Re: [PATCH 6/8] DWARF-5: call sites

[Jan Kratochvil <jan.kratochvil@redhat.com>]

On Fri, 17 Feb 2017 12:57:42 +0100, Pedro Alves wrote:
> On 02/12/2017 08:23 PM, Jan Kratochvil wrote:
> > --- a/gdb/testsuite/gdb.arch/amd64-entry-value-param.exp
> > +++ b/gdb/testsuite/gdb.arch/amd64-entry-value-param.exp
> > @@ -19,7 +19,7 @@ set opts {}
> >  if [info exists COMPILE] {
> >      # make check RUNTESTFLAGS="gdb.arch/amd64-entry-value-param.exp COMPILE=1"
> >      set srcfile ${srcfile2}
> > -    lappend opts debug optimize=-O2
> > +    lappend opts optimize=-O2 additional_flags=-gdwarf-5
> 
> Did you mean to update the .S file?
> 
> I wonder whether it makes sense to run the test twice, once against
> the current .S file using the GNU version of the opcodes, and
> another against standard DWARF5 opcodes.

So I have added: gdb.arch/amd64-entry-value-param-dwarf5.exp

It is not perfect as gdb.arch/amd64-entry-value-param-dwarf5.c there is the
same as gdb.arch/amd64-entry-value-param.c .


> That raises the question
> of why we don't do that with the c based tests, where we'll handle
> whatever format the compiler outputs.  So maybe not go there...

We do that, unaware if Sergio does that with buildbot but at least I do that
with my GDB tester which I updated for DWARF-5:
	http://git.jankratochvil.net/?p=nethome.git;a=blob_plain;f=bin/hammock
	!defined $dwarf or ($dwarf>=2 && $dwarf<=5) or die "--dwarf requires DWARF version number";

Many years ago I was running the nightly testsuite runs for all DWARF
versions.

Nowadays I have at least regression tested this DWARF-5 patchset so that
-gdwarf-4 and -gdwarf-5 do produce the same testsuite results.  It has caught
some bugs in my patchset.


Jan