[gentoo-soc] Application for "Tree-wide collision checking and provided files database"

[gentoo-soc] Application for "Tree-wide collision checking and provided files database"

[Stanislav Ochotnický]

Hi everyone,

I am applying for gentoo SoC project idea "Tree-wide collision
checking and provided files database". This is my student application,
which you can read also here:
http://student.fiit.stuba.sk/~ochotnic04/files/sochotnicky_gsoc_application.pdf

I am welcome to suggestions to improve it (however with so little time
I do not really expect anything major :-) ).


*Project mission*

Main goal of this project is to create easy way for Gentoo developers
to maintain database of files present in packages residing in Portage
tree. This will open new possibilities for both Gentoo team and users
alike. Gentoo QA team will be able to improve existing ebuilds and
common users will be able to check package contents before installing
packages.

*Abstract*

Gentoo emerge is checking for file collisions before installing
packages. However this collision checking only works after package has
been compiled. This is natural since Gentoo is source based and
package contents are different for every architecture, set of USE
flags and compiler options. Proposed project would provide users and
more importantly Gentoo QA team access to database of files in
packages from Portage tree. Information about package contents would
enable Gentoo QA team to improve existing ebuilds by solving problems
resulting from collisions. Users would be also able to see approximate
size of package and query package contents before installing it. When
invoking non-existing command on shell, shell handler could check file
database and offer packages to install. There are definitely other
uses waiting to be discovered.

*Deliverables*

** Tool for aggregating package contents **

Also information from stat() sycall for files. Most probably this part
will be database and wrapper set of tools. More sources of package
contents would be possible:

       1. tinderbox set-up exactly to collect this information
(semi-automatic, fixing compile errors manually)

       2. binary packages (perhaps interaction with GSoC project
Improved binary package support?)

       3. restricted user supplied package contents (this would need
to be analyzed more deeply from security/privacy point of view)

Public API for read-only access and/or commiting package contents
could be provided depending on analysis and gentoo team preferences.

** Web interface for package contents database **

After database of packages is succesfully created this web interface
would provide world-readable access to package contents through web
browser with ability to search for files and packages.

** Console tool for querying package contents **

Depending on the size of database, it could be downloaded from server
or client could query public server for package contents and/or other
information provided. I would like to keep dependencies of this tool
to the minimum (Python, depending on the database format maybe
sqlite).

All deliverables will include documentation.


* Timeline *

15. April - 23. May: Getting to know Gentoo community spirit a bit
more and some lightweight design discussions for first deliverable

24. May - 7. June (2 weeks): Discussion to design first deliverable

8. June - 21. June (2 weeks): Coding First deliverable with basic
documentation and design for 2nd and 3rd deliverable

22. June - 13. July (3 weeks): Second deliverable (web interface)

14. July - 4. August (3 weeks): Third deliverable coding (client utility)

4. August - 17. August (2 weeks): Finalizing, fixing bugs, cleaning up
code and documentation

Obviously this is only very rough idea. Lot of things will depend on
analysis and design decisions for first deliverable.

* Biography*

I am grad student at Slovak University of Technology in Bratislava
studying Software Engineering. I have participaded on few OSS projects
(musicpd, gstfs, gstreamer) before, mostly with bugreports and
(admittedly very few) patches. My bachelor's project was "Userspace
process access restriction in Linux and FreeBSD". I am quite confident
with C/C++, Java, basic shell scripting and Python. I have also used
quite a few other technologies, but usually only for one or two
projects. Of course I would like to learn more and try something new.
As far as my Linux usage goes, I have been happy Gentoo user for
almost 4 years now. Before that I used Linux From Scratch based
distribution for one year, so I have a lot of experience with build
errors, which should come in handy :-).

My work experience includes participation on coding part of Ripple
control system for energetics companies (together with more smaller
projects for MicroStep-HDO, Slovakia). I've also spent 6 months as
intern in Edisoft (Portugal) where I worked on model-driven
development tool for FPGAs.


--
Stanislav Ochotnicky