[PATCH, testsuite] Fix usage of sizeof in testsuite/g++.dg/cpp0x/rv[1..8]p.C

[PATCH, testsuite] Fix usage of sizeof in testsuite/g++.dg/cpp0x/rv[1..8]p.C

[Uros Bizjak]

[-- Attachment #1: Type: text/plain, Size: 1364 bytes --]

On Fri, Aug 13, 2010 at 6:15 AM, Ian Lance Taylor <iant@google.com> wrote:

>> A problem arises with the code in testsuite/g++.dg/cpp0x/rv[1..8]p.C.
>> These tests use "sizeof(..character array...) == <n>", but sizeof char
>> array depends heavily on the value of #define STRUCTURE_SIZE_BOUNDARY.
>> Targets that define this value to i.e. 32 (for performance reasons,
>> instead of default BITS_PER_UNIT) will fail all these checks.
>>
>> Would it be acceptable to change all these checks from
>>
>>     "sa<sizeof(foo(a)) == 1> t1;"
>>
>> to
>>
>>     "sa<sizeof(foo(a)) != 0> t1;" ?
>
> Sounds fine to me.

Attached is the mechanical patch that implements proposed change.

2010-08-13  Uros Bizjak  <ubizjak@gmail.com>

	* g++.dg/cast.C: Change "sizeof(...) = <n>" checks
	to "sizeof(...) != 0" to pass on targets with
	STRUCTURE_SIZE_BOUNDARY != BITS_PER_UNIT.
	* g++.dg/cpp0x/iop.C: Ditto.
	* g++.dg/cpp0x/named_refs.C: Ditto.
	* g++.dg/cpp0x/rv1p.C: Ditto.
	* g++.dg/cpp0x/rv2p.C: Ditto.
	* g++.dg/cpp0x/rv3p.C: Ditto.
	* g++.dg/cpp0x/rv4p.C: Ditto.
	* g++.dg/cpp0x/rv5p.C: Ditto.
	* g++.dg/cpp0x/rv6p.C: Ditto.
	* g++.dg/cpp0x/rv7p.C: Ditto.
	* g++.dg/cpp0x/rv8p.C: Ditto.
	* g++.dg/cpp0x/unnamed_refs.C: Ditto.

Tested on x86_64-pc-linux-gnu {,-m32} and my private target.

OK for 4.4, 4.5 and SVN mainline?

Uros.

[-- Attachment #2: t.diff.txt.bz2 --]
[-- Type: application/x-bzip2, Size: 7946 bytes --]

Re: [PATCH, testsuite] Fix usage of sizeof in testsuite/g++.dg/cpp0x/rv[1..8]p.C

[Ian Lance Taylor]

Uros Bizjak <ubizjak@gmail.com> writes:

> On Fri, Aug 13, 2010 at 6:15 AM, Ian Lance Taylor <iant@google.com> wrote:
>
>>> A problem arises with the code in testsuite/g++.dg/cpp0x/rv[1..8]p.C.
>>> These tests use "sizeof(..character array...) == <n>", but sizeof char
>>> array depends heavily on the value of #define STRUCTURE_SIZE_BOUNDARY.
>>> Targets that define this value to i.e. 32 (for performance reasons,
>>> instead of default BITS_PER_UNIT) will fail all these checks.
>>>
>>> Would it be acceptable to change all these checks from
>>>
>>>     "sa<sizeof(foo(a)) == 1> t1;"
>>>
>>> to
>>>
>>>     "sa<sizeof(foo(a)) != 0> t1;" ?
>>
>> Sounds fine to me.
>
> Attached is the mechanical patch that implements proposed change.
>
> 2010-08-13  Uros Bizjak  <ubizjak@gmail.com>
>
> 	* g++.dg/cast.C: Change "sizeof(...) = <n>" checks
> 	to "sizeof(...) != 0" to pass on targets with
> 	STRUCTURE_SIZE_BOUNDARY != BITS_PER_UNIT.
> 	* g++.dg/cpp0x/iop.C: Ditto.
> 	* g++.dg/cpp0x/named_refs.C: Ditto.
> 	* g++.dg/cpp0x/rv1p.C: Ditto.
> 	* g++.dg/cpp0x/rv2p.C: Ditto.
> 	* g++.dg/cpp0x/rv3p.C: Ditto.
> 	* g++.dg/cpp0x/rv4p.C: Ditto.
> 	* g++.dg/cpp0x/rv5p.C: Ditto.
> 	* g++.dg/cpp0x/rv6p.C: Ditto.
> 	* g++.dg/cpp0x/rv7p.C: Ditto.
> 	* g++.dg/cpp0x/rv8p.C: Ditto.
> 	* g++.dg/cpp0x/unnamed_refs.C: Ditto.
>
> Tested on x86_64-pc-linux-gnu {,-m32} and my private target.
>
> OK for 4.4, 4.5 and SVN mainline?

Hmmm, actually, looking at the patch, I'm not sure this is the best
approach.  The problem is that your patch means that some of the tests
no longer actually test anything.  E.g., in rv2p.C, we have this:

    sa<sizeof(sink_2_13(a))           == 1> t1;
    sa<sizeof(sink_2_13(va))          == 3> t3;

The point of this test is to check function overloading.  Change both
instances to != 0 means that the test is no longer testing anything.

So a better approach here is going to be to change the fields from,
e.g., "char" to "long", and change the tests to be N * sizeof(long).

Sorry about that.

Ian

Re: [PATCH, testsuite] Fix usage of sizeof in testsuite/g++.dg/cpp0x/rv[1..8]p.C

[Uros Bizjak]

[-- Attachment #1: Type: text/plain, Size: 1390 bytes --]

On Fri, Aug 13, 2010 at 10:43 PM, Ian Lance Taylor <iant@google.com> wrote:

> Hmmm, actually, looking at the patch, I'm not sure this is the best
> approach.  The problem is that your patch means that some of the tests
> no longer actually test anything.  E.g., in rv2p.C, we have this:
>
>    sa<sizeof(sink_2_13(a))           == 1> t1;
>    sa<sizeof(sink_2_13(va))          == 3> t3;
>
> The point of this test is to check function overloading.  Change both
> instances to != 0 means that the test is no longer testing anything.
>
> So a better approach here is going to be to change the fields from,
> e.g., "char" to "long", and change the tests to be N * sizeof(long).

Thanks, attached is fixed patch that implements your suggestion.

2010-08-14  Uros Bizjak  <ubizjak@gmail.com>

	* g++.dg/cast.C: Change fields of structures to "long" to pass
	the test on targets with STRUCTURE_SIZE_BOUNDARY != BITS_PER_UNIT.
	* g++.dg/cpp0x/iop.C: Ditto.
	* g++.dg/cpp0x/named_refs.C: Ditto.
	* g++.dg/cpp0x/rv1p.C: Ditto.
	* g++.dg/cpp0x/rv2p.C: Ditto.
	* g++.dg/cpp0x/rv3p.C: Ditto.
	* g++.dg/cpp0x/rv4p.C: Ditto.
	* g++.dg/cpp0x/rv5p.C: Ditto.
	* g++.dg/cpp0x/rv6p.C: Ditto.
	* g++.dg/cpp0x/rv7p.C: Ditto.
	* g++.dg/cpp0x/rv8p.C: Ditto.
	* g++.dg/cpp0x/unnamed_refs.C: Ditto.

Patch was tested on x86_64-pc-linux-gnu and committed to mainline SVN.

Uros.

[-- Attachment #2: t.diff.txt.bz2 --]
[-- Type: application/x-bzip2, Size: 9331 bytes --]