On Saturday 17 November 2007 08:24:35 Aaron Crane wrote: > The attached patch adds PARROT_API to two IO-related symbols. This > makes the following tests in t/src/io.t pass for me on Linux x86 with > gcc 4.1.3: > > not ok 16 - PIO_make_offset > not ok 17 - PIO_seek > not ok 19 - stdio-layer > > This is my first Parrot patch; please let me know if I'm doing > anything wrong.
The patch looks perfect and it will make the tests pass, but I think the tests are doing the wrong thing, just like the intlist tests. If we have to write tests in C to test various pieces of Parrot that aren't part of the official external API, then we have to slap PARROT_API on all of our symbols to export them in our shared libraries, we should rethink our testing strategy. The IO subsystem is up for a refactoring soon anyway, per a new IO PDD, so I think our best approach is to whack these tests and learn from them what not to do in the future. I'm going to mark them as TODO again to get rid of the failures for now. Anyone else have strong feelings one way or the other? -- c
