Re: [PATCH] PR25756 more debuginfod metrics

["Frank Ch. Eigler" <fche@redhat.com>]

Hi -

> I found this somewhat hard to review since the indenting isn't
> following the (GNU) style used in the rest of the file. Please place
> curly brackets on their own line when possible.

OK.

> > +2020-10-20  Frank Ch. Eigler  <fche@redhat.com>
> > +
> > +	PR26756: more prometheus metrics
> > +	* debuginfod.cxx (*_exception): Add counters for error occurrences.
> > +	(fdcache::*): Add counters for fdcache operations and status.
> 
> Also a new public set_metrics() function.

OK.

> In general these changes look correct.
> Unfortunately the test crashes on my RHEL7 with DTS9 system:
> [...]
> #11 0x00007f924ecb7ce9 in __run_exit_handlers (status=0, 
>     listp=0x7f924f0456c8 <__exit_funcs>, 
>     run_list_atexit=run_list_atexit@entry=true) at exit.c:77
> #12 0x00007f924ecb7d37 in __GI_exit (status=<optimized out>) at exit.c:99
> #13 0x00007f924eca055c in __libc_start_main (
>     main=0x40ad40 <main(int, char**)>, argc=10, argv=0x7ffc8ebaf6a8, 
>     init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, 
>     stack_end=0x7ffc8ebaf698) at ../csu/libc-start.c:300
> #14 0x000000000040c3e5 in _start ()
>     at /opt/rh/devtoolset-9/root/usr/include/c++/9/bits/stl_tree.h:211
> Which seems to be caused by the libarchive_fdcache destructor calling
> limit (0,0) triggering the new inc_metric which somehow cannot use the
> metrics map?

Ah it's a global variable destructor sequencing issue.
Hmm what a pain, ok will think of a solution.


> > +# create a 000 empty .rpm file to evoke a metric-visible error
> > +touch R/nothing.rpm
> > +chmod 000 R/nothing.rpm
> > +
> 
> Does this cause any issues during cleanup?
> I would have expected something like:
> 
> if [ -f R/nothing.rpm ]; then chmod 644 R/nothing.rpm; fi
> 
> in cleanup(). But maybe the rm -f takes care of that?

There's a "rm -rf R" in cleanup, so is covered.


- FChE