| Bug #56601 | Test uses Unix path for temporary file, fails, and writes misleading message | ||
|---|---|---|---|
| Submitted: | 6 Sep 2010 19:06 | Modified: | 16 Nov 2010 2:02 |
| Reporter: | Joerg Bruehe | Email Updates: | |
| Status: | Closed | Impact on me: | |
| Category: | Tests: Engine | Severity: | S3 (Non-critical) |
| Version: | 5.5.6-m3 | OS: | Windows |
| Assigned to: | Mattias Jonsson | CPU Architecture: | Any |
[15 Sep 2010 8:24]
Bugs System
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/118277 3207 Mattias Jonsson 2010-09-15 Bug#56659: Mismatch of CAPITAL vs small letters in "unified filelist" partitioning output Bug#56657: Test still uses "--exec rm -f ..." which is non-portable Bug#56601: Test uses Unix path for temporary file, fails, and writes misleading message Several tests that was written in a non portable way (failed on windows) Fixed by 1) backporting the fix for replace_result to also apply to list_files (mysqltest from mysql-trunk) 2) replacing all #p#/#sp#/#tmp# to #P#/#SP#/#TMP#/ (innodb always converts filenames to lower case in windows). 3) replacing '--exec rm -f' with '--remove_files_wildcard' 4) replacing a perl snippet with '--write_file' @ client/mysqltest.cc backport from mysql-trunk to allow replace_result to apply also on list_files @ mysql-test/suite/parts/inc/partition_check_drop.inc Compensate for differences between innodb on windows vs unix. Using mysqltest command, instead of unix command to remove files. @ mysql-test/suite/parts/inc/partition_crash.inc compensate for differences between innodb on windows vs unix @ mysql-test/suite/parts/inc/partition_fail.inc compensate for differences between innodb on windows vs unix @ mysql-test/suite/parts/inc/partition_layout.inc compensate for differences between innodb on windows vs unix @ mysql-test/suite/parts/inc/partition_layout_check1.inc compensate for differences between innodb on windows vs unix @ mysql-test/suite/parts/inc/partition_layout_check2.inc compensate for differences between innodb on windows vs unix @ mysql-test/suite/parts/r/partition_recover_myisam.result updated result @ mysql-test/suite/parts/r/partition_special_myisam.result updated result @ mysql-test/suite/parts/t/part_supported_sql_func_innodb.test Test takes very long time, require --big flag @ mysql-test/suite/parts/t/partition_alter1_1_2_innodb.test Test takes very long time, require --big flag @ mysql-test/suite/parts/t/partition_alter1_2_innodb.test Test takes very long time, require --big flag @ mysql-test/suite/parts/t/partition_alter2_1_1_innodb.test Test takes very long time, require --big flag @ mysql-test/suite/parts/t/partition_alter2_1_2_innodb.test Test takes very long time, require --big flag @ mysql-test/suite/parts/t/partition_alter2_2_2_innodb.test Test takes very long time, require --big flag @ mysql-test/suite/parts/t/partition_alter4_innodb.test Test takes very long time, require --big flag @ mysql-test/suite/parts/t/partition_debug_sync_innodb.test compensate for differences between innodb on windows vs unix @ mysql-test/suite/parts/t/partition_recover_myisam.test more generic suppression (failed in windows) @ mysql-test/suite/parts/t/partition_special_myisam.test Using portable mysqltest command 'write_file' instead of perl snippet.
[21 Sep 2010 10:33]
Joerg Bruehe
I checked the attached patch. The fixes for bug#56601, bug#56657, and bug#56659 are approved. The backport in "client/mysqltest.cc" is approved. The generic suppression in test "partition_recover_myisam" is approved. I have mixed feelings about classifying "partition_alter*" and "part_supported_sql_func_innodb" as "big" tests: - Yes, these tests may take quite long, so they may qualify as "big". Depending on the platform, their combined run time may reach 1 hour. - OTOH, I fear we have a tendency to disable tests rather than fix them. This may cause us to lose test coverage. I have no firm opinion whether we should do this change. Please check whether we have continuous PB2 coverage for "big" tests on all platforms in at least the main trees, so that it doesn't get lost. If we do *not* have this, get a second review for this change.
