Re: recomputation and PR 109154

[Andrew MacLeod <amacleod@redhat.com>]

I committed it.   ran it again for fun.  sigh.  Looks like its also 
triggering another issue now in g++.dg/warn/Wstringop-overflow-4.C
where its issuing:

b.C: In function ‘void test_strcpy_new_int16_t(size_t, const size_t*)’:
b.C:76:12: warning: ‘void* __builtin_memcpy(void*, const void*, long 
unsigned int)’ writing 3 bytes into a region of size 0 
[-Wstringop-overflow=]
    76 |     strcpy (d, s);                              \
       |     ~~~~~~~^~~~~~
b.C:90:3: note: in expansion of macro ‘T’
    90 |   T (S (2), new int16_t[r_imin_imax + 1]); // { dg-bogus "into 
a region of size" "pr106120" { xfail { ilp32 && c++98_only } } }


previously, by VRP2 time we hadn't figured out the edge case, couldn't 
recalculate _29 and iftmp.1_38 was unknown.

    iftmp.1_38 = _29 * 2;
   _40 = operator new [] (iftmp.1_38);
   __builtin_memcpy (_40, &MEM <const char[37]> [(void 
*)"0123456789abcdefghijklmnopqrstuvwxyz" + 34B], 3);

Now, by VRP2 we have figured it out...

     _40 = operator new [] (0);
     __builtin_memcpy (_40, &MEM <const char[37]> [(void 
*)"0123456789abcdefghijklmnopqrstuvwxyz" + 34B], 3);

And that info is available earlier for the warnings,  just hasn't  been 
explcitly exposed:

Do we want to change the cfail to always? something like:

diff --git a/gcc/testsuite/g++.dg/warn/Wstringop-overflow-4.C 
b/gcc/testsuite/g++.dg/warn/Wstringop-overflow-4.C
index 35fb59e0232..faad5bed074 100644
--- a/gcc/testsuite/g++.dg/warn/Wstringop-overflow-4.C
+++ b/gcc/testsuite/g++.dg/warn/Wstringop-overflow-4.C
@@ -141,7 +141,7 @@ void test_strcpy_new_int16_t (size_t n, const size_t 
vals[])

    int r_imin_imax = SR (INT_MIN, INT_MAX);
    T (S (1), new int16_t[r_imin_imax]);
-  T (S (2), new int16_t[r_imin_imax + 1]); // { dg-bogus "into a region 
of size" "pr106120" { xfail { ilp32 && c++98_only } } }
+  T (S (2), new int16_t[r_imin_imax + 1]); // { dg-bogus "into a region 
of size" "pr106120" { xfail { c++98_only } } }
    T (S (9), new int16_t[r_imin_imax * 2 + 1]);

    int r_0_imax = SR (0, INT_MAX);


Of course, I dont know what this is doing on other arches... perhaps 
wait for the fallout to be complete?

Andrew

On 3/30/23 12:05, Jakub Jelinek wrote:
> On Thu, Mar 30, 2023 at 11:58:19AM -0400, Andrew MacLeod wrote:
>> On 3/30/23 09:41, Jakub Jelinek wrote:
>>> On Wed, Mar 29, 2023 at 01:22:27PM -0400, Andrew MacLeod wrote:
>>>> however, as seems to be the case often, better ranges result in, I now get:
>>>>
>>>> FAIL: 23_containers/vector/bool/allocator/copy.cc (test for excess errors)
>>> Our middle-end warnings are just badly designed :(, the better value ranges
>>> are, the more false positives they have.
>>>
>>>> commit 358d0ca44faf2e20fbacd0f74386308b5ca52cd4
>>>> Author: Andrew MacLeod <amacleod@redhat.com>
>>>> Date:   Tue Mar 28 12:16:34 2023 -0400
>>>>
>>>>       Add recursive GORI recompuations with a depth limit.
>>> LGTM for trunk, let's do with the regression incrementally.
>>> Or as Richard mentioned on IRC, one possibility would be to force this
>>> param temporarily to 1 (or whatever matches previous behavior) for the
>>> diagnostic range queries).
>>>
>>> You need a ChangeLog entry though...
>>>
>> Attached. I also removed the bogus warning in Walloc-13.c that no longer
>> happens
>>
>> So incrementally deal with it.. what? just let it fail?
> For today?  Yes.
>
> Ok for trunk.
>
>> commit debb8ce1f9b9d5a72d88d0ae90a6b4da5130ff59
>> Author: Andrew MacLeod <amacleod@redhat.com>
>> Date:   Tue Mar 28 12:16:34 2023 -0400
>>
>>      Add recursive GORI recompuations with a depth limit.
>>      
>>              PR tree-optimization/109154
>>              gcc/
>>              * gimple-range-gori.cc (gori_compute::may_recompute_p): Add depth limit.
>>              * gimple-range-gori.h (may_recompute_p): Add depth param.
>>              * params.opt (ranger-recompute-depth): New param.
>>      
>>              gcc/testsuite/
>>              * gcc.dg/Walloca-13.c: Remove bogus warning that is now fixed.
> 	Jakub
>