On 13 March 2018 at 16:15, Ryan Schmidt wrote:
> On Mar 13, 2018, at 07:50, Rainer Müller wrote:
>> On 2018-03-13 13:27, Ryan Schmidt wrote:
>>> On Mar 13, 2018, at 02:43, Mojca Miklavec wrote:
>>>> On 13 March 2018 at 05:07, Ryan Schmidt wrote:
>>>>> Maybe we want to exclude some of them...
>>>> I would probably only keep the port watchers.
>>> Surely we want to keep port builders, not port watchers?
>> Hm, but we run multiple builders per commit.
> Oh right. It just seems like if a commit builds just one port, then its
> annoying to have to click on the portwatcher details, only to have to click
> another link to get the portbuilder details to see what happened. And if more
> than one port builds, then once you get to portwatcher and you have multiple
> logs to click, but there's no indication of which log is for which port.
We can probably change context and show the port name, we just need to
keep in mind that GitHub might then show hundreds of lines and at some
point we'll no longer be able to see some of them anyway.
> At the moment, portwatcher can also be marked as failed because of a failed
> mirroring job, and that might be annoying. Currently, for example, any port
> depending indirectly on mesa will fail to mirror, because its python26
> variant is broken. This will clear itself up once distfiles are mirrored
> first, before portwatcher.
>> I guess the context would
>> then have to include the portname or the results would be overwritten again.
> That could work. I kind of don't like the tiny little box that GitHub
> restricts the status information to. It's barely wide enough to contain the
> context string, and not wide enough to show the build status string. But if
> we make the context shorter, by removing "ports-" and "-portbuilder" and
> maybe even "-x86_64" we'd have more room for a port name.
We could also shorten the "buildbot" label itself.