(Bug report for version 2.5.7, which was not available on bugzilla)
Python bindings for inputBuffer/Reader interface use high-level file access
(like fread() in C) when low-level access (like reade()) could (and should)
be used. Reader waits for some (hardcoded?) number of characters instead of
processing what is available. This makes reader.Read() block even if there
are full tags available, which could be returned.

This is easy to workaround with the following code, but IMHO it should be
fixed in the bindings.

workaround example:

class MyFile:                                                             
            
    def __init__(self,file):                                              
                   
        self.fd=file.fileno()
        self.file=file
    def read(self,len):
        return os.read(self.fd,len)

input=libxml2.inputBuffer(MyFile(sys.stdin))
reader=input.newTextReader("ble")