[22 Sep 2010 14:22]
John Embretsen
Patch conditionally approved. See review comment for Bug#56659 (changeset fixes multiple bugs).
[1 Oct 2010 19:50]
Mattias Jonsson
Pushed into mysql-5.5-bugteam and mysql-trunk-merge
[3 Oct 2010 1:18]
Paul DuBois
Changes to test suite. No changelog entry needed.
[9 Nov 2010 19:46]
Bugs System
Pushed into mysql-5.5 5.5.7-rc (revid:sunanda.menon@sun.com-20101109182959-otkxq8vo2dcd13la) (version source revid:sunanda.menon@sun.com-20101109182959-otkxq8vo2dcd13la) (merge vers: 5.5.7-rc) (pib:21)
[13 Nov 2010 16:19]
Bugs System
Pushed into mysql-trunk 5.6.99-m5 (revid:alexander.nozdrin@oracle.com-20101113155825-czmva9kg4n31anmu) (version source revid:alexander.nozdrin@oracle.com-20101113152450-2zzcm50e7i4j35v7) (merge vers: 5.6.1-m4) (pib:21)
[13 Nov 2010 16:36]
Bugs System
Pushed into mysql-next-mr (revid:alexander.nozdrin@oracle.com-20101113160336-atmtmfb3mzm4pz4i) (version source revid:vasil.dimov@oracle.com-20100629074804-359l9m9gniauxr94) (pib:21)
Description: Found in the release build of 5.5.6-m3 on Windows: ===== parts.partition_special_myisam [ fail ] Test ended at YYYY-MM-DD HH:MM:SS CURRENT_TEST: parts.partition_special_myisam Could not open file for writing /test/load.in at G:/pb2/test/sb_1-2235634-1283491829.74/mysql-5.5.6-rc-win32/mysql-test/var\tmp34A6.tmp line 1. mysqltest: At line NNN: command "perl" failed with error 2 The result from queries just before the failure was: < snip > ... # # Bug#53770: Server crash at handler.cc:2076 on LOAD DATA # after timed out COALESCE PARTITION CREATE TABLE t1 ( i INT NOT NULL AUTO_INCREMENT PRIMARY KEY, f INT ) ENGINE = MyISAM PARTITION BY HASH(i) PARTITIONS 3; # Connection 1 starts transaction and gets lock START TRANSACTION; SELECT * FROM t1; i f SET lock_wait_timeout = 2; # Connection 2 tries to coalesce partitions (timeout): ALTER TABLE t1 COALESCE PARTITION 2; ERROR HY000: Lock wait timeout exceeded; try restarting transaction More results from queries before failure can be found in /PATH/mysql-test/var/log/partition_special_myisam.log ===== This happens when an embedded Perl snippet cannot write a temporary file. Note that the path name reported is "/test/load.in" which should not be used and is very likely to fail everywhere, so why does it work on most platforms? Well, in reality it isn't used - here is the code snippet from "mysql-test/suite/parts/t/partition_special_myisam.test": ===== --connect (con3,localhost,root,,) perl; open( LD, ">" . "/tmp/load.in" ) || die "Could not open file for writing " . $ENV{'MYSQLTEST_DATADIR'} . "/test/load.in"; print LD "1\n2\n3\n"; close( LD ); EOF ===== So the real path name is "/tmp/load.in" which will succeed on all Unix platforms (that is why the test passes there), but will still fail on Windows (as experienced in the build). How to repeat: Run the suite on Windows. Suggested fix: The reason is non-portable test code in the fix for bug#53770. Proposed fixes: 1) Search the test files for other Perl snippets opening a temporary file. They typically use path names built from $ENV{MYSQL_TMP_DIR} or $ENV{'MYSQLTEST_VARDIR'} 2) By all means, don't write error messages which write different path names from those used in the code!