Reproduction steps:

---
$ python2.7 -c "import zlib;zlib.crc32('a'*(1<<31))"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
OverflowError: size does not fit in an int
---

We ran into this bug in zlib.crc32 when using zipfile.writestr() with a very large string; as soon as zipfile tried to write the crc checksum, it raised this error.


Python 3 does not appear to suffer from this bug.

This bug prevents zipfile's writestr() from writing large data (longer than UINT_MAX) to a 64-bit zip file.

The zlib.crc32 function which, as written, cannot accept input with a size larger than an unsigned int.

https://hg.python.org/cpython/file/94ec4d8cf104/Modules/zlibmodule.c#l964

Python 3 has updated this to call the zlib crc32 function multiple times in this situation:

https://hg.python.org/cpython/file/93888975606b/Modules/zlibmodule.c#l1210

so the fix exists, we just need to do this in 2.7.

Unfortunately, fixing just zlib.crc32 isn't quite enough for our purposes.  We still will see OverflowErrow in zipfile if compression is selected.


Demonstration code:

############################################
import zipfile

## Possible workaround: monkey-patch crc32 from binascii?!
import binascii
zipfile.crc32 = binascii.crc32

content = 'a'*(1<<31)
filename = '/tmp/zip_test.zip'

zf = zipfile.ZipFile(filename, "w",
                     compression=zipfile.ZIP_DEFLATED,
                     allowZip64=True)
zf.writestr('big', content)
zf.close()

zf = zipfile.ZipFile(filename, "r", allowZip64=True)
print zf.open('big').read() == content
#############################################


This will raise the following error under Python 2.7.6:

#############################################
$ python zip_test.py
Traceback (most recent call last):
  File "zip_test.py", line 13, in <module>
    zf.writestr('big', content)
  File "/usr/lib/python2.7/zipfile.py", line 1228, in writestr
    bytes = co.compress(bytes) + co.flush()
OverflowError: size does not fit in an int
#############################################



If we use compression=zipfile.ZIP_STORED, we don't see this error, but it kind of misses a major point of using zipfile.

Apparently crc32() was fixed in Python 3 via Issue 10276.

See also Issue 27130 about 64-bit support more generally in zlib.

This appears to have been fixed by at least issue27130's https://hg.python.org/cpython/rev/2192edcfea02 recent commits.

greg:cpython/build27$ ./python -c "import zlib;zlib.crc32('a'*(1<<31))"
greg:cpython/build27$ ./python ../zipfile_2gb_test.py
True
greg:cpython/build27$ ls -al /tmp/zip_test.zip
-rw-rw-r-- 1 greg greg 2087407 Aug  7 09:28 /tmp/zip_test.zip
greg:~/sandbox/python/cpython/build27$ unzip -t /tmp/zip_test.zip
Archive:  /tmp/zip_test.zip
    testing: big                      OK
No errors detected in compressed data of /tmp/zip_test.zip.
greg:cpython/build27$ unzip -l /tmp/zip_test.zip
Archive:  /tmp/zip_test.zip
  Length      Date    Time    Name
---------  ---------- -----   ----
2147483648  2016-08-07 09:27   big
---------                     -------
2147483648                     1 file