[Bug java/12586] New: gnu.java.net name mangling bug

["bryce at mckinlay dot net dot nz" <gcc-bugzilla at gcc dot gnu dot org>]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12586

           Summary: gnu.java.net name mangling bug
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: java
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: bryce at mckinlay dot net dot nz
                CC: gcc-bugs at gcc dot gnu dot org

Michael Koch found this bug in GCJ's name mangler:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Bryce,


I found a weird bug in gcj in namespace mangling. I added a new 
package-private constructor in java.net.ServerSocket:

ServerSocket (gnu.java.net.PlainSocketImpl impl)
{
}

This gets mangled to:

_ZN4java3net12ServerSocketC1EPNS0_15PlainSocketImplE
java::net::ServerSocket::ServerSocket[in-charge]
(java::net::PlainSocketImpl*)

(See the missing "gnu::").

I debugged this a bit and my tests showed that when disabling the 
usage of compression_table the name gets mangled correctly (but not 
compresssed). I think the bug is either in 
find_compression_record_match or compression_table_add (both in gcc/
java/mangle.c).

I saw that you fixed PR4717 two years ago. This may be a corner case 
of this PR. (cv diff -r 1.17 -r 1.18 mangle.c).

It would be really nice if you could help me fixing this. 
Unfortunately I'm not really experienced with GDB and GCC debugging.

I you want I can send you a simple testcase too.


Michael