MySQL Bugs: #33559: Again: "select hex(ascii(a)) ... order by a" results in different order

Bug #33559	Again: "select hex(ascii(a)) ... order by a" results in different order
Submitted:	28 Dec 2007 18:25	Modified:	18 Jan 2008 20:37
Reporter:	Joerg Bruehe	Email Updates:
Status:	Duplicate	Impact on me:	None
Category:	Tests: Engine	Severity:	S3 (Non-critical)
Version:	6.0.4-alpha	OS:	Any
Assigned to:	Assigned Account	CPU Architecture:	Any

Description:
This is similar to bug#31556, but on a much larger scale.
I have (till now) only found it in 6.0, cannot (yet) comment on 5.1 or below.

Similar differences in the "charset" tests for all table handlers:
innodb, memory, myisam, ndb.
(NDB may have additional issues on some platforms, not covered in this report.)

The difference is nearly 500 lines long, here comes the start:

funcs_2.innodb_charset         [ fail ]

--- /PATH/mysql-test/suite/funcs_2/r/innodb_charset.result
+++ /PATH/mysql-test/suite/funcs_2/r/innodb_charset.reject
@@ -1867,42 +1867,42 @@
 40     1
 41     1
 61     1
+A5     1
+B9     1
 C1     1
-E1     1
 C2     1
-E2     1
 C3     1
-E3     1
 C4     1
+E1     1
+E2     1
+E3     1
 E4     1
-A5     1
-B9     1
 42     1
 62     1
 43     1
 63     1
 C6     1
-E6     1
 C7     1
+E6     1
 E7     1
 C8     1
 E8     1
 44     1
 64     1
 CF     1
-EF     1
 D0     1
+EF     1
 F0     1
 45     1
 65     1
 C9     1
-E9     1
-CC     1
-EC     1
-CB     1
-EB     1
 CA     1
+CB     1
+CC     1
+E9     1
 EA     1
+EB     1
+EC     1
 46     1
 66     1
 47     1
...

How to repeat:
Run the full tests, including the "extra" suite.

Suggested fix:
We will not get stability here unless we make it mandatory for developers to also run these tests when adding patches.

The way we handle it now, changed code meets these tests the first time during release build. I report a bug, some QA developer checks it, some code developer says "yes, harmless, result changed because of ...", the test is adapted -
and this cycle repeats in the next release build.

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/41028

ChangeSet@1.2788, 2008-01-14 11:36:12-05:00, pcrews@pcrews-mac-local.local +8 -0
  Result file re-recorded due to changes in the following collations:
  latin2_czech_cs
  cp1250_czech_cs
  
  Changes made for Bug#33559: Again: "select hex(ascii(a))...order by a" results in different order

The test failures a code bug relating to the czech collations for latin2 and cp1250 character sets.

Bug#33791 has been opened to address this issue.

Tests will continue to fail until the new bug has been corrected.

Signing this off as a duplicate and letting Bug#33791 take it's place