A pleasant low-hanging fruit

A pleasant low-hanging fruit

[Ludovic Courtès]

Hi!

I just had a brain wave and couldn’t resist: commit ef010c0 changes
‘guix package --install’ such that, when installing a GNU package, it
automatically reports the availability of a new upstream version.

So, for instance, if you run it today:

--8<---------------cut here---------------start------------->8---
$ ./pre-inst-env guix package -i idutils -n
[nothing special here]

$ ./pre-inst-env guix package -i gettext -n
gnu/packages/gettext.scm:27:3: note: using gettext-0.18.1.1 but gettext-0.18.2.1 is available upstream
--8<---------------cut here---------------end--------------->8---

Isn’t it cool?  :-)

This is fast (~1 sec.), except for packages like GCC with many versions
and a deep FTP tree structure (~1 m!).  It’s a rare operation, so it
doesn’t have to be blazingly fast, but definitely faster than for GCC.
Room for improvement.

The next step is to offer the option to automatically install from the
upstream tarball, by downloading it and substituting it in the ‘package’
record.  The only bit that is missing for this is a GPG keyring of GNU
maintainer keys so that we can automatically authenticate tarballs (I’ve
asked Karl about this.)

The next next step is some fancy stuff that, upon successful
installation of the new upstream version, offers to automatically submit
a patch for inclusion in Guix.  This way, every user would end up
participating in the maintenance of the distro in a simple yet effective
fashion.  :-)

Ludo’.

Re: A pleasant low-hanging fruit

[Alex Sassmannshausen]

The mind boggles... Awesome.


Alex

On Tue,   5 Mar 2013, 19:47:21 GMT, Ludovic Courtès <ludo@gnu.org> wrote:

> Hi!
> 
> I just had a brain wave and couldn’t resist: commit ef010c0 changes
> ‘guix package --install’ such that, when installing a GNU package, it
> automatically reports the availability of a new upstream version.
> 
> So, for instance, if you run it today:
> 
> --8<---------------cut here---------------start------------->8---
> $ ./pre-inst-env guix package -i idutils -n
> [nothing special here]
> 
> $ ./pre-inst-env guix package -i gettext -n
> gnu/packages/gettext.scm:27:3: note: using gettext-0.18.1.1 but
> gettext-0.18.2.1 is available upstream --8<---------------cut
> here---------------end--------------->8---
> 
> Isn’t it cool?   :-)
> 
> This is fast (~1 sec.), except for packages like GCC with many versions
> and a deep FTP tree structure (~1 m!).   It’s a rare operation, so it
> doesn’t have to be blazingly fast, but definitely faster than for GCC.
> Room for improvement.
> 
> The next step is to offer the option to automatically install from the
> upstream tarball, by downloading it and substituting it in the ‘package’
> record.   The only bit that is missing for this is a GPG keyring of GNU
> maintainer keys so that we can automatically authenticate tarballs (I’ve
> asked Karl about this.)
> 
> The next next step is some fancy stuff that, upon successful
> installation of the new upstream version, offers to automatically submit
> a patch for inclusion in Guix.   This way, every user would end up
> participating in the maintenance of the distro in a simple yet effective
> fashion.   :-)
> 
> Ludo’.
> 

Re: A pleasant low-hanging fruit

[Cyril Roelandt]

On 03/05/2013 08:47 PM, Ludovic Courtès wrote:
> Hi!
>
> I just had a brain wave and couldn’t resist: commit ef010c0 changes
> ‘guix package --install’ such that, when installing a GNU package, it
> automatically reports the availability of a new upstream version.
>
> So, for instance, if you run it today:
>
> --8<---------------cut here---------------start------------->8---
> $ ./pre-inst-env guix package -i idutils -n
> [nothing special here]
>
> $ ./pre-inst-env guix package -i gettext -n
> gnu/packages/gettext.scm:27:3: note: using gettext-0.18.1.1 but gettext-0.18.2.1 is available upstream
> --8<---------------cut here---------------end--------------->8---
>

This works well with GNU packages, but breaks other packages:

$ ./pre-inst-env ./scripts/guix package -i w3m
Backtrace:
In ice-9/boot-9.scm:
  149: 13 [catch #t #<catch-closure 98682c0> ...]
  157: 12 [#<procedure 98318c0 ()>]
In unknown file:
    ?: 11 [catch-closure]
In ice-9/boot-9.scm:
   63: 10 [call-with-prompt prompt0 ...]
In ice-9/eval.scm:
  407: 9 [eval # #]
In ice-9/boot-9.scm:
2111: 8 [save-module-excursion #<procedure 981e080 at 
ice-9/boot-9.scm:3646:3 ()>]
3653: 7 [#<procedure 981e080 at ice-9/boot-9.scm:3646:3 ()>]
In unknown file:
    ?: 6 [load-compiled/vm 
"/home/cyril/.cache/guile/ccache/2.0-LE-4-2.0/home/cyril/guix/scripts/guix.go"]
In ice-9/boot-9.scm:
  149: 5 [catch srfi-34 #<procedure 9c39f10 at guix/ui.scm:104:2 ()> ...]
  157: 4 [#<procedure 9c101e0 ()>]
In guix/scripts/package.scm:
  722: 3 [#<procedure 9c24660 at guix/scripts/package.scm:715:8 ()>]
  560: 2 [process-actions (# #)]
In srfi/srfi-1.scm:
  664: 1 [filter-map #<procedure 9c34270 at 
guix/scripts/package.scm:588:37 (expr)> ...]
In guix/scripts/package.scm:
  592: 0 [#<procedure 9c34270 at guix/scripts/package.scm:588:37 (expr)> #]

guix/scripts/package.scm:592:39: In procedure #<procedure 9c34270 at 
guix/scripts/package.scm:588:37 (expr)>:
guix/scripts/package.scm:592:39: In procedure string-split: Wrong type 
argument in position 1 (expecting string): #f


Cyril.

