[review,v3] testsuite, cp: increase the coverage of testing pass-by-ref arguments

  On Monday, January 13, 2020 9:21 PM, Aktemur, Tankut Baris wrote:
> 
> On Monday, January 13, 2020 8:38 PM, Luis Machado wrote:
> >
> > On 1/13/20 4:35 PM, Aktemur, Tankut Baris wrote:
> > > On Monday, January 13, 2020 7:21 PM, Luis Machado wrote:
> > >>
> > >> Hi,
> > >>
> > >> I noticed these new tests fail (in large numbers) for GCC 5.4.x. Is it
> > >> expected? If so, it may be worth making this test conditional on newer
> > >> GCC versions.
> > >>
> > >
> > > Yes, this is expected. Older GCC versions did not emit certain DWARF attributes
> > > (DW_AT_deleted, DW_AT_defaulted).  This prevents GDB from making the right
> > > pass-by-reference decision.  I'll submit a patch for this.
> >
> > Thanks for clarifying this.
> >
> > I can submit that patch if you like. I have a box running an older GCC,
> > so i noticed that and thought i'd check.
> >
> > Luis
> 
> Yes, sure.
> 
> Thank you.
> 
> -Baris

I've investigated GCC and Clang for this.  GCC started emitting DW_AT_deleted and
DW_AT_defaulted with version 7.  Clang does not emit these attributes; however, it
has been emitting DW_AT_calling_convention starting with version 7.  This
attribute helps the debugger make the right decision in some cases.

Based on this, I think the test cases have to be filtered in a somewhat
fine-granular manner.  Therefore I thought I could save you from the burden of
having to go through the code-generating test definition.  Below is a patch proposal.

-Baris



From 86017121c8f1c7d0f0020c7a8da22dba64dad3fd Mon Sep 17 00:00:00 2001
From: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Date: Tue, 14 Jan 2020 13:36:09 +0100
Subject: [PATCH] testsuite, cp: add pass-by-ref test expected failures for
 certain compilers

There exist expected failures in the pass-by-ref.exp and
pass-by-ref-2.exp tests based on the GCC and Clang version.

* GCC version <= 6 and Clang do not emit DW_AT_deleted and
  DW_AT_defaulted.

* Clang version >= 7 emits DW_AT_calling_convention, which helps the
  debugger make the right calling convention decision in some cases
  despite lacking the 'defaulted' and 'deleted' attributes.

Mark the related tests as XFAIL based on the compiler version.

Tested on X86_64 using GCC 5.5.0, 6.5.0, 7.4.0, 8.3.0, 9.2.1;
and Clang 5.0.1, 6.0.0, 7.0.0, 8.0.0, 9.0.1, 10.0.0.

gdb/testsuite/ChangeLog:
2020-01-14  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>

	* gdb.cp/pass-by-ref.exp: Mark some tests as XFAIL based on the
	GCC/Clang version.
	* gdb.cp/pass-by-ref-2.exp: Ditto.

Change-Id: I1d8440aa438049f7c4da7f4f76f201c48550f1e4

---
 gdb/testsuite/gdb.cp/pass-by-ref-2.exp |  6 ++++++
 gdb/testsuite/gdb.cp/pass-by-ref.exp   | 26 ++++++++++++++++++++++++++
 2 files changed, 32 insertions(+)

-- 
2.17.1


Intel Deutschland GmbH
Registered Address: Am Campeon 10-12, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de
Managing Directors: Christin Eisenschmid, Gary Kershaw
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928
  

On 1/14/20 9:52 AM, Aktemur, Tankut Baris wrote:
> On Monday, January 13, 2020 9:21 PM, Aktemur, Tankut Baris wrote:
>>
>> On Monday, January 13, 2020 8:38 PM, Luis Machado wrote:
>>>
>>> On 1/13/20 4:35 PM, Aktemur, Tankut Baris wrote:
>>>> On Monday, January 13, 2020 7:21 PM, Luis Machado wrote:
>>>>>
>>>>> Hi,
>>>>>
>>>>> I noticed these new tests fail (in large numbers) for GCC 5.4.x. Is it
>>>>> expected? If so, it may be worth making this test conditional on newer
>>>>> GCC versions.
>>>>>
>>>>
>>>> Yes, this is expected. Older GCC versions did not emit certain DWARF attributes
>>>> (DW_AT_deleted, DW_AT_defaulted).  This prevents GDB from making the right
>>>> pass-by-reference decision.  I'll submit a patch for this.
>>>
>>> Thanks for clarifying this.
>>>
>>> I can submit that patch if you like. I have a box running an older GCC,
>>> so i noticed that and thought i'd check.
>>>
>>> Luis
>>
>> Yes, sure.
>>
>> Thank you.
>>
>> -Baris
> 
> I've investigated GCC and Clang for this.  GCC started emitting DW_AT_deleted and
> DW_AT_defaulted with version 7.  Clang does not emit these attributes; however, it
> has been emitting DW_AT_calling_convention starting with version 7.  This
> attribute helps the debugger make the right decision in some cases.
> 
> Based on this, I think the test cases have to be filtered in a somewhat
> fine-granular manner.  Therefore I thought I could save you from the burden of
> having to go through the code-generating test definition.  Below is a patch proposal.
> 
> -Baris

Thanks! I've checked this on my box with an older GCC and i see the 
XFAIL's now. So it looks good to me.

