On 7/14/06, Knut Anders Hatlen (JIRA) <[email protected]> wrote:
[ http://issues.apache.org/jira/browse/DERBY-323?page=all ]Knut Anders Hatlen updated DERBY-323: ------------------------------------- Attachment: derby-323-v1.diff derby-323-v1.stat Attached a patch (derby-323-v1.diff) which makes derbyall run cleanlier with non-US locales. The patch does: 1) Rewrite tools/derbyrunjartest.java to use jvm.getCommandLine() to get the command line to start the test with. 2) Move the setting of user.language and user.country from RunTest and NetServer to jvm.getCommandLine(). This way, all of the java processes that are forked from the test harness or from a test get the same locale. 3) Add Locale.setDefault(Locale.US) to RunTest.main(). This fixes the issue with the nist tests. With this patch, derbyall runs cleanly with most locales, for instance pt_BR, de_DE and ja_JP.UTF-8. However, three tests in i18nTest fail with some variants of Japanese locale. ja_JP.ujis and ja_JP.SJIS have these three failures. ja_JP.eucJP has these failures on some platforms. I would assume other locales with non-Unicode/non-Latin character sets also could see these failures. The failing tests are: i18n/urlLocale.sql i18n/messageLocale.sql i18n/iepnegativetests_ES.sql All the failures are caused by non-ascii characters that are printed as '?' because they don't exist in the default character set for the locale. I believe this could be fixed by a combination of running the i18n tests with -Dfile.encoding=something and specifying encoding to streams/readers in RunTest.
I think those three failures fall under 'my' favorite DERBY-244. Would you like to take it? :-) I've been sitting on it without much progress for a long, long time, just so it doesn't slip off my radar. Myrna
