ci for wip-haskell-updates

ci for wip-haskell-updates

[Robert Vollmert]

Hi all,

just pushed my first commit to the new branch wip-haskell-updates,
yay!

Now I was wondering how to get CI to test that. From a trip through
the sources, the following would appear to work:

In guix/gnu/ci.scm:
  Define %haskell-packages, e.g. as all packages defined in
  gnu/packages/haskell.scm together with anything that uses the
  haskell build system.
  Add a corresponding subset ‘haskell to hydra-jobs.
maintenance/modules/sysadmin/services.scm:
  Add a cuirass spec referencing gnu-system.scm and the haskell
  subset.

Is that a reasonable way to go? Should I send patches to do that?

(By the way there’s a branch wip-haskell that hasn’t been touched
in a while, should we delete it?)

Cheers
Robert

Re: ci for wip-haskell-updates

[Ricardo Wurmus]

Hi Robert,

> just pushed my first commit to the new branch wip-haskell-updates,
> yay!

Yay!

> Now I was wondering how to get CI to test that.

Adding branches is currently done manually.  I’ll just update the
table of branches to build on ci.guix.gnu.org in a few minutes.

> (By the way there’s a branch wip-haskell that hasn’t been touched
> in a while, should we delete it?)

Looks like I last touched it.  It should probably be deleted.

-- 
Ricardo

Re: ci for wip-haskell-updates

[Robert Vollmert]

Hi Ricardo,

> On 22. Jul 2019, at 21:01, Ricardo Wurmus <rekado@elephly.net> wrote:
> 
>> Now I was wondering how to get CI to test that.
> 
> Adding branches is currently done manually.  I’ll just update the
> table of branches to build on ci.guix.gnu.org in a few minutes.

nice, thanks!

The current status of the evaluation of wip-haskell-updates is quite
confusing. There are a number of jobs listed as “Scheduled” with a red
X mark at the front. Is this a bug?

>> (By the way there’s a branch wip-haskell that hasn’t been touched
>> in a while, should we delete it?)
> 
> Looks like I last touched it.  It should probably be deleted.

I tried:

$ git push guix :wip-haskell
remote: fatal: bad object 0000000000000000000000000000000000000000
remote: Sending notification emails to: guix-commits@gnu.org
^C

I have a bad feeling about the fact that it hung sending notification
emails though — hope this didn’t spam guix-commits…

Robert

Re: ci for wip-haskell-updates

[Robert Vollmert]

> On 22. Jul 2019, at 21:50, Robert Vollmert <rob@vllmrt.net> wrote:
> 
> Hi Ricardo,
> 
>> On 22. Jul 2019, at 21:01, Ricardo Wurmus <rekado@elephly.net> wrote:
>> 
>>> Now I was wondering how to get CI to test that.
>> 
>> Adding branches is currently done manually.  I’ll just update the
>> table of branches to build on ci.guix.gnu.org in a few minutes.
> 
> nice, thanks!
> 
> The current status of the evaluation of wip-haskell-updates is quite
> confusing. There are a number of jobs listed as “Scheduled” with a red
> X mark at the front. Is this a bug?

I think I found what causes this: Those jobs were actually in state
“started”, which fell through to failed.

The patch below should complete the list, and make falling through
noticeable:

diff --git a/src/cuirass/templates.scm b/src/cuirass/templates.scm
index ab1b85c..71183b2 100644
--- a/src/cuirass/templates.scm
+++ b/src/cuirass/templates.scm
@@ -130,27 +130,40 @@
   (define status (assq-ref build #:status))
   (define display-status
     (cond
-     ((= (build-status succeeded) status)
-      `(span (@ (class "oi oi-check text-success")
-                (title "Succeeded"))
-             " Success"))
      ((= (build-status scheduled) status)
       `(span (@ (class "oi oi-clock text-warning")
                 (title "Scheduled")
                 (aria-hidden "true"))
              " Scheduled"))
-     ((= (build-status canceled) status)
-      `(span (@ (class "oi oi-question-mark text-warning")
-                (title "Canceled"))
-             " Canceled"))
+     ((= (build-status started) status)
+      `(span (@ (class "oi oi-clock text-warning")
+                (title "Started")
+                (aria-hidden "true"))
+             " Started"))
+     ((= (build-status succeeded) status)
+      `(span (@ (class "oi oi-check text-success")
+                (title "Succeeded"))
+             " Success"))
+     ((= (build-status failed) status)
+      `(span (@ (class "oi oi-x text-danger")
+                (title "Failed"))
+             " Failed"))
      ((= (build-status failed-dependency) status)
       `(span (@ (class "oi oi-warning text-danger")
                 (title "Dependency failed"))
              " Dependency failed"))
+     ((= (build-status failed-other) status)
+      `(span (@ (class "oi oi-warning text-danger")
+                (title "Other failure"))
+             "  failed"))
+     ((= (build-status canceled) status)
+      `(span (@ (class "oi oi-question-mark text-warning")
+                (title "Canceled"))

Re: ci for wip-haskell-updates

[Robert Vollmert]

