stat opcode causes program crash on Win32 with invalid file descriptor

[whiteknight]

The test file t/dynoplibs/io.t has a few tests for the dynop stat_i_i_i and family that cause the test file to crash when using a nonsensical value for the file descriptor. Parrot terminates when the test is run with no warning. In dubgging, it appears that there is some kind of error happening at the win32-level, which isn't being handled and causes the program to terminate.

These tests are new additions from GCI.

I'm commenting out those tests for now while we investigate the cause and possible solutions.

[whiteknight]

I've commented out the offending tests in 5d54e34

[nwellnhof]

The passed INTVAL is converted to a Windows file handle which is a void *. I don't think it's possible to check if a handle is valid, so there's no way to avoid crashes.

IMO, we simply shouldn't use raw filehandles and deprecate that opcode.

[whiteknight]

+1 from me to remove.

[nwellnhof]

I removed this dynop in branch nwellnhof/unicode_dynpmcs.

[nwellnhof]

stat_i_i_i has been removed. I can't imagine this op has been used, because afaics there's no way to get a valid filehandle. So we shouldn't need a deprecation notice.