mz_zip_add_mem_to_archive_file_in_place in Android will trigger the fdsan mechanism to kill the process.

[initlist]

My test codes

int test_zip()
    {
        mz_zip_archive zip;
        memset(&zip, 0, sizeof(zip));

        if (!mz_zip_writer_init_file(&zip, "/sdcard/hanwang/wordResult/test.zip", 0))
        {
            LOGI("Failed to initialize zip writer\n");
            return -1;
        }
        if (!mz_zip_writer_finalize_archive(&zip))
        {
            LOGE("Failed to finalize zip archive\n");
            mz_zip_writer_end(&zip);
            return -1;
        }

        for (int i = 0; i < 50; i++)
        {
            const char *str = "hello world";
            LOGI("%d add string to zip: %s\n", i, str);
            if (!mz_zip_add_mem_to_archive_file_in_place(
                    std::string("/sdcard/hanwang/wordResult/test.zip").c_str(), ("file1.txt" + std::to_string(i)).c_str(),
                    str, (mz_uint16)strlen(str),
                    NULL, 0,
                    MZ_BEST_COMPRESSION))
            {
                LOGI("Failed to add file1.txt to zip\n");
                return -1;
            }
            // sleep 1s
            // std::this_thread::sleep_for(std::chrono::seconds(1));
        }
        return 0;
        // 关闭 ZIP 文件
        if (!mz_zip_writer_finalize_archive(&zip))
        {
            LOGE("Failed to finalize zip archive\n");
            mz_zip_writer_end(&zip);
            return -1;
        }

        mz_zip_writer_end(&zip);
        LOGE("test_zip() returns 0\n");

        return 0;
    }

The above code will crash if I don't sleep. I found that fdsan is triggered because of repeated reading and writing of the same fd. If I sleep, fdsan will not be triggered.

[RealForMe]

I have also encountered the same problem. Can I avoid android_fdsan_close_with_tag by simply sleeping?

[uroni]

Looks like an Android (bionic) issue to me.

Does this commit fix this? 11e7c39