> On 22. Jul 2019, at 22:44, Robert Vollmert <rob@vllmrt.net> wrote:
> 
> 
> 
>> On 22. Jul 2019, at 21:50, Robert Vollmert <rob@vllmrt.net> wrote:
>> 
>> Hi Ricardo,
>> 
>>> On 22. Jul 2019, at 21:01, Ricardo Wurmus <rekado@elephly.net> wrote:
>>> 
>>>> Now I was wondering how to get CI to test that.
>>> 
>>> Adding branches is currently done manually.  I’ll just update the
>>> table of branches to build on ci.guix.gnu.org in a few minutes.
>> 
>> nice, thanks!
>> 
>> The current status of the evaluation of wip-haskell-updates is quite
>> confusing. There are a number of jobs listed as “Scheduled” with a red
>> X mark at the front. Is this a bug?
> 
> I think I found what causes this: Those jobs were actually in state
> “started”, which fell through to failed.
> 
> The patch below should complete the list, and make falling through
> noticeable:

Apologies, I’d truncated the patch. A dedicated “started” icon would
be nice, too, but I didn’t find anything fitting in the font.

diff --git a/src/cuirass/templates.scm b/src/cuirass/templates.scm
index ab1b85c..7fab879 100644
--- a/src/cuirass/templates.scm
+++ b/src/cuirass/templates.scm
@@ -130,27 +130,40 @@
   (define status (assq-ref build #:status))
   (define display-status
     (cond
-     ((= (build-status succeeded) status)
-      `(span (@ (class "oi oi-check text-success")
-                (title "Succeeded"))
-             " Success"))
      ((= (build-status scheduled) status)
       `(span (@ (class "oi oi-clock text-warning")
                 (title "Scheduled")
                 (aria-hidden "true"))
              " Scheduled"))
-     ((= (build-status canceled) status)
-      `(span (@ (class "oi oi-question-mark text-warning")
-                (title "Canceled"))
-             " Canceled"))
+     ((= (build-status started) status)
+      `(span (@ (class "oi oi-clock text-warning")
+                (title "Started")
+                (aria-hidden "true"))
+             " Started"))
+     ((= (build-status succeeded) status)
+      `(span (@ (class "oi oi-check text-success")
+                (title "Succeeded"))
+             " Success"))
+     ((= (build-status failed) status)
+      `(span (@ (class "oi oi-x text-danger")
+                (title "Failed"))
+             " Failed"))
      ((= (build-status failed-dependency) status)
       `(span (@ (class "oi oi-warning text-danger")
                 (title "Dependency failed"))
              " Dependency failed"))
+     ((= (build-status failed-other) status)
+      `(span (@ (class "oi oi-warning text-danger")
+                (title "Other failure"))
+             " Other failure"))
+     ((= (build-status canceled) status)
+      `(span (@ (class "oi oi-question-mark text-warning")
+                (title "Canceled"))
+             " Canceled"))
      (else
-      `(span (@ (class "oi oi-x text-danger")
-                (title "Failed"))
-             " Failed"))))
+      `(span (@ (class "oi oi-warning text-danger")
+                (title "Invalid status"))
+             " Invalid status"))))
   (define blocking-outputs
     (or (and-let* (((= (build-status failed-dependency) status))
                    (drv (false-if-exception

Re: ci for wip-haskell-updates

[Ludovic Courtès]

Hi,

Robert Vollmert <rob@vllmrt.net> skribis:

>>> The current status of the evaluation of wip-haskell-updates is quite
>>> confusing. There are a number of jobs listed as “Scheduled” with a red
>>> X mark at the front. Is this a bug?
>> 
>> I think I found what causes this: Those jobs were actually in state
>> “started”, which fell through to failed.
>> 
>> The patch below should complete the list, and make falling through
>> noticeable:
>
> Apologies, I’d truncated the patch. A dedicated “started” icon would
> be nice, too, but I didn’t find anything fitting in the font.

Good catch!

> diff --git a/src/cuirass/templates.scm b/src/cuirass/templates.scm
> index ab1b85c..7fab879 100644
> --- a/src/cuirass/templates.scm
> +++ b/src/cuirass/templates.scm
> @@ -130,27 +130,40 @@
>    (define status (assq-ref build #:status))
>    (define display-status
>      (cond
> -     ((= (build-status succeeded) status)
> -      `(span (@ (class "oi oi-check text-success")
> -                (title "Succeeded"))
> -             " Success"))

In general please try to minimize patches; in this case, I think this
hunk can remain here.

>       (else
> -      `(span (@ (class "oi oi-x text-danger")
> -                (title "Failed"))
> -             " Failed"))))
> +      `(span (@ (class "oi oi-warning text-danger")
> +                (title "Invalid status"))
> +             " Invalid status"))))

Much better!

Please push to cuirass.git (with a commit log).

Thank you!

Ludo’.

Re: ci for wip-haskell-updates

[Robert Vollmert]

On 25. Jul 2019, at 19:38, Ludovic Courtès <ludo@gnu.org> wrote:
> Robert Vollmert <rob@vllmrt.net> skribis:
> 
>>>> The current status of the evaluation of wip-haskell-updates is quite
>>>> confusing. There are a number of jobs listed as “Scheduled” with a red
>>>> X mark at the front. Is this a bug?
>>> 
>>> I think I found what causes this: Those jobs were actually in state
>>> “started”, which fell through to failed.
>>> 
>>> The patch below should complete the list, and make falling through
>>> noticeable:
>> 
>> Apologies, I’d truncated the patch. A dedicated “started” icon would
>> be nice, too, but I didn’t find anything fitting in the font.

> Much better!
> 
> Please push to cuirass.git (with a commit log).
> 
> Thank you!

I extended the patch a bit, to fix some similar places. I couldn’t test it
besides compiling, though. Submitted here:

http://debbugs.gnu.org/cgi/bugreport.cgi?bug=36874

Cheers
Robert