Mailing-List: contact dev-help@stdcxx.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@stdcxx.apache.org
Received-SPF: pass (athena.apache.org: local policy)
Message-ID: <47F5650E.9000401@roguewave.com>
Date: Thu, 03 Apr 2008 17:15:26 -0600
From: Martin Sebor <sebor@roguewave.com>
Organization: Rogue Wave Software, Inc.
User-Agent: Thunderbird 2.0.0.12 (X11/20080226)
MIME-Version: 1.0
To: dev@stdcxx.apache.org
Subject: Re: [jira] Commented: (STDCXX-761) [HP aCC 6.16] Out of bound access
 in new.cpp
References: <1804206317.1207257804910.JavaMail.jira@brutus>
 <CFFDD219128FD94FB4F92B99F52D0A49010A36D0@exchmail01.Blue.Roguewave.Com>
In-Reply-To: 
 <CFFDD219128FD94FB4F92B99F52D0A49010A36D0@exchmail01.Blue.Roguewave.Com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Eric Lemings wrote:
>  
> Travis,
> 
> If you could, give the following patch a whirl (or quick review
> at least).

Is there an easier way to silence the warning than by adding
all these loops? E.g., by assigning the address of the first
member to a local pointer? (Or is the compiler too smart for
that?)

Also, there is no need to use the _RWSTD_SIZE_T macro in .cpp
files. The macro is useful in library and test suite headers
to reduce namespace pollution (so we don't have to #include
<stddef.h>.

Last thing: the code formatting convention calls for a space
before each open parenthesis. It might take some getting used
to but once you do, you'll never go back -- just ask Travis ;-)

Martin

> 
> -----
> $ svn diff
> Index: tests/src/new.cpp
> ===================================================================
> --- tests/src/new.cpp   (revision 644444)
> +++ tests/src/new.cpp   (working copy)
> @@ -604,6 +604,17 @@
>      return ret;
>  }
> 
> +static void
> +rwt_checkpoint_compare (_RWSTD_SIZE_T* diffs, _RWSTD_SIZE_T n,
> +                        const _RWSTD_SIZE_T* st0, const _RWSTD_SIZE_T*
> st1,
> +                        bool& diff_0)
> +{
> +    for (_RWSTD_SIZE_T i = 0; i != n; ++i) {
> +        diffs [i] = st1 [i] - st0 [i];
> +        if (diffs [i])
> +            diff_0 = false;
> +    }
> +}
> 
>  _TEST_EXPORT rwt_free_store*
>  rwt_checkpoint (const rwt_free_store *st0, const rwt_free_store *st1)
> @@ -616,21 +627,24 @@
>          // of the free_store specified by the arguments
> 
>          static rwt_free_store diff;
> -
>          memset (&diff, 0, sizeof diff);
> 
> -        size_t*       diffs    = diff.new_calls_;
> -        const size_t* st0_args = st0->new_calls_;
> -        const size_t* st1_args = st1->new_calls_;
> -
>          bool diff_0 = true;   // difference of 0 (i.e., none)
> 
> -        for (size_t i = 0; i != 16; ++i) {
> -            diffs [i] = st1_args [i] - st0_args [i];
> -            if (diffs [i])
> -                diff_0 = false;
> -        }
> +#define EXTENT(array) (sizeof (array) / sizeof(*array))
> +#define COMPARE(member) \
> +        rwt_checkpoint_compare (diff.member, EXTENT(diff.member),\
> +                                st0->member, st1->member, diff_0)
> 
> +        COMPARE(new_calls_);
> +        COMPARE(delete_calls_);
> +        COMPARE(delete_0_calls_);
> +        COMPARE(blocks_);
> +        COMPARE(bytes_);
> +        COMPARE(max_blocks_);
> +        COMPARE(max_bytes_);
> +        COMPARE(max_block_size_);
> +
>          if (diff_0)
>              return 0;
> -----
> 
> The 0.new test seems to be okay with it.
> 
> Thanks,
> Brad.
> 
>> -----Original Message-----
>> From: Travis Vitek (JIRA) [mailto:jira@apache.org] 
>> Sent: Thursday, April 03, 2008 3:23 PM
>> To: Eric Lemings
>> Subject: [jira] Commented: (STDCXX-761) [HP aCC 6.16] Out of 
>> bound access in new.cpp
>>
>>
>>     [ 
>> https://issues.apache.org/jira/browse/STDCXX-761?page=com.atla
>> ssian.jira.plugin.system.issuetabpanels:comment-tabpanel&focus
>> edCommentId=12585290#action_12585290 ] 
>>
>> Travis Vitek commented on STDCXX-761:
>> -------------------------------------
>>
>> It looks like someone is being tricky. They are assuming that 
>> every member in {{rwt_free_store}} is contiguous [i.e. no 
>> padding], and they are actually comparing each of 
>> {{new_calls_}}, {{delete_calls_}}, {{blocks_}}, {{bytes_}}, 
>> {{max_blocks_}}, {{max_bytes_}} and {{max_block_size_}} with 
>> that one loop. If you decide to fix this, you'll end up 
>> probably end up writing eight loops of two iterations each.
>>
>> Also, I'd prefer that the 2 isn't hardcoded and that you use 
>> something like {{sizeof (array) / sizeof (*array)}}, or a 
>> size macro instead.
>>
>>> [HP aCC 6.16] Out of bound access in new.cpp
>>> --------------------------------------------
>>>
>>>                 Key: STDCXX-761
>>>                 URL: 
>> https://issues.apache.org/jira/browse/STDCXX-761
>>>             Project: C++ Standard Library
>>>          Issue Type: Sub-task
>>>          Components: Test Driver
>>>    Affects Versions: 4.2.0
>>>         Environment: $ uname -sr && aCC -V
>>> HP-UX B.11.31
>>> aCC: HP C/aC++ B3910B A.06.16 [Nov 26 2007]
>>>            Reporter: Scott (Yu) Zhong
>>>            Assignee: Eric Lemings
>>>             Fix For: 4.2.1
>>>
>>>   Original Estimate: 2h
>>>  Remaining Estimate: 2h
>>>
>>> When compiled with HP aCC 6.16 +w +O, the test driver 
>> source file 
>> [new.cpp|http://svn.apache.org/repos/asf/stdcxx/trunk/tests/sr
>> c/new.cpp] produces the warnings below:
>>> {noformat}
>>> aCC -c     -I$(TOPDIR)/include -I$(BUILDDIR)/include 
>> -I$(TOPDIR)/tests/include -AA  +O2  +DD64 +w \
>>>     +W392 +W655 +W684 +W818 +W819 +W849 +W2193 +W2236 
>> +W2261 +W2340 +W2401 +W2487 +W4227 \
>>>     +W4229 +W4231 +W4235 +W4237 +W4249 +W4255 +W4272 +W4284 
>> +W4285 +W4286 +W4296 +W4297 +W3348 \
>>>     $(TOPDIR)/tests/src/new.cpp
>>> "$(TOPDIR)/tests/src/new.cpp", line 629, procedure 
>> rwt_checkpoint: warning #20206-D: Out of bound access (In 
>> expression "(unsigned long long*)(&diff)->new_calls_+i", 
>> (&diff)->new_calls_ (type: unsigned long long [2]) has byte 
>> range [0 .. 15], writing byte range [0 .. 127].)
>>> {noformat}
>> -- 
>> This message is automatically generated by JIRA.
>> -
>> You can reply to this email to add a comment to the issue online.
>>
>>