On Mon, Jul 14, 2014 at 4:52 PM, Frank Shearar <frank.shearar at gmail.com>
wrote:

> On 14 July 2014 22:36, Chris Muller <asqueaker at gmail.com> wrote:
> > Why not just cleanly list the name of the failing test and nothing else?
> > Tests are only understood and debugged in the running image, not from a
> > cryptic message reported on the build website.
>
> I realise that _you_ don't use CI to debug tests. I do. I find value
> in these error messages, especially since I don't have a Linux
> machine, and as it happens, a test (see Tests-fbs.301) fails in
> Spur+Linux but passes in Spur+Windows.

>  It certainly can't be fixed
> > and committed without running the test in the image anyway.
>
> Yes, and so what?
>

Don't worry, if you find them enormously valuable then I want you to have
them.  I'm just trying to understand the CI use cases with and without the
debugging messages.  Seems like, either way, the process is the same..  (go
into image, run, fix, run, commit -- where did the messages help?).

> So why do we want to see that debugging noise in the build-server reports?
>
> How is a decent error message noise? I find ENORMOUS value in these
> error messages. How could I possibly report a problem on a Linux box
> to Eliot otherwise?
>

The problem was still reported by CI from a Linux run.  When do you consume
these messages and how putting them to use before getting into an image
with full debugger?  It's not for trying to debug an OS-specific failure
from a different OS, even I know better..  :)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20140714/59bb96e0/attachment.htm