On Wed, Sep 5, 2012 at 1:35 AM, Romain d'Alverny <rdalve...@gmail.com> wrote: > On Wed, Sep 5, 2012 at 10:03 AM, Colin Guthrie <mag...@colin.guthr.ie> wrote: >> 'Twas brillig, and Pascal Terjan at 04/09/12 17:57 did gyre and gimble: >>> On Tue, Sep 4, 2012 at 9:52 AM, Romain d'Alverny <rdalve...@gmail.com> >>> wrote: >>> >>>> By the way, there are some builds that get a strangely reported build >>>> duration (15588 days; always the same value when it occurs). I didn't >>>> investigate yet. >> >> So 15588 would be days since epoch. I guess it's the same as Pascal's >> comment. If it was submitted over 2 days ago we might not read the >> submit time and it ends up getting a timestamp of 0 and thus 15588 days >> to build. > > Ok. So at this point (where we do a find to get the files), what's the > good approach: just leave the package/build out of the list if we > can't get its name (that's the new behaviour in my pending new index > page) or fix the find to get all? I'd say the former.
Yes that's fine > Two others things that are aside this very point: > > * we could gain more insight over the whole submission/build process > if we could store data about way past builds (not necessarily the full > details of each build, at least not right now, but at least the > basics: when, who, what package, how long it took on which host, what > was the result). One way to do it would be to have a database in which > we can push and update info about each submission, as it happens > (today, everything is only in the filesystem). What control points > would need to be modified to push information into such a database? Ulri would probably be enough expect for upload, so maybe ulri + emi But then it would be better to instead use that as reference in ulri/emi instead of looking through the filesystem :) > * today, submissions are shown simply by src package. Should we show > submission results per arch? (what if i586 passes and x86_64 fails for > instance). It can not happen unless there is a bug, they are submitted together