Re: CI, Java 17 and Javadoc

2023-01-12 Thread Rémy Maucherat
On Thu, Jan 12, 2023 at 3:03 PM Mark Thomas  wrote:
>
> On 12/01/2023 13:41, Rémy Maucherat wrote:
> > On Thu, Jan 12, 2023 at 1:19 PM Mark Thomas  wrote:
> >>
> >> I tried switching the CI over to use Java 17 last night. This exposed an
> >> unexpected Javadoc behaviour that is currently breaking the 9.0.x and
> >> 8.5.x builds.
> >>
> >> Ant is configured to run Javadoc with failonwarning="true"
> >>
> >> When running the Javadoc task with Java 17:
> >>
> >> - if source="11" (Tomcat 10.1.x and 11.0.x) no warnings are generated
> >> regarding SecurityManager use
> >>
> >> - if source="8" (Tomcat 9.0.x) warnings are generated for
> >> SecurityManager use
> >>
> >> - if source="7" (Tomcat 8.5.x) warnings are generated for
> >> SecurityManager use
> >>
> >> I don't understand why the warning generation is only generated for
> >> older source values. It would make more sense to me if it were only
> >> generated for newer values.
> >
> > Yes, ok, that looks very odd ...
> >
> >> Possible options:
> >> - only use Java 17 for 11.0.x builds
> >> - use failonwarning="false"
> >
> > Yes please. If you remember, I sent an update earlier on Javadoc 18+.
> > There will now be some warnings that will make it counterproductive
> > for us (in particular, having to have an empty constructor on
> > everything just so that it can be javadoc-ed). We should keep the
> > setting around however to see how these things evolve, maybe it would
> > be improved/fixed eventually ...
>
> Ah yes. I remember that now. I'll parameterise the setting and default
> it to false. Early indications are that the CheckStyle checks are more
> useful anyway. They have already uncovered a bunch of issues. Some are
> cosmetic but others are more significant such as whole sets of constants
> having the wrong descriptions.

Actually this javadoc validation through checkstyle looks better
indeed. It's nice you were able to find these options, I had no idea
it existed.

Rémy

> Mark
>
> >
> > Rémy
> >
> >> - something else
> >>
> >> I'll switch the CI system back to Java 11 while we discuss options. I
> >> also plan to look at CheckStyle options for Javadoc validation to see if
> >> they could be an alternative approach if we use failonwarning="false"
> >>
> >> Thoughts?
> >
> >
> >
> >> Mark
> >>
> >> -
> >> To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
> >> For additional commands, e-mail: dev-h...@tomcat.apache.org
> >>
> >
> > -
> > To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
> > For additional commands, e-mail: dev-h...@tomcat.apache.org
> >
>
> -
> To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
> For additional commands, e-mail: dev-h...@tomcat.apache.org
>

-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org



Re: CI, Java 17 and Javadoc

2023-01-12 Thread Christopher Schultz

Mark,

On 1/12/23 07:19, Mark Thomas wrote:
I tried switching the CI over to use Java 17 last night. This exposed an 
unexpected Javadoc behaviour that is currently breaking the 9.0.x and 
8.5.x builds.


Ant is configured to run Javadoc with failonwarning="true"

When running the Javadoc task with Java 17:

- if source="11" (Tomcat 10.1.x and 11.0.x) no warnings are generated
   regarding SecurityManager use

- if source="8" (Tomcat 9.0.x) warnings are generated for
   SecurityManager use

- if source="7" (Tomcat 8.5.x) warnings are generated for
   SecurityManager use

I don't understand why the warning generation is only generated for 
older source values. It would make more sense to me if it were only 
generated for newer values.


Possible options:
- only use Java 17 for 11.0.x builds
- use failonwarning="false"
- something else

>
I'll switch the CI system back to Java 11 while we discuss options. I 
also plan to look at CheckStyle options for Javadoc validation to see if 
they could be an alternative approach if we use failonwarning="false"


Thoughts?


I'm +1 for only moving-forward with Tomcat 11 for the time being. It 
makes sense to ratchet things forward at least there.


-chris

-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org



Re: CI, Java 17 and Javadoc

2023-01-12 Thread Mark Thomas

On 12/01/2023 13:41, Rémy Maucherat wrote:

On Thu, Jan 12, 2023 at 1:19 PM Mark Thomas  wrote:


I tried switching the CI over to use Java 17 last night. This exposed an
unexpected Javadoc behaviour that is currently breaking the 9.0.x and
8.5.x builds.

Ant is configured to run Javadoc with failonwarning="true"

When running the Javadoc task with Java 17:

- if source="11" (Tomcat 10.1.x and 11.0.x) no warnings are generated
regarding SecurityManager use

- if source="8" (Tomcat 9.0.x) warnings are generated for
SecurityManager use

- if source="7" (Tomcat 8.5.x) warnings are generated for
SecurityManager use

I don't understand why the warning generation is only generated for
older source values. It would make more sense to me if it were only
generated for newer values.


Yes, ok, that looks very odd ...


Possible options:
- only use Java 17 for 11.0.x builds
- use failonwarning="false"


Yes please. If you remember, I sent an update earlier on Javadoc 18+.
There will now be some warnings that will make it counterproductive
for us (in particular, having to have an empty constructor on
everything just so that it can be javadoc-ed). We should keep the
setting around however to see how these things evolve, maybe it would
be improved/fixed eventually ...


Ah yes. I remember that now. I'll parameterise the setting and default 
it to false. Early indications are that the CheckStyle checks are more 
useful anyway. They have already uncovered a bunch of issues. Some are 
cosmetic but others are more significant such as whole sets of constants 
having the wrong descriptions.


Mark



Rémy


- something else

I'll switch the CI system back to Java 11 while we discuss options. I
also plan to look at CheckStyle options for Javadoc validation to see if
they could be an alternative approach if we use failonwarning="false"

Thoughts?





Mark

-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org



-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org



-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org



Re: CI, Java 17 and Javadoc

2023-01-12 Thread Rémy Maucherat
On Thu, Jan 12, 2023 at 1:19 PM Mark Thomas  wrote:
>
> I tried switching the CI over to use Java 17 last night. This exposed an
> unexpected Javadoc behaviour that is currently breaking the 9.0.x and
> 8.5.x builds.
>
> Ant is configured to run Javadoc with failonwarning="true"
>
> When running the Javadoc task with Java 17:
>
> - if source="11" (Tomcat 10.1.x and 11.0.x) no warnings are generated
>regarding SecurityManager use
>
> - if source="8" (Tomcat 9.0.x) warnings are generated for
>SecurityManager use
>
> - if source="7" (Tomcat 8.5.x) warnings are generated for
>SecurityManager use
>
> I don't understand why the warning generation is only generated for
> older source values. It would make more sense to me if it were only
> generated for newer values.

Yes, ok, that looks very odd ...

> Possible options:
> - only use Java 17 for 11.0.x builds
> - use failonwarning="false"

Yes please. If you remember, I sent an update earlier on Javadoc 18+.
There will now be some warnings that will make it counterproductive
for us (in particular, having to have an empty constructor on
everything just so that it can be javadoc-ed). We should keep the
setting around however to see how these things evolve, maybe it would
be improved/fixed eventually ...

Rémy

> - something else
>
> I'll switch the CI system back to Java 11 while we discuss options. I
> also plan to look at CheckStyle options for Javadoc validation to see if
> they could be an alternative approach if we use failonwarning="false"
>
> Thoughts?



> Mark
>
> -
> To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
> For additional commands, e-mail: dev-h...@tomcat.apache.org
>

-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org



CI, Java 17 and Javadoc

2023-01-12 Thread Mark Thomas
I tried switching the CI over to use Java 17 last night. This exposed an 
unexpected Javadoc behaviour that is currently breaking the 9.0.x and 
8.5.x builds.


Ant is configured to run Javadoc with failonwarning="true"

When running the Javadoc task with Java 17:

- if source="11" (Tomcat 10.1.x and 11.0.x) no warnings are generated
  regarding SecurityManager use

- if source="8" (Tomcat 9.0.x) warnings are generated for
  SecurityManager use

- if source="7" (Tomcat 8.5.x) warnings are generated for
  SecurityManager use

I don't understand why the warning generation is only generated for 
older source values. It would make more sense to me if it were only 
generated for newer values.


Possible options:
- only use Java 17 for 11.0.x builds
- use failonwarning="false"
- something else

I'll switch the CI system back to Java 11 while we discuss options. I 
also plan to look at CheckStyle options for Javadoc validation to see if 
they could be an alternative approach if we use failonwarning="false"


Thoughts?

Mark

-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org