BTW, to reproduce, you can go pull source from: https://github.com/pixie-lang/pixie Then run ./make-with-jit The build script assumes that above the pixie directory you have pypy checkout in a directory named "pypy". On Sat, Oct 18, 2014 at 5:11 PM, Timothy Baldridge <tbaldridge@gmail.com> wrote:

What's also odd, is that there is something non-deterministic about these errors, sometimes I get the one error, but if I translate again (without changing any code) I'll get the other.

On Sat, Oct 18, 2014 at 2:08 PM, Timothy Baldridge <tbaldridge@gmail.com> wrote:

...

My interpreter is built using mostly unicode for symbols and strings, but recently I've been getting some really weird translation errors. The first is this: https://gist.github.com/halgari/0d57dd87434968561705

I tracked this error down to being caused whenever I try an isinstance of unicode like this:

isinstance(x, unicode)

This is really annoying as I'd love to have a single unified wrap function:

@specialize.argtype(0) def wrap(x): if isinstance(x, int): return numbers.Integer(x) if isinstance(x, unicode): return String(x)

And as of this morning I started getting errors like this:

[translation:ERROR] TyperError: don't know how to convert from <UnicodeRepr * GcStruct rpy_unicode { hash, chars }> to <UniCharRepr UniChar> [translation:ERROR] .. (pixie.vm.reader:47)PromptReader.read [translation:ERROR] .. block@82 with 1 exits [translation:ERROR] .. v235 = simple_call(v234)

What is a UniChar? My code only deals with unicode strings, so I'm not sure what's happening here.

Thanks for any help. I've had unicode working perfectly with my interpreter for weeks, and suddenly in the past two days I've started getting these errors.

Timothy

-- “One of the main causes of the fall of the Roman Empire was that–lacking zero–they had no way to indicate successful termination of their C programs.” (Robert Firth)

-- “One of the main causes of the fall of the Roman Empire was that–lacking zero–they had no way to indicate successful termination of their C programs.” (Robert Firth)

BTW, to reproduce, you can go pull source from: https://github.com/pixie-lang/pixie Then run ./make-with-jit The build script assumes that above the pixie directory you have pypy checkout in a directory named "pypy". On Sat, Oct 18, 2014 at 5:11 PM, Timothy Baldridge <tbaldridge@gmail.com> wrote:

What's also odd, is that there is something non-deterministic about these errors, sometimes I get the one error, but if I translate again (without changing any code) I'll get the other.

On Sat, Oct 18, 2014 at 2:08 PM, Timothy Baldridge <tbaldridge@gmail.com> wrote:

...

My interpreter is built using mostly unicode for symbols and strings, but recently I've been getting some really weird translation errors. The first is this: https://gist.github.com/halgari/0d57dd87434968561705

I tracked this error down to being caused whenever I try an isinstance of unicode like this:

isinstance(x, unicode)

This is really annoying as I'd love to have a single unified wrap function:

@specialize.argtype(0) def wrap(x): if isinstance(x, int): return numbers.Integer(x) if isinstance(x, unicode): return String(x)

And as of this morning I started getting errors like this:

[translation:ERROR] TyperError: don't know how to convert from <UnicodeRepr * GcStruct rpy_unicode { hash, chars }> to <UniCharRepr UniChar> [translation:ERROR] .. (pixie.vm.reader:47)PromptReader.read [translation:ERROR] .. block@82 with 1 exits [translation:ERROR] .. v235 = simple_call(v234)

What is a UniChar? My code only deals with unicode strings, so I'm not sure what's happening here.

Thanks for any help. I've had unicode working perfectly with my interpreter for weeks, and suddenly in the past two days I've started getting these errors.

Timothy

-- “One of the main causes of the fall of the Roman Empire was that–lacking zero–they had no way to indicate successful termination of their C programs.” (Robert Firth)

-- “One of the main causes of the fall of the Roman Empire was that–lacking zero–they had no way to indicate successful termination of their C programs.” (Robert Firth)