Re: A pleasant low-hanging fruit

[Andreas Enge]

Am Sonntag, 10. März 2013 schrieb Cyril Roelandt:
> This works well with GNU packages, but breaks other packages:

Ah, I noticed the same thing with a new package; I thought it was a mistake 
in my recipe...

Andreas

Re: A pleasant low-hanging fruit

[Ludovic Courtès]

Cyril Roelandt <tipecaml@gmail.com> skribis:

> On 03/05/2013 08:47 PM, Ludovic Courtès wrote:
>> Hi!
>>
>> I just had a brain wave and couldn’t resist: commit ef010c0 changes
>> ‘guix package --install’ such that, when installing a GNU package, it
>> automatically reports the availability of a new upstream version.
>>
>> So, for instance, if you run it today:
>>
>> --8<---------------cut here---------------start------------->8---
>> $ ./pre-inst-env guix package -i idutils -n
>> [nothing special here]
>>
>> $ ./pre-inst-env guix package -i gettext -n
>> gnu/packages/gettext.scm:27:3: note: using gettext-0.18.1.1 but gettext-0.18.2.1 is available upstream
>> --8<---------------cut here---------------end--------------->8---
>>
>
> This works well with GNU packages, but breaks other packages:
>
> $ ./pre-inst-env ./scripts/guix package -i w3m
> Backtrace:

I can’t reproduce the problem with v0.1-237-gef86c39.

Could you check whether it happens with current master, and if it does,
add some ‘pk’ calls in there to see what happens?

Ludo’.

Re: A pleasant low-hanging fruit

[Cyril Roelandt]

On 03/14/2013 02:44 PM, Ludovic Courtès wrote:
> I can’t reproduce the problem with v0.1-237-gef86c39.
>
> Could you check whether it happens with current master, and if it does,
> add some ‘pk’ calls in there to see what happens?
>

I can confirm this happens with current master.

I don't really know what you mean by "add 'pk' calls", or how to debug a 
Guile program step-by-step, could you point me to some documentation ?

Looking at the backtrace and at the code, I think the problem comes from 
"check-package-freshness". This line :

      "(when (gnu-package? package)"

might explain why the code works only for GNU packages. Isn't an 
unspecified value returned for non-GNU packages ?

Cyril.

Re: A pleasant low-hanging fruit

[Ludovic Courtès]

Cyril Roelandt <tipecaml@gmail.com> skribis:

> guix/scripts/package.scm:592:39: In procedure #<procedure 9c34270 at
> guix/scripts/package.scm:588:37 (expr)>:
> guix/scripts/package.scm:592:39: In procedure string-split: Wrong type
> argument in position 1 (expecting string): #f

This is actually ‘string-split’ as called by ‘official-gnu-packages’,
called from ‘gnu-package?’.

The problem is with Guile <= 2.0.5, whose HTTP client did not support
chunked encoding.

Commit 993fb66 should fix that.  Can you confirm?

After that, make sure to upgrade to a more recent Guile.  :-)

Ludo’.

Re: A pleasant low-hanging fruit

[Ludovic Courtès]

Cyril Roelandt <tipecaml@gmail.com> skribis:

> On 03/14/2013 02:44 PM, Ludovic Courtès wrote:
>> I can’t reproduce the problem with v0.1-237-gef86c39.
>>
>> Could you check whether it happens with current master, and if it does,
>> add some ‘pk’ calls in there to see what happens?
>>
>
> I can confirm this happens with current master.
>
> I don't really know what you mean by "add 'pk' calls", or how to debug
> a Guile program step-by-step, could you point me to some documentation
> ?

  (pk args ...)
     Print ARGS and return the last element of ARGS.

Ludo’.

Re: A pleasant low-hanging fruit

[Cyril Roelandt]

On 03/16/2013 01:01 AM, Ludovic Courtès wrote:
> Cyril Roelandt<tipecaml@gmail.com>  skribis:
>
>> guix/scripts/package.scm:592:39: In procedure #<procedure 9c34270 at
>> guix/scripts/package.scm:588:37 (expr)>:
>> guix/scripts/package.scm:592:39: In procedure string-split: Wrong type
>> argument in position 1 (expecting string): #f
>
> This is actually ‘string-split’ as called by ‘official-gnu-packages’,
> called from ‘gnu-package?’.
>
> The problem is with Guile<= 2.0.5, whose HTTP client did not support
> chunked encoding.
>
> Commit 993fb66 should fix that.  Can you confirm?
>

It works!

Cyril.