Small nit below...


> 
> 
> 
>  From 86017121c8f1c7d0f0020c7a8da22dba64dad3fd Mon Sep 17 00:00:00 2001
> From: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
> Date: Tue, 14 Jan 2020 13:36:09 +0100
> Subject: [PATCH] testsuite, cp: add pass-by-ref test expected failures for
>   certain compilers
> 
> There exist expected failures in the pass-by-ref.exp and
> pass-by-ref-2.exp tests based on the GCC and Clang version.
> 
> * GCC version <= 6 and Clang do not emit DW_AT_deleted and
>    DW_AT_defaulted.
> 
> * Clang version >= 7 emits DW_AT_calling_convention, which helps the
>    debugger make the right calling convention decision in some cases
>    despite lacking the 'defaulted' and 'deleted' attributes.
> 
> Mark the related tests as XFAIL based on the compiler version.
> 
> Tested on X86_64 using GCC 5.5.0, 6.5.0, 7.4.0, 8.3.0, 9.2.1;
> and Clang 5.0.1, 6.0.0, 7.0.0, 8.0.0, 9.0.1, 10.0.0.
> 
> gdb/testsuite/ChangeLog:
> 2020-01-14  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>
> 
> 	* gdb.cp/pass-by-ref.exp: Mark some tests as XFAIL based on the
> 	GCC/Clang version.
> 	* gdb.cp/pass-by-ref-2.exp: Ditto.
> 
> Change-Id: I1d8440aa438049f7c4da7f4f76f201c48550f1e4
> 
> ---
>   gdb/testsuite/gdb.cp/pass-by-ref-2.exp |  6 ++++++
>   gdb/testsuite/gdb.cp/pass-by-ref.exp   | 26 ++++++++++++++++++++++++++
>   2 files changed, 32 insertions(+)
> 
> diff --git a/gdb/testsuite/gdb.cp/pass-by-ref-2.exp b/gdb/testsuite/gdb.cp/pass-by-ref-2.exp
> index a83ce8d5d7d..e7b04771eb2 100644
> --- a/gdb/testsuite/gdb.cp/pass-by-ref-2.exp
> +++ b/gdb/testsuite/gdb.cp/pass-by-ref-2.exp
> @@ -42,6 +42,10 @@ if {![runto_main]} {
>       return -1
>   }
>   
> +# GCC version <= 6 and Clang do not emit DW_AT_defaulted and DW_AT_deleted
> +set is_gcc_6_or_older [test_compiler_info {gcc-[0-6]-*}]
> +set is_clang [test_compiler_info {clang-*}]
> +
>   set bp_location [gdb_get_line_number "stop here"]
>   gdb_breakpoint $bp_location
>   gdb_continue_to_breakpoint "end of main" ".*return .*;"

It seems to be a mixed bag, but i see more examples of having a period 
after the sentence than not having it. Multiple cases of this on the patch.

