Re: Can't install wine, curl build fails at 'check' stage

[zimoun]

Hi,

On Mon, 04 Jul 2022 at 15:40, Ludovic Courtès <ludo@gnu.org> wrote:

> Indeed, ‘guix build
> /gnu/store/v3kc0z1mx7zai6hky8kzlp171cxr2ccj-curl-7.84.0.drv’ fails with:

[...]

> Now, I don’t see where this .drv comes from.  Here’s what I see on
> x86_64:

I think it comes from a twist with i686:

--8<---------------cut here---------------start------------->8---
$ guix time-machine --commit=81cf674             \
     -- build  --no-grafts                     \
     -e '(@@ (gnu packages curl) curl-7.84.0)' \
     -s i686-linux -d
/gnu/store/v3kc0z1mx7zai6hky8kzlp171cxr2ccj-curl-7.84.0.drv
--8<---------------cut here---------------end--------------->8---


> --8<---------------cut here---------------start------------->8---
> $ guix build curl -d --no-grafts
> /gnu/store/vygnm458v0vpnpqfcgdgk7i46gy7znv3-curl-7.79.1.drv
> $ guix build curl -d 
> /gnu/store/zny336h8l7h5vjslck2f40a6cdw8mn0m-curl-7.84.0.drv
> $ guix describe
> Generation 221  Jul 03 2022 23:52:07    (current)
>   guix e069de4
>     repository URL: https://git.savannah.gnu.org/git/guix.git
>     branch: master
>     commit: e069de452a2c923868f5137421b4b6349c38d754
> --8<---------------cut here---------------end--------------->8---

Well, from [1], it is reported 81cf674, in case something had been
changed (see various fixes as aeb8c4fbea8197f5c568efab800a45aba53d4dd6
or 89fe091bb4f70c7f929c770ed5ca8e8ad52f8e09 or
c145e51844bf52eb77cdc969a0fe30a48755b29e).


Just a step by step to find the issue.  Maybe it can help.


Let check the derivations:

--8<---------------cut here---------------start------------->8---
$ guix time-machine --commit=81cf674       \
       -- build                            \
       -e '(@@ (gnu packages curl) curl)' -d
/gnu/store/381k3irnlmb9iqqwlabsk1c3xