Re: [Gzz] jnistrconv.test

[Asko Soukka]

Wed, 13 Nov 2002, Benja Fallenstein kirjoitti:
> I get a single Failure in jnistrconv.test:
> testUniScandJavaSubstr_without_stringlength fails, the rest goes
> through. (I'm using the same VM as before.)

Great! It won't fail with my VM (Sun's 1.4).

> Does that mean the bug is fixed?
> Why is LinebreakingCellContentView broken?

It should mean that. Can you still show me the AssertionError's
message? Did it look familiar?

The bug was on the C++ side as you guessed earlier.

Try to consider a couple code examples below. In the first versio
a string is created from a pointer got from JNI, but without defining
the exact length of the string. I believe, the only way to defined the
length of the string, is to find the first null-char next to the
pointers address - and that why it propably fails when using
String.substr().

Now all JNI-string are converted into C++-strings through the
same function (look at the second example), which asks the length
of the string from JNI. All the other tests than "_withouth_stringlength"
use this function.

Failing version:
  const jchar *strptr = env->GetStringChars(jstr, 0);
  unistring unistr(strptr);

Working version (although, unistring is currenty unicodecharvector):
  const jchar *strptr = env->GetStringChars(jstr, 0);
  unistring unistr(strptr, env->GetStringLength(jstr));

-- 
Asko Soukka     | Taitoniekantie 9 A 603 | address@hidden
+358-40-8235947 | FIN-40740 JYVÄSKYLÄ    | http://www.iki.fi/asko.soukka/