Bug 19647
| Summary: | 8859-1 and other encodings fail with iconv (endianess) | ||
|---|---|---|---|
| Product: | [Retired] Red Hat Linux | Reporter: | krip |
| Component: | jikes | Assignee: | Trond Eivind Glomsrxd <teg> |
| Status: | CLOSED RAWHIDE | QA Contact: | Aaron Brown <abrown> |
| Severity: | medium | Docs Contact: | |
| Priority: | medium | ||
| Version: | 7.0 | ||
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | i386 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2000-11-15 11:13:14 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
I've asked one of the jikes developers to have a look at it.... Here's the answer from Mo DeJong (mdejong): Yeah, we know about this one. It is the single worst problem in 1.12. Just as we fixed the CLASSPATH crashing problem in 1.11, another nasty one shows up (Ugh). Fixing this problem does not look too hard for any one locale, the problem is we need to get regression tests in place before a bug fix can be checked into the CVS. We need tests for every single unicode char and its escaped equiv, so that we can be sure it works everywhere. I am sure I will get around to fixing it soon. ok, the whole bug report is somewhat confusing and while we're waiting for a proper fix from the jikes people here is an explanation of what i've come to understand. The first part of the patch is needed for jikes to compile, and incorporated into redhat's src.rpm. The second patch is needed for the -encoding to work. Thus if you want to compile java-files with for example ISO-8859-1 encoding, you need to apply the second patch and provide '-encoding iso-8859-1' as an argument to jikes when compiling. I think it would be nice to have the ISO-8859-1 as the implicit default encoding (as javac and jikes on other systems work) but that isn't really needed. The patch has been applied to jikes-1.12-2 - a total fix should be part of 1.13, whenever that is released. |
the redhat 7.0 jikes-1.12-1 dist doesn't grok chars >128 when build with iconv (binary as well as naive source build). without an encoding specified, they are invalid (jikes say \uFFF6 ...), with -encoding option, the whole file is broken :( to make it work, I had to patch line 895 in stream.cpp v 1.40 (char ** cast on 2nd (!) arg to iconv) and line 552 in option.cpp v 1.43 (explicitly demanding UTF-16BE !?). diff -c orig/jikes-1.12/src/stream.cpp jikes-1.12/src/stream.cpp *** orig/jikes-1.12/src/stream.cpp Tue Jul 25 13:32:33 2000 --- jikes-1.12/src/stream.cpp Tue Oct 24 01:41:40 2000 *************** *** 890,896 **** size_t chl = 2; size_t srcl = 1; size_t n = iconv(control.option.converter, ! &source_ptr, &srcl, (char **)&chp, &chl ); --- 890,896 ---- size_t chl = 2; size_t srcl = 1; size_t n = iconv(control.option.converter, ! (char **)&source_ptr, &srcl, (char **)&chp, &chl ); diff -c orig/jikes-1.12/src/option.cpp jikes-1.12/src/option.cpp *** orig/jikes-1.12/src/option.cpp Mon Jul 31 00:56:31 2000 --- jikes-1.12/src/option.cpp Tue Oct 24 01:53:54 2000 *************** *** 251,257 **** #elif defined(HAVE_ICONV_H) encoding = new char[strlen(arguments.argv[++i]) + 1]; strcpy(encoding, arguments.argv[i]); ! converter=iconv_open("utf-16", encoding); if(converter==(iconv_t)-1) { converter = NULL; --- 251,257 ---- #elif defined(HAVE_ICONV_H) encoding = new char[strlen(arguments.argv[++i]) + 1]; strcpy(encoding, arguments.argv[i]); ! converter=iconv_open("UTF-16BE", encoding); if(converter==(iconv_t)-1) { converter = NULL;