Package Update

josef Reidinger jreidinger at suse.com
Mon Mar 15 21:15:17 UTC 2021


On Mon, 15 Mar 2021 14:05:36 -0300 (BRT)
Luiz Angelo Daros de Luca <luizluca at tre-sc.jus.br> wrote:

> Hi Josef,
> 
> Yes, a new validator is a right move. The problem is that it blocks existing workflows,
> although with errors, with a generic error message. It would be fantastic if xmllint or 
> ruby rexml or nokorigi relaxng validators could be improved to give a better user feedback.
> It would also help the user to decide to ignore the existing autoyast errors or not.

I agree. We consider all options but ruby REXML implementation is not complete and does not work with our schema.
and nokogiri and xmllint is both based on libxml2 and it is long time issue in it. See https://gitlab.gnome.org/GNOME/libxml2/-/issues/61 where one of bug reports are from year 2003.

> 
> It is the first time I read jing will provide me with better error messages. I'll try it.
> Most yast users are not familiar with relaxng to decide the better validator. Yast only says
> it is an alternative solution, just like xmllint. It should be more clear that it is better.
> Anyway, I still thing it would be better to improve existing validators as they can be used
> inside installer with little extra weight. Nobody wants to bring java into the installer.

I agree, but as it is long standing issue in libxml2 I have higher hope that rexml implementation will be complete and with better errors.

> 
> It's good to know you can ignore it, but it should be easier to do, like a "Ignore", "Cancel",
> "Retry (redownloading)".

That sounds good for me. I will suggest it in team planning.

Thanks for feedback.

> 
> Thanks,

Josef

> 
> --
> Luiz Angelo Daros de Luca
> Tribunal Regional Eleitoral de Santa Catarina
> STI/CSIT/Seção de Administração de Redes e de Servidores
> e-mail: luizluca at tre-sc.jus.br
> jabber: luizluca at tre-sc.gov.br
> fone: +55 48 3251-7458
> 
> ----- Mensagem original -----
> > De: "josef Reidinger" <jreidinger at suse.com>
> > Para: "sle-beta" <sle-beta at lists.suse.com>
> > Enviadas: Segunda-feira, 15 de março de 2021 7:00:57
> > Assunto: Re: Package Update  
> 
> > On Fri, 12 Mar 2021 20:33:52 -0300 (BRT)
> > Luiz Angelo Daros de Luca <luizluca at tre-sc.jus.br> wrote:
> >   
> >> The new "validator" will still give us a lot of pain.  
> > 
> > Hi Luiz,
> > main goal of new validator is early catch of issues in autoyast profile so you
> > do not waste time with invalid profile. As you can read in previous reply you
> > can also disable it.
> > Now let me give a bit inside what happens here:
> >   
> >> 
> >> It would be nice to test every example config:
> >> 
> >> <?xml version="1.0" encoding="UTF-8"?>
> >> <!DOCTYPE profile>
> >> <profile xmlns="http://www.suse.com/1.0/yast2ns"  
> >> xmlns:config="http://www.suse.com/1.0/configns">  
> >> <general>
> >> <semi-automatic config:type="list">
> >> <semi-automatic_entry>networking</semi-automatic_entry>
> >> <semi-automatic_entry>scc</semi-automatic_entry>
> >> <semi-automatic_entry>partitioning</semi-automatic_entry>
> >> </semi-automatic>
> >> </general>
> >> </profile>
> >> https://documentation.suse.com/sles/15-SP2/single-html/SLES-autoyast/index.html#CreateProfile-General-semi-automatic
> >>   
> > 
> > It is in fact bug in documentation, as it was always "module" as child element,
> > but both options works. Why? Because parser for child of element that has type
> > "list" ignores name and just parse its value.
> >   
> >> xmllint returns a very limited diagnostic answer:
> >> 
> >> xmllint --noout --relaxng /usr/share/YaST2/schema/autoyast/rng/profile.rng test
> >> test :4: element general: Relax-NG validity error : Element profile has extra
> >> content: general
> >> test fails to validate  
> > 
> > I think even in error report suggestion is to use jing as validator to get
> > better error message. Sadly we cannot use it directly in installer as it is
> > java based and we do not have space/RAM for running it in installation.
> >   
> >> 
> >> Reading profile.rng, it looks like "semi-automatic_entry" does not exist anymore
> >> and now it is "module".  
> > 
> > As said before it was that way always in schema, just bug in documentation.
> > Good part is that we are now discussion to allow both in schema as both works
> > and it is already in documentation.  
> >> 
> >> We have autoyast syntax changes, an enforced syntax validator, no updated doc
> >> and a very limited diagnostic answer.
> >> I'm doing some fixes in my autoyast profiles based mostly on trial and error and
> >> manual bisect. It's taking a lot of time,
> >> specially for items deep inside the tree.  
> > 
> > I suggest to just use yast2-schema package and run jing on your profile to get
> > quick xml schema validation as profile the fastest way to adapt autoyast
> > profile.
> > 
> > Josef
> >   
> >> 
> >> Regards,  
> 



More information about the sle-beta mailing list