18:14 <agentzh> fche: is it something doable? https://sourceware.org/bugzilla/show_bug.cgi?id=23506

18:15 <fche> yeah; try with stap -k -p4 and check out what it already collects

18:15 <fche> though are you sure you want @var() as opposed to & @var() ?

18:16 <fche> @var returns values, not addresses

18:18 <fche> ah I see, the c program accounts for that

18:25 <fche> agentzh, so yeah it should work

18:51 <agentzh> fche: yeah, stap already resolves function symbols. i guess we would have to encode the global symbol addresses into the kernel module in a similar way?

18:52 <agentzh> fche: btw, i've added much more detailed info to this CPU stuck during test suite run PR: https://sourceware.org/bugzilla/show_bug.cgi?id=23493

18:52 <agentzh> i attached the raw output files to avoid any confusions.

18:52 <agentzh> please let me know if you also need all those systemtap.log files generated by that test run.

18:52 <fche> yup thanks.

18:53 <fche> re. function symbols, it should not be a big deal to include data symbols too ... they're coming from the same source

18:55 <agentzh> fche: yeah, i was just a bit worried about the kernel module size ;)

18:56 <fche> yeah, it's not that bad

18:56 <agentzh> okay

18:56 <fche> probably translate.cxx line 6848 STT_FUNC area ... that may be where the filtering is happening

18:57 <agentzh> fche: i wonder if we could have a way to disable the add_global_var_display() thing. according to the source, there is no way to disable that for the kernel/dyninst runtimes (it's not supported by bpf runtime anyway).

18:57 <agentzh> fche: re STT_FUNC: thanks for the hint!

18:58 <agentzh> we'll have a look.

19:02 <agentzh> fche: created a PR for the --no-global-var-display thing: https://sourceware.org/bugzilla/show_bug.cgi?id=23507

19:04 <fche> sure. or .... your generator can emit a probe never { } with some reads of the globals

19:05 <agentzh> fche: i tried the fake reads too, but it is hard to suppress those "side-effect free" warnings from the translator for those fake reads and fake probes.

19:06 <agentzh> so not ideal.

19:06 <fche> probe never { print(global) }

19:06 <fche> won't have warnings there

19:06 <agentzh> ah, okay, print() never runs, so it does the trick :)

19:06 <agentzh> thanks for the tip

19:06 <fche> righto