构建共享对象 - 使用 ninja 编译在 osx 上工作,ubuntu crush 在 windows 上工作

building shared object - compiling with ninja works on osx and ubuntu crush on windows

我正在尝试为基于 JVM 的语言构建一个 cld3 包装器,使用 Java Abstracted Foreign Function Layer .

我创建了一个小的 class 将答案从 c++ 库转换为缓冲区。

void detect(long ptr, const char *text, const char *into) {

    NNetLanguageIdentifier *nptr = (NNetLanguageIdentifier *) ptr;
    NNetLanguageIdentifier::Result res = nptr->FindLanguage(text);

    long current = (long) into;

    *((float *) current) = res.probability;
    current += sizeof(float);
    *((float *) current) = res.proportion;
    current += sizeof(float);
    *((short *) current) = res.is_reliable;
    current += sizeof(short);
    *((int *) current) = res.language.size();
    current += sizeof(int);

    memccpy(reinterpret_cast<void *>(current), res.language.c_str(), res.language.size() + 1, sizeof(res.language));

}

然后在 JVM 端:

   fun detect(text: String): LangDetectResponse {
        val buffer = ByteArray(SINGLE_LANGUAGE_DETECTION_BUFFER_SIZE)
        detector.detect(ptr, text, buffer)

        ByteBuffer.wrap(buffer).order(ByteOrder.LITTLE_ENDIAN).let {
            return extractResponseFromByteBuffer(it)
        }
    }

    private fun extractResponseFromByteBuffer(byteBuffer: ByteBuffer): LangDetectResponse {
        val probability = byteBuffer.float
        val proportion = byteBuffer.float
        val isReliable = byteBuffer.short
        val sizeOfLanguage = byteBuffer.int
        val languageBuffer = ByteArray(sizeOfLanguage) { 127.toByte() }
        byteBuffer.get(languageBuffer, 0, sizeOfLanguage)
        val language = String(languageBuffer, StandardCharsets.UTF_8)
        return LangDetectResponse(probability, proportion, isReliable.toInt() == 1, language)
    }

我设法使用 The Ninja 构建系统编译了这段代码 .

并创建了 so 和 dylib 文件:

https://github.com/ntedgi/cld3-kotlin#operations-systems-support

一切正常。

我试图在 windows 上创建一个共享对象,使用我用来编译其他操作系统的相同代码,但我得到了这个错误:

../../third_party/cld_3/src/src/Cld3LangDetector.cc(19,36): error: cast to 'chrome_lang_id::NNetLanguageIdentifier *' from smaller integer type 'long' [-Werror,-Wint-to-pointer-cast]
    NNetLanguageIdentifier *nptr = (NNetLanguageIdentifier *) ptr;
                                   ^
../../third_party/cld_3/src/src/Cld3LangDetector.cc(25,36): error: cast to 'chrome_lang_id::NNetLanguageIdentifier *' from smaller integer type 'long' [-Werror,-Wint-to-pointer-cast]
    NNetLanguageIdentifier *nptr = (NNetLanguageIdentifier *) ptr;
                                   ^
../../third_party/cld_3/src/src/Cld3LangDetector.cc(31,7): error: cast to 'float *' from smaller integer type 'long' [-Werror,-Wint-to-pointer-cast]
    *((float *) current) = res.probability;
      ^
../../third_party/cld_3/src/src/Cld3LangDetector.cc(33,7): error: cast to 'float *' from smaller integer type 'long' [-Werror,-Wint-to-pointer-cast]
    *((float *) current) = res.proportion;
      ^
../../third_party/cld_3/src/src/Cld3LangDetector.cc(35,7): error: cast to 'short *' from smaller integer type 'long' [-Werror,-Wint-to-pointer-cast]
    *((short *) current) = res.is_reliable;
      ^
../../third_party/cld_3/src/src/Cld3LangDetector.cc(37,7): error: cast to 'int *' from smaller integer type 'long' [-Werror,-Wint-to-pointer-cast]
    *((int *) current) = res.language.size();
      ^

我已经尝试将 cflags 传递给忍者版本。

我很高兴知道如何解决这个问题,也很高兴了解为什么其他操作系统容忍这个错误。

谢谢。

long 不是存储指针值的合适类型。在 64 位上 Windows long 通常只有 32 位。您应该使用保证足够大的 ::std::uintptr_t