Automated linting before master branch commit

Automated linting before master branch commit

[mikadoZero]

# Automated linting

What do people think about automating package linting?  This would be to
make sure a package is linted before it is committed to the Guix
repository master branch?

I am not familiar with what Guix uses for continuous integration /
deployment.

I do not know if automated linting is already in place.  The packages
on the master branch that have linting issues could have been committed
before automated linting was in place.

# Acknowledgment

It was brought to my attention by Tobias Geerinckx-Rice:

"... tarballs aren't guaranteed to be stable over time (GitHub can
regenerate them, changing the metadata and hence the hash, and has done
so in the past). 

They must not be used.  ..."

https://lists.gnu.org/archive/html/help-guix/2019-03/msg00096.html

# Guix master branch

There are package in the Guix repository master branch that use an
autogenerated tarball and have this lint output:

"the source URI should not be an autogenerated tarball"

# Number of packages

Estimated number of packages with autogenerated tarball:

389

This is a low estimate as it is only counts GitHub autogenerated
tarballs.

This estimate is from this command run in the Guix repository.

`grep github.*archive gnu/packages/*scm | wc -l`

Re: Automated linting before master branch commit

[Danny Milosavljevic]

[-- Attachment #1: Type: text/plain, Size: 836 bytes --]

Hi,

> What do people think about automating package linting?  This would be to
> make sure a package is linted before it is committed to the Guix
> repository master branch?

A linter by its nature will in some cases flag things which are not wrong,
otherwise one could just check the stuff while compiling in the first place.

> I am not familiar with what Guix uses for continuous integration /
> deployment.

We can manually make cuirass and/or Hydra evaluate custom branches--but it's
not automated.

Christopher Baines is working on making patch review better--see thread
"Patchwork + automated checking and testing of patches".

> Estimated number of packages with autogenerated tarball:
> 
> 389

Yeah, those should be updated.  With that many packages, writing a script
to do it would be less annoying.

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: Automated linting before master branch commit

[Ricardo Wurmus]

Danny Milosavljevic <dannym@scratchpost.org> writes:

>> Estimated number of packages with autogenerated tarball:
>> 
>> 389
>
> Yeah, those should be updated.  With that many packages, writing a script
> to do it would be less annoying.

Most of these packages were added before we realized that these
generated tarballs are not stable.  There has been an effort to remove
them.  Linting would not have prevented this.

-- 
Ricardo

Re: Automated linting before master branch commit

[mikadoZero]

Danny Milosavljevic writes:

> Hi,
>
>> What do people think about automating package linting?  This would be to
>> make sure a package is linted before it is committed to the Guix
>> repository master branch?
>
> A linter by its nature will in some cases flag things which are not wrong,
> otherwise one could just check the stuff while compiling in the first place.
>
>> I am not familiar with what Guix uses for continuous integration /
>> deployment.
>
> We can manually make cuirass and/or Hydra evaluate custom branches--but it's
> not automated.
>
> Christopher Baines is working on making patch review better--see thread
> "Patchwork + automated checking and testing of patches".

Thanks for sharing that thread.

>> Estimated number of packages with autogenerated tarball:
>> 
>> 389
>
> Yeah, those should be updated.  With that many packages, writing a script
> to do it would be less annoying.