Issue2324

This shows up in testing ZODB under Jython. ZODB creates a StringIO object and repeatedly uses it to pickle objects, resetting it each time. Given the write sequence of calls, this results in a StringIndexOutOfBoundsException.

>>> from cStringIO import StringIO
>>> before = 'cpersistent.mapping\nPersistentMapping\nq\x00.}q\x01U\x04dataq\x02}q\x03U\x04testq\x04(U\x08\x00\x00\x00\x00\x00\x00\x00\x01q\x05h\x00tq\x06Qss.'
>>> sio.write(before)
>>> sio.truncate()
>>> sio.seek(0)
strs = ['cpersistent.mapping\nPersistentMapping\n',
... 'q\x00',
... '.',
... '}',
... 'q\x01',
... 'U\x04data',
... 'q\x02',
... '}',
... 'q\x03',
... '(',
... 'K\x00',
... '(',
... 'U\x08\x00\x00\x00\x00\x00\x00\x00\x02',
... 'q\x04',
... 'h\x00',
... 't',
... 'q\x05',
... 'Q',
... 'K\x01',
... '(',
... 'U\x08\x00\x00\x00\x00\x00\x00\x00\x03']
>>> for i in strs:
...  sio.write(i)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
	at java.lang.AbstractStringBuilder.replace(AbstractStringBuilder.java:828)
	at java.lang.StringBuilder.replace(StringBuilder.java:262)
	at org.python.modules.cStringIO$StringIO.write(cStringIO.java:375)
	at sun.reflect.GeneratedMethodAccessor8.invoke(Unknown Source)
	at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
	at java.lang.reflect.Method.invoke(Method.java:497)

java.lang.StringIndexOutOfBoundsException: java.lang.StringIndexOutOfBoundsException: start > end

The exception happens on the very last string in the list.

Wow, ZODB? That's crazy talk on Jython! ;)

I will take a look and see if we can fix quickly, since that would be very cool.

ZODB does have a workaround in using `io.BytesIO`, so it's not super-urgent from that perspective (although I think StringIO is probably faster?). It does seem like a general bug though that anyone could hit.

So it's a problem in using seek. and how that interacts with a previous allocation of bytes in terms of determining additional capacity.

Easy fix - the replacement, then subsequent append of the StringBuilder is just not computing correct parameters. Fix forthcoming.

Fixed as of https://hg.python.org/jython/rev/ce136d0b598f

Thank you. I can confirm that ce136d0b598f does fix the ZODB tests.