Store image files on the filesystem

[patriotyk]

Signed-off-by: Serhiy Stetskovych patriotyk@gmail.com

[codecov-io]

Codecov Report

Merging #1091 into master will increase coverage by 0.06%.
The diff coverage is 81.39%.

@@            Coverage Diff             @@
##           master    #1091      +/-   ##
==========================================
+ Coverage   75.66%   75.72%   +0.06%     
==========================================
  Files         158      158              
  Lines        9430     9392      -38     
==========================================
- Hits         7135     7112      -23     
+ Misses       2295     2280      -15

Impacted Files	Coverage Δ
src/libaktualizr/storage/sqlstorage.h	50% <ø> (ø)	⬆️
src/libaktualizr/storage/sqlstorage_base.h	100% <ø> (ø)	⬆️
src/libaktualizr/storage/invstorage.h	97.43% <ø> (ø)	⬆️
src/libaktualizr/uptane/fetcher.cc	96.9% <100%> (ø)	⬆️
src/libaktualizr/storage/sqlstorage.cc	73.04% <80.48%> (+0.61%)	⬆️
src/libaktualizr/storage/sql_utils.h	80.59% <0%> (+0.31%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 096e7b5...b08bce8. Read the comment docs.

[eu-siemann]

It looks to me that we will enter unrecoverable state if the binary file disappears from the filesystem:
checkTargetFile() will return true as it finds the database entry, but subsequent SQLTargetRHandle constructor will throw.

[eu-siemann]

I'm also not so sure about storing the files under the user-provided filename. As there can be additional directories in the filename, we either risk to end up with hundreds of empty dirs, or need to implement some directory cleanup logic.
What do you think about having just one folder for all binaries and using hash as a filename?

[pattivacek]

What do you think about having just one folder for all binaries and using hash as a filename?

I think that is not a bad idea. However, TUF seems to disagree. See section 5.5.2. Uptane seems to be on the same page.

[eu-siemann]

@patrickvacek

the client MUST write the file to non-volatile storage as FILENAME.EXT

Well, if we say that this requirement is about the name in a filesystem, then storing binary blobs in sqlite db is also wrong :)
I guess it's more about to be able to associate FILENAME.EXT with the corresponding binary data, and the rest is an implementation detail.

[pattivacek]

Well, if we say that this requirement is about the name in a filesystem, then storing binary blobs in sqlite db is also wrong :)

Indeed!

I guess it's more about to be able to associate FILENAME.EXT with the corresponding binary data, and the rest is an implementation detail.

Yes, I agree. I can't think of a good reason why storing the files by their hash would be a bad thing. I can see some benefit in using the names as provided for visibility and maybe maintaining a "nice" directory structure, but that could be as much of a curse as it could be a blessing, and dealing with special characters could be annoying, and no one is really supposed to be looking at the hard drive anyway, so the perceived benefit would be minimal at best.

[patriotyk]

Please check, should be fine now.

[eu-siemann]

Thanks, Serhiy! Looks fine to me.

[lbonn]

I've rebased the PR and added a small tweaks commit for the minor style issues we've found.

[pattivacek]

Looks good!