Bug #42965 isinf() on 32bit x86 with gcc 4.3 can produce incorrect results for ROUND()

Description:
In Drizzle, we were hit directly by this GCC bug (that I found):
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39228

Theoretically MySQL could be hit by the same problem, depending on how the code is compiled (and with what compiler).

How to repeat:
see gcc bug

Suggested fix:
The proposed patch for Drizzle is below (will add a comment about the gcc bug before committing. will paste final commit here). It would be in sql/item_func.cc in mysql

=== modified file 'drizzled/function/math/round.cc'
--- drizzled/function/math/round.cc	2008-12-16 06:21:46 +0000
+++ drizzled/function/math/round.cc	2009-02-18 09:59:40 +0000
@@ -111,14 +111,22 @@
     tmp2 is here to avoid return the value with 80 bit precision
     This will fix that the test round(0.1,1) = round(0.1,1) is true
   */
-  volatile double tmp2;
+  double tmp2;
 
   tmp=(abs_dec < array_elements(log_10) ?
        log_10[abs_dec] : pow(10.0,(double) abs_dec));
 
+  double value_times_tmp= value * tmp;
+
+  if(sizeof(double) < sizeof(double_t))
+  {
+    volatile double t= value_times_tmp;
+    value_times_tmp= t;
+  }
+
   if (dec_negative && isinf(tmp))
     tmp2= 0;
-  else if (!dec_negative && isinf(value * tmp))
+  else if (!dec_negative && isinf(value_times_tmp))
     tmp2= value;
   else if (truncate)
   {

No code in MySQL 5.0/5.1/6.0 is affected by the mentioned GCC bug, since all isinf() calls either are in C++ code, or do not use intermediate values as the argument.

But just to be on the safe side, I will add a configure check and a workaround in case isinf() is unreliable.

note that the drizzle code affected was being built by g++

Stewart, how could drizzle be affected by the GCC bug then, if g++ transforms isinf() to the __isinf() call which always produces correct results?

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/70137

2741 Alexey Kopytov	2009-03-24
      Fix for bug #42965: isinf() on 32bit x86 with gcc 4.3 can    
                          produce incorrect results for ROUND() 
      
      Added a workaround and a configure check to test whether 
      isinf() is affected by the GCC bug #39228. 
             
      Since no code in MySQL server is currently affected by that 
      bug, the patch is actually a safeguard for possible future 
      code modifications. No test cases or changelog entries are 
      needed. 
      modified:
        configure.in
        include/my_global.h

i did look at a later tree than when i originally worked on the bug, so perhaps we changed what we were building the file with.

you can always check the compiled code.

Pushed into 5.0.82 (revid:davi.arnaut@sun.com-20090505184158-dvmedh8n472y8np5) (version source revid:davi.arnaut@sun.com-20090505184158-dvmedh8n472y8np5) (merge vers: 5.0.82) (pib:6)

Pushed into 5.1.35 (revid:davi.arnaut@sun.com-20090505190206-9xmh7dlc6kom8exp) (version source revid:davi.arnaut@sun.com-20090505190206-9xmh7dlc6kom8exp) (merge vers: 5.1.35) (pib:6)

Pushed into 6.0.12-alpha (revid:svoj@sun.com-20090506125450-yokcmvqf2g7jhujq) (version source revid:horst@mysql.com-20090327184517-25eq077q2beocs6y) (merge vers: 6.0.11-alpha) (pib:6)

No user-visible effects. No changelog entry needed.

Pushed into 5.1.35-ndb-6.3.26 (revid:jonas@mysql.com-20090615074202-0r5r2jmi83tww6sf) (version source revid:jonas@mysql.com-20090615070837-9pccutgc7repvb4d) (merge vers: 5.1.35-ndb-6.3.26) (pib:6)

Pushed into 5.1.35-ndb-7.0.7 (revid:jonas@mysql.com-20090615074335-9hcltksp5cu5fucn) (version source revid:jonas@mysql.com-20090615072714-rmfkvrbbipd9r32c) (merge vers: 5.1.35-ndb-7.0.7) (pib:6)

Pushed into 5.1.35-ndb-6.2.19 (revid:jonas@mysql.com-20090615061520-sq7ds4yw299ggugm) (version source revid:jonas@mysql.com-20090615054654-ebgpz7elwu1xj36j) (merge vers: 5.1.35-ndb-6.2.19) (pib:6)