Ticket #1551 (assigned defect)
filestore.retrieve(): chmod() don't make sense
| Reported by: | sascha_silbe | Owned by: | alsroot |
|---|---|---|---|
| Priority: | Unspecified by Maintainer | Milestone: | Unspecified by Release Team |
| Component: | sugar-datastore | Version: | Git as of bugdate |
| Severity: | Unspecified | Keywords: | |
| Cc: | Distribution/OS: | Unspecified | |
| Bug Status: | Unconfirmed |
Description
While fixing #1550, I stumbled over this code again:
# Try to hard link from the original file to the targetpath. This can
# fail if the file is in a different filesystem. Do a symlink instead.
try:
os.link(file_path, destination_path)
except OSError, e:
if e.errno == errno.EXDEV:
os.symlink(file_path, destination_path)
else:
raise
# Try to make the original file readable. This can fail if the file is
# in a FAT filesystem.
try:
os.chmod(file_path, 0604)
except OSError, e:
if e.errno != errno.EPERM:
raise
I guess the chmod() is supposed to make the file readable to the caller in case we use a symlink instead of a hardlink.
The permissions don't make much sense:
* allow read+write for the owner (data store user) - OK
* no access to the group - ?
* read access to world - overrides group settings
Was it supposed to be 0640 (rw- r-- ---) instead?
Why do we always do the chmod() instead of only when we actually symlink, BTW? A hardlink doesn't share permissions, only data.
Do we support data stores on FAT filesystems at all? AFAICT our optimizer will fail hard (at least in filestore.hard_link_entry()) on those.
