Fix erroneous install errors when iterating on an archive (preserving its name but changing its contents).
Review Request #243 - Created March 29, 2011 and submitted
| Alain Linden | Reviewers | ||
| oz.linden, brad.linden, nat_linden | |||
| None | autobuild | ||
Fixed a bug where an error would be thrown if one tried to install an archive with a filename that matches a cached version but is updated. One could hit this if they updated the same archive multiple times in a day (like when doing development work). Now if the cached archive doesn't match the hash, a download will be attempted so only true download errors will result in an exception.
Expand AllPosted (March 29, 2011, 1:18 p.m.)
Couple comments.
First, I think I missed a bet in my small change. Probably there should be a local hash_algorithm var:
hash_algorithm = archive.hash_algorithm or "md5"
if not hash_algorithms.verify_hash(hash_algorithm, cachefile, archive.hash): ...
Second, I'm not enthused about the download_required flag. Three cases, right?
* no such tarball in cache
* tarball in cache by that name has wrong MD5
* tarball in cache by that name is correct
So what about logic like this?
if os.path.exists(cachefile):
if hash_algorithms.verify_hash(archive.hash_algorithm, cachefile, archive.hash):
logger.debug("found in cache: " + cachefile)
else:
# Cache contains tarball by that name, but wrong hash. Try downloading again.
common.remove_package(archive.url)
if not os.path.exists(cachefile):
# download the package to the cache
...
Other reviews