Recipe Renames, etc
Matt Spinler
mspinler at linux.vnet.ibm.com
Thu Oct 11 06:19:14 AEDT 2018
On 2018-10-10 14:02, Patrick Venture wrote:
> On Wed, Oct 10, 2018 at 12:01 PM Patrick Venture <venture at google.com>
> wrote:
>>
>> On Wed, Oct 10, 2018 at 12:00 PM Matt Spinler
>> <mspinler at linux.vnet.ibm.com> wrote:
>> >
>> > On 2018-10-10 13:50, Patrick Venture wrote:
>> > > On Wed, Oct 10, 2018 at 11:46 AM Patrick Venture <venture at google.com>
>> > > wrote:
>> > >>
>> > >> On Wed, Oct 10, 2018 at 11:45 AM Patrick Venture <venture at google.com>
>> > >> wrote:
>> > >> >
>> > >> > On Wed, Oct 10, 2018 at 11:38 AM Matt Spinler
>> > >> > <mspinler at linux.vnet.ibm.com> wrote:
>> > >> > >
>> > >> > > On 2018-10-02 10:28, Patrick Venture wrote:
>> > >> > > > I've sent a bunch of recipe renames, but I'm still updating each with
>> > >> > > > an individual tested footer. I will try marking them all as WIP now
>> > >> > > >
>> > >> > > > Patrick
>> > >> > >
>> > >> > >
>> > >> > > Hi Patrick,
>> > >> > >
>> > >> > > I think you broke master with these. For example, the phosphor-rest
>> > >> > > recipe - it has a default url-config.json, and then a layer specific file gets
>> > >> > > applied from the meta-ibm layer, like:
>> > >> > >
>> > >> > > meta-ibm/recipes-phosphor/interfaces/phosphor-rest/url_config.json
>> > >> > >
>> > >> > > However, from the log.do_unpack from the phosphor-rest build, it doesn't
>> > >> > > look there, it now only looks in phosphor-rest%. So any override directory without
>> > >> > > a % isn't being found.
>> > >> >
>> > >> > Interesting that the CI didn't catch the break since witherspoon is
>> > >> > one of the systems built via it. But I'll take a look.
>> >
>> > Yea, our current automated test bucket can't catch this. (shows there
>> > is still room for improvement)
>> >
>> >
>> > >>
>> > >> Although I have had configuration files disappear when the wildcard is
>> > >> missing from the directory, the recipe normally complains about being
>> > >> unable to find the file. For a file override, if there's already that
>> > >> file - presumably then it wouldn't complain.
>> >
>> > The latter is what's happening here.
>> >
>> > >
>> > > I've fired off a witherspoon build locally from openbmc HEAD, and will
>> > > report back. Very few of the recipes were updated in this fashion so
>> > > I'm not terribly worried about a general breakage.
>> >
>> > Thanks for looking. I'm not sure if there's an easy way to find these
>> > cases.
>>
>> This specific case is difficult to catch but, it's something that's
>> avoided for most of the renames, so at least there's that. Future
>> changes though -- one will just have to be more cognizant of matching
>> output.
>>
>> >
>> > >
>> > >>
>> > >> >
>> > >> > >
>> > >> > > I haven't looked to see what all is broken yet and in how many layers,
>> > >> > > at least our phosphor-rest and phosphor-gevent, which is why the web UI stopped
>> > >> > > working on witherspoon. Not sure if any other systems are also having problems.
>> > >> > >
>> > >> > >
>> >
>
> Given it looks like I've already staged the fixes to this, we just
> need to move those forward and when it settles, it should all be
> copacetic.
Those patches do indeed fix it. Glad everything is already done.
Thanks!
More information about the openbmc
mailing list