Re: [Xen-devel] [PATCH OSSTEST v1] mg-all-branch-statuses: Show how up to date each branch is

[Ian Campbell <ian.campbell@xxxxxxxxxx>]

On Mon, 2015-06-29 at 15:15 +0100, Ian Jackson wrote:
> Ian Campbell writes ("Re: [PATCH OSSTEST v1] mg-all-branch-statuses: Show how 
> up to date each branch is"):
> > On Mon, 2015-06-29 at 13:25 +0100, Ian Jackson wrote:
> > > It's probably fairly straightforward to search through the database or
> > > history to find some relevant date.  What do want to know ?
> > > 
> > >  1. `Commit' timestamp of last pass ?
> > >  2. `Commit' timestamp of first commit after last pass ?
> > >  3. `Commit' timestamp of tip ?
> > 
> > 2 sounds tricky, especially since there could potentially be several.
> 
> Yes.  Although `earliest-dated nontrivial descendant of the last pass
> commit' is unique, it ...
> 
> > In general commit timestamps can be a bit misleading, since various
> > things preserve them even over long periods of time before they are
> > published and they don't necessarily correspond at all with when
> > something became visible in the tree one is looking at.
> 
> ... is likely to exacerbate this problem.  (Although I did say `commit
> timestamp', which in git is a separate timestamp updated by rebase,
> cherry pick, etc., as opposed to `author timestamp'.)

I knew there was two timestamps but not that they differed in this way,
so good.

> > Maybe better than nothing though.
> > 
> > >  4. Start date of last flight for current baseline ?
> > >  5. Start date of first flight after last flight for current baseline ?
> > >  6. Start date of first flight for current tip ?
> > > 
> > > 5 ?
> > 
> > You mean out of all 6 options you think 5 is all we need?
> 
> Yes.
> 
> > The date of the of the flight which established the current baseline
> > seems like a no-brainer (I suppose that is #4).
> 
> That doesn't tell us how out of date it is.  If the tree is updated
> rarely but the tip was updated yesterday it might look like the test
> has been failing for months.

Ah yes, that's a useful way of thinking about things which I wasn't
doing...

> > I'm not really sure what date of a flight relating to tip would be most
> > useful, first try of something newer (#5?) vs most recent attempt (#6).
> 
> The former tells you how long it has been failing.

Right and is therefore much more interesting.

> > Maybe even the _number_ of bites of the cherry would be useful?
> 
> Yes.  You'd want both `number of times _this version_ has been tried'
> and `total number of flights since last pass'.

True.

> Perhaps I should make a library function in Osstest::Executive which
> provides this date and count.  After all they would be useful in the
> email reports.

That would be good, thanks.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel