use of git-fetch

use of git-fetch

[Federico Beffa]

[-- Attachment #1: Type: text/plain, Size: 576 bytes --]

Hi,

I'm trying to package some emacs libraries that appear not to have a
GNU style installation procedure. Given that MELPA is unusable, I'm
trying to use 'git-fetch'.

The package 'emacs-dash' (see attachment) works as expected. However,
the package 'emacs-s' also downloads 'dash' and not 's'. I cloned with
git the two repositories with the given URLs and they definitely are
different.

Am I doing something wrong?

Thanks,
Fede

P.S. The hash of 'emacs-s' is wrong. It is a copy of the one for
'emacs-dash' and I will change it once I manage to download the right
repo.

[-- Attachment #2: git-fetch-dash-s.scm --]
[-- Type: text/x-scheme, Size: 1462 bytes --]

(define-public emacs-dash
  (package
    (name "emacs-dash")
    (version "eadb2651bb")
    (source (origin
              (method git-fetch)
              (uri (git-reference
                    (commit version)
                    (url "https://github.com/magnars/dash.el.git")))
              (sha256
               (base32
                "06mc7kh3fzdh2mqkyynjnp0xpv30yfaiik8bqv8z5b6hldji3cky"))))
    (build-system emacs-build-system)
    (arguments
     `(#:phases
       (modify-phases %standard-phases
         (add-before 'install 'check
                     (lambda _
                       (zero? (system* "./run-tests.sh")))))))
    (home-page "https://github.com/magnars/dash.el")
    (synopsis "Modern list library for Emacs")
    (description "An Emacs library providing a modern list API.")
    (license license:gpl3+)))

(define-public emacs-s
  (package
    (name "emacs-s")
    (version "e59915ec6f")
    (source (origin
              (method git-fetch)
              (uri (git-reference
                    (commit version)
                    (url "https://github.com/magnars/s.el.git")))
              (sha256
               (base32
                "06mc7kh3fzdh2mqkyynjnp0xpv30yfaiik8bqv8z5b6hldji3cky"))))
    (build-system emacs-build-system)
    (home-page "https://github.com/magnars/s.el")
    (synopsis "Emacs string manipulation library.")
    (description "An Emacs library for manipulating strings.")
    (license license:gpl3+)))

Re: use of git-fetch

[Thompson, David]

On Wed, Jul 22, 2015 at 12:32 PM, Federico Beffa <beffa@ieee.org> wrote:
> Hi,
>
> I'm trying to package some emacs libraries that appear not to have a
> GNU style installation procedure. Given that MELPA is unusable, I'm
> trying to use 'git-fetch'.
>
> The package 'emacs-dash' (see attachment) works as expected. However,
> the package 'emacs-s' also downloads 'dash' and not 's'. I cloned with
> git the two repositories with the given URLs and they definitely are
> different.
>
> Am I doing something wrong?

Are you deleting the '.git' directory before computing the hash of the
directory?  That is a necessary step.

Consider this a call for someone to extend 'guix download' to work
with Git repositories so people don't have to do this tedious process.
Any takers?

- Dave

Re: use of git-fetch

[Ludovic Courtès]

Federico Beffa <beffa@ieee.org> skribis:

>               (sha256
>                (base32
>                 "06mc7kh3fzdh2mqkyynjnp0xpv30yfaiik8bqv8z5b6hldji3cky"))))

[...]

>               (sha256
>                (base32
>                 "06mc7kh3fzdh2mqkyynjnp0xpv30yfaiik8bqv8z5b6hldji3cky"))))

Both recipes are telling that they use the same source.  So the daemon
cleverly saves one download since it already has the thing with that
hash on disk.  See?  :-)

HTH,
Ludo’.

Re: use of git-fetch

[Federico Beffa]

On Wed, Jul 22, 2015 at 7:00 PM, Thompson, David
<dthompson2@worcester.edu> wrote:
> On Wed, Jul 22, 2015 at 12:32 PM, Federico Beffa <beffa@ieee.org> wrote:
>> Hi,
>>
>> I'm trying to package some emacs libraries that appear not to have a
>> GNU style installation procedure. Given that MELPA is unusable, I'm
>> trying to use 'git-fetch'.
>>
>> The package 'emacs-dash' (see attachment) works as expected. However,
>> the package 'emacs-s' also downloads 'dash' and not 's'. I cloned with
>> git the two repositories with the given URLs and they definitely are
>> different.
>>
>> Am I doing something wrong?
>
> Are you deleting the '.git' directory before computing the hash of the
> directory?  That is a necessary step.

Thanks for the suggestion! Probably I didn't explain well... see the
reply to Ludovic.

>
> Consider this a call for someone to extend 'guix download' to work
> with Git repositories so people don't have to do this tedious process.
> Any takers?
>
> - Dave

Re: use of git-fetch

[Federico Beffa]

On Wed, Jul 22, 2015 at 11:10 PM, Ludovic Courtès <ludo@gnu.org> wrote:
> Federico Beffa <beffa@ieee.org> skribis:
>
>>               (sha256
>>                (base32
>>                 "06mc7kh3fzdh2mqkyynjnp0xpv30yfaiik8bqv8z5b6hldji3cky"))))
>
> [...]
>
>>               (sha256
>>                (base32
>>                 "06mc7kh3fzdh2mqkyynjnp0xpv30yfaiik8bqv8z5b6hldji3cky"))))
>
> Both recipes are telling that they use the same source.  So the daemon
> cleverly saves one download since it already has the thing with that
> hash on disk.  See?  :-)

Thanks for the reply!

I thought about this and, before posting, I also tried with an hash
where the first character was changed from '0' to '6'. It gave the
same result, even after deleting any existing 'emacs-dash' and
'emacs-s' derivation in the store. In my understanding that shouldn't
happen. Is that correct?

By changing the hash to a totally different one (a string of '6's) it
downloads the right repo. As expected, it complains about the hash and
gives me the right one which I can copy into the package recipe.
(That's The Lazy work-flow :-).)

Regards,
Fede

>
> HTH,
> Ludo’.

Re: use of git-fetch

[Ludovic Courtès]

Federico Beffa <beffa@ieee.org> skribis:

> On Wed, Jul 22, 2015 at 11:10 PM, Ludovic Courtès <ludo@gnu.org> wrote:
>> Federico Beffa <beffa@ieee.org> skribis:
>>
>>>               (sha256
>>>                (base32
>>>                 "06mc7kh3fzdh2mqkyynjnp0xpv30yfaiik8bqv8z5b6hldji3cky"))))
>>
>> [...]
>>
>>>               (sha256
>>>                (base32
>>>                 "06mc7kh3fzdh2mqkyynjnp0xpv30yfaiik8bqv8z5b6hldji3cky"))))
>>
>> Both recipes are telling that they use the same source.  So the daemon
>> cleverly saves one download since it already has the thing with that
>> hash on disk.  See?  :-)
>
> Thanks for the reply!
>
> I thought about this and, before posting, I also tried with an hash
> where the first character was changed from '0' to '6'. It gave the
> same result, even after deleting any existing 'emacs-dash' and
> 'emacs-s' derivation in the store. In my understanding that shouldn't
> happen. Is that correct?

Right, but we’d have to check the exact sequence of actions that you
took.

> By changing the hash to a totally different one (a string of '6's) it
> downloads the right repo. As expected, it complains about the hash and
> gives me the right one which I can copy into the package recipe.
> (That's The Lazy work-flow :-).)

Heh.  :-)

It’s a good idea to clone by hand though and check the signatures if
there are signed tags or commits (“trust on first clone”), but I’m
afraid this is not commonplace.

Ludo’.

Re: use of git-fetch

[Federico Beffa]

On Fri, Jul 24, 2015 at 12:50 AM, Ludovic Courtès <ludo@gnu.org> wrote:
> Federico Beffa <beffa@ieee.org> skribis:
>
>> On Wed, Jul 22, 2015 at 11:10 PM, Ludovic Courtès <ludo@gnu.org> wrote:
>>> Federico Beffa <beffa@ieee.org> skribis:
>>>
>>>>               (sha256
>>>>                (base32
>>>>                 "06mc7kh3fzdh2mqkyynjnp0xpv30yfaiik8bqv8z5b6hldji3cky"))))
>>>
>>> [...]
>>>
>>>>               (sha256
>>>>                (base32
>>>>                 "06mc7kh3fzdh2mqkyynjnp0xpv30yfaiik8bqv8z5b6hldji3cky"))))
>>>
>>> Both recipes are telling that they use the same source.  So the daemon
>>> cleverly saves one download since it already has the thing with that
>>> hash on disk.  See?  :-)
>>
>> Thanks for the reply!
>>
>> I thought about this and, before posting, I also tried with an hash
>> where the first character was changed from '0' to '6'. It gave the
>> same result, even after deleting any existing 'emacs-dash' and
>> 'emacs-s' derivation in the store. In my understanding that shouldn't
>> happen. Is that correct?
>
> Right, but we’d have to check the exact sequence of actions that you
> took.

Basically I did:

1. write the package for emacs-dash
2. guix build emacs-dash
3. copy emacs-dash and make emacs-s out of it, keeping the hash unchanged
4. guix build emacs-s

Boom: in the generated derivation I find dash.el instead of s.el.

5. change the first char of the hash 0->6

Boom: same.

6. change the full hash to a sequence of 6s.

OK, now I get the expected result.

7. change hash, rebuild and inspect.

Regards,
Fede

Re: use of git-fetch

[Alex Kost]

Federico Beffa (2015-07-22 19:32 +0300) wrote:

> Hi,
>
> I'm trying to package some emacs libraries that appear not to have a
> GNU style installation procedure. Given that MELPA is unusable, I'm
> trying to use 'git-fetch'.

May I ask, why are you trying to use 'git-fetch'?  There are releases
for both "dash" and "s":

https://github.com/magnars/dash.el/releases
https://github.com/magnars/s.el/releases

-- 
Alex

Re: use of git-fetch

[Ludovic Courtès]

Federico Beffa <beffa@ieee.org> skribis:

> Basically I did:
>
> 1. write the package for emacs-dash
> 2. guix build emacs-dash
> 3. copy emacs-dash and make emacs-s out of it, keeping the hash unchanged
> 4. guix build emacs-s
>
> Boom: in the generated derivation I find dash.el instead of s.el.
>
> 5. change the first char of the hash 0->6
>
> Boom: same.

So you’re saying the two hashes that differed by one digit were
considered the same, essentially.

Could you check if there’s a simple way to reproduce it?

Thanks,
Ludo’.