GSoC Status: Week 8

Matthew Windsor mbw500 at york.ac.uk
Sun Aug 11 22:48:13 UTC 2013


Hi,

This week has mainly been spent refactoring the job and query code,
and trying (and failing, for now) to figure out how to make jobs
provide more information when they fail.

Spending a week with Code Complete took its toll on me as I've spent a
lot of time trying to subdivide functions, clean up code and generally
make the job/query code less of an unbridled nightmare to look at.  C
files are being split up and fanned out, and hopefully the end result
should be an improvement in both maintainability and (wishfully
thinking) correctness as complexity is shuffled out into easier to
check portions.

The old query-based jobs code has been removed completely, with
nothing using that anymore.  The new jobs code seems to work fine for
removing and installing trivial packages, but again updating is not
yet properly tested.

It turns out that checking the return value of pkg_jobs_apply isn't
going to work for finding out more information about application
errors, and I haven't yet figured out how to check the events system
for errors without causing a job to bail out on minor things such as
checksum mismatches.  This will need some looking into at the pkgng
code side (I might ask on #pkgng too), as I imagine I'll need to
filter through errors either by strcmping the messages or finding some
way of extracting an integer value for the error (does the errno event
have anything to do with this, or is this just for normal unix
errnos?)

Hopefully I'll have more to share next week, by my own admission this
week has been somewhat unproductive in terms of the things I wanted to
do.

~ Matt


More information about the soc-status mailing list