A Perl CORBA::ORBit test revealed a problem with the static initializers of typecodes in ORBIT-0.5.16 typecodes.c - there was no repository ID and some of the names were inconsistent with IDL. Unless you attempt to introspect the system typecodes you won't see this, which wouldn't be common in C but is something people tend to do in dynamic environments like Perl. The enclosed patch relative to 0.5.16 fixes this, and also fixes the alloca() configure problem on Solaris without impacting linux as far as I can test. Be sure to regenerate config.h.in from acconfig.h before distributing. Rgds, -Huw Rogers > This needs a patch to ORBit-0.5.16, which I am testing now and will post > to orbit-list and here too. > > > I've added a few tests to start off a CORBA::ORBit test suite. There is > > a test of orb initialisation, a test of the recent typecode work, and a > > simple any encode/decode. You can run tests using the usual "make test" > > rule. > > > > Huw, the typecodesys.t test shows a couple of errors for me when testing > > the CORBA::Long case. (i.e. its id is empty and the name is "long" > > instead of "Long".). I get the same errors with both ORBit 0.5.15 and > > 0.5.16. Is this expected? > > > > The user defined typecodeuser.t test cases are fine. > > > > Cheers, > > Alex. > > -- > Huw Rogers <count0 metavector com> -- Huw Rogers <count0 metavector com>
Attachment:
ORBit-0.5.16.patch
Description: Binary data