Crash in resource manager

[ddyer0]

The attached test program (with the bugs.res file) runs on the simultor and on android, but crashes
on ios with a RunTimeException "decoding error".

The exact nature of the crash depends on mysterious factors, including not crashing if the sequence
of resource operations is perturbed slightly.

bugs.zip

[shai-almog]

repaint() is not thread-safe and you're invoking it from two threads. Things might work, but this is a clear violation of the EDT. repaint() in paint() isn't something we support.

[ddyer0]

that's an incidental feature of the demo program. remove it.

[shai-almog]

Please include a smaller sample in the issue body. If you have a crash:

• Is it on the simulator - include logs
• Is it on device only - which device? Include logs (ideally for debug build)

[ddyer0]

This is IOS only; works fine on simulator and android. The general code and procedures for getting
data out of .res files has been in use for years; but since perturbations in the reading procedure made
the problem go away, I can't guarantee that it's a genuine IOS-only problem.

It was only in the final stages of making the demo program that the nice "decoding error' error
occurred, prior to that it was a "hangs and does nothing" type problem that behaved as though
the reading were stuck in a loop.

As for size, I think this is as small as it gets, in particular the inclusion of the bufferedreader class in the tool
chain is necessary. But at least the error occurs immediately and predictably. Bufferedreader uses an 8K
buffer, and the data size is a lot more than 8K, I suspect that's a necessary condition.

[ddyer0]

More experimental evidence:

I ran the same binary that fails on my Ipad on an Iphone, and it also fails there.

I switched the demo to use a different (separately constructed) .res file containing the same data, and it works on IOS.

[shai-almog]

If you got a "decoding error" then I don't know. If you got a "Decoding Error" then it's probably this:

CodenameOne/vm/ByteCodeTranslator/src/nativeMethods.m

Line 273 in 98febe6

java_lang_RuntimeException___INIT_____java_lang_String(CN1_THREAD_STATE_PASS_ARG ex, newStringFromCString(CN1_THREAD_STATE_PASS_ARG "Decoding Error"));

That means a UTF-8 string failed to decode on iOS which might mean one of the strings within there is broken or some of the other logic led to a point where we're trying to read something that isn't a string.

[ddyer0]

I agree that might be the error I am hitting, but there's no request for utf8 decoding in the tool chain,
and the error still occurs if I add the optional argument "ASCII" as the encoding type to InputStreamReader.

[shai-almog]

All strings in Codename One are UTF-8. It's possible that when you copied a string into the resource file a problematic character slipped in and somehow this corrupted the file. It's also possible there's a bug in reading the file or in our UTF-8 decoding logic.

[ddyer0]

that's exactly what I discovered when developing my workaround; the text I was reading contained a few rogue
ascii chacters in the 128-255 range, which desktop java, the simulator, and android just shrug and pass on
the unusual but legal ascii values. ios throws a fit.