try attaching to php in general with gdb and break in the vicinity of that function_return probe (by line number or such), see what's going on

17:30 <irker083> systemtap: smakarov systemtap.git:refs/heads/master * release-4.1-57-ge037dc7 / bpf-translate.cxx testsuite/systemtap.bpf/bpf_tests/foreach_string.stp: stapbpf pr23875 bugfix :: allocate actual keysize in foreach to avoid stack clobber http://tinyurl.com/y5n3ybew

17:30 <irker083> systemtap: smakarov systemtap.git:refs/heads/master * release-4.1-58-gebfc300 / testsuite/systemtap.bpf/bpf_tests/foreach.stp: stapbpf pr24811 test :: also test non-constant iteration limit in foreach.stp http://tinyurl.com/y5s7qzbq

18:02 <fche> franks2, sorry, haven't been able to reproduce the problem here just yet; I see the same sdt.h markers in my fedora30 php7.so tho

18:58 <gromero> fche: giveup_all(), which we use to flush fp regs live state to fp thread structures, was introduced in v4.10. So I would like to limit fp support to >= 4.10. What do you suggest for that?

19:04 <fche> couple of ways

19:04 <fche> in the embedded-c code, you hcan have an #if kernel-version kind of thing

19:04 <fche> in the tapset code, you can have a %( kernel_v %? %: %) kind of things

19:22 <gromero> fche: got it

22:18 <franks2> Fche: thanks for input. Will try tomorrow