> @@ -65,6 +69,7 @@ set sig "\"Inlined\:\:Inlined\\(.*Inlined const\&\\)\""
>   gdb_test "print cbvInlined (inlined)" \
>       "expression cannot be evaluated .* \\(maybe inlined\\?\\)"
>   
> +if {$is_gcc_6_or_older || $is_clang} {setup_xfail "*-*-*"}
>   gdb_test "print cbvDtorDel (*dtorDel)" \
>       ".* cannot be evaluated .* 'DtorDel' is not destructible" \
>       "type not destructible"
> @@ -94,6 +99,7 @@ gdb_test "print cbvTwoMCtor (twoMctor)" \
>       ".* cannot be evaluated .* 'TwoMCtor' is not copy constructible" \
>       "copy ctor is implicitly deleted"
>   
> +if {$is_gcc_6_or_older || $is_clang} {setup_xfail "*-*-*"}
>   gdb_test "print cbvTwoMCtorAndCCtor (twoMctorAndCctor)" "12" \
>       "call cbvTwoMCtorAndCCtor"
>   gdb_test "print twoMctorAndCctor.x" "2" \
> diff --git a/gdb/testsuite/gdb.cp/pass-by-ref.exp b/gdb/testsuite/gdb.cp/pass-by-ref.exp
> index f500710fd43..aff9f9a3c1e 100644
> --- a/gdb/testsuite/gdb.cp/pass-by-ref.exp
> +++ b/gdb/testsuite/gdb.cp/pass-by-ref.exp
> @@ -369,6 +369,12 @@ proc test_for_class { prefix states cbvfun data_field length} {
>       global ADDED
>       global TRACE
>   
> +    # GCC version <= 6 and Clang do not emit DW_AT_defaulted and DW_AT_deleted
> +    set is_gcc_6_or_older [test_compiler_info {gcc-[0-6]-*}]
> +    set is_clang [test_compiler_info {clang-*}]
> +    # But Clang version >= 7 emits DW_AT_calling_convention for types
> +    set is_clang_6_or_older [test_compiler_info {clang-[0-6]-*}]
> +
>       with_test_prefix $name {
>   	if {[is_copy_constructible $states]} {
>   	    set expected [expr {$ORIGINAL + $ADDED}]
> @@ -378,6 +384,19 @@ proc test_for_class { prefix states cbvfun data_field length} {
>   	    if {$dtor == "explicit"} {
>   		gdb_test "print tracer = 0" " = 0" "reset the tracer"
>   	    }
> +
> +	    if {$cctor == "defaultedIn" || $dtor == "defaultedIn"} {
> +		if {$is_gcc_6_or_older || $is_clang_6_or_older} {
> +		    setup_xfail "*-*-*"
> +		} elseif {$is_clang} {
> +		    # If this is a pass-by-value case, Clang >= 7's
> +		    # DW_AT_calling_convention leads to the right decision.
> +		    # Otherwise, it is expected to fail.
> +		    if {"defaultedOut" in $states || "explicit" in $states} {
> +			setup_xfail "*-*-*"
> +		    }
> +		}
> +	    }
>   	    gdb_test "print ${cbvfun}<$name> (${name}_var)" " = $expected" \
>   		"call '$cbvfun'"
>   	    gdb_test "print ${name}_var.${data_field}\[0\]" " = $ORIGINAL" \
> @@ -389,10 +408,17 @@ proc test_for_class { prefix states cbvfun data_field length} {
>   		    "cbv argument should not change (item $last_index)"
>   	    }
>   	    if {$dtor == "explicit"} {
> +		if {$cctor == "defaultedIn"
> +		    && ($is_gcc_6_or_older || $is_clang)} {
> +		    setup_xfail "*-*-*"
> +		}
>   		gdb_test "print tracer" " = $TRACE" \
>   		    "destructor should be called"
>   	    }
>   	} else {
> +	    if {$cctor == "deleted" && ($is_gcc_6_or_older || $is_clang)} {
> +		setup_xfail "*-*-*"
> +	    }
>   	    gdb_test "print ${cbvfun}<$name> (${name}_var)" \
>   		".* cannot be evaluated .* '${name}' is not copy constructible" \
>   		"calling '$cbvfun' should be refused"
>
  

On Tuesday, January 14, 2020 2:13 PM, Luis Machado wrote:
> 

> On 1/14/20 9:52 AM, Aktemur, Tankut Baris wrote:

> >

> > I've investigated GCC and Clang for this.  GCC started emitting DW_AT_deleted and

> > DW_AT_defaulted with version 7.  Clang does not emit these attributes; however, it

> > has been emitting DW_AT_calling_convention starting with version 7.  This

> > attribute helps the debugger make the right decision in some cases.

> >

> > Based on this, I think the test cases have to be filtered in a somewhat

> > fine-granular manner.  Therefore I thought I could save you from the burden of

> > having to go through the code-generating test definition.  Below is a patch proposal.

> >

> > -Baris

> 

> Thanks! I've checked this on my box with an older GCC and i see the

> XFAIL's now. So it looks good to me.

> 

> Small nit below...

> 


Thank you.  I'll fix that and wait for an official approval.

-Baris

> > +# GCC version <= 6 and Clang do not emit DW_AT_defaulted and DW_AT_deleted

> > +set is_gcc_6_or_older [test_compiler_info {gcc-[0-6]-*}]

> > +set is_clang [test_compiler_info {clang-*}]

> > +

> >   set bp_location [gdb_get_line_number "stop here"]

> >   gdb_breakpoint $bp_location

> >   gdb_continue_to_breakpoint "end of main" ".*return .*;"

> 

> It seems to be a mixed bag, but i see more examples of having a period

> after the sentence than not having it. Multiple cases of this on the patch.

> 

Intel Deutschland GmbH
Registered Address: Am Campeon 10-12, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de
Managing Directors: Christin Eisenschmid, Gary Kershaw
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928