[ptweezy]

After reading a gzip file whose header has a modification time of 0, IGzipFile.mtime is None. gzip.GzipFile gives 0, and so did python-isal 1.3.0. The getter of _GzipReader._last_mtime returns None when the stored time is 0. That check came with the C reader in 1.4.0 (25e1449), where _last_mtime starts at 0, and it makes mtime None before the first read, as test_mtime in the CPython tests expects. It cannot tell a time of 0 from a reader that has not read a header yet.

RFC 1952 says that an MTIME of 0 means no time stamp is available, but the gzip documentation describes mtime after a read as an integer, and gzip.GzipFile returns 0. A modification time of 0 is common. gzip -n, python -m isal.igzip -n, and igzip_threaded writers write it, and so does gzip.compress() by default on Python 3.14. For example, on Python 3.14:

>>> import gzip, io
>>> from isal import igzip
>>> f = igzip.IGzipFile(fileobj=io.BytesIO(gzip.compress(b"data")))
>>> f.read(), f.mtime
(b'data', None)

This PR adds a flag to _GzipReader that is set when it reads the first header, and the getter checks the flag. mtime is still None before the first read. The new test_igzipfile_mtime fails on develop for a time of 0.

Checklist

• Pull request details were added to CHANGELOG.rst
• Documentation was updated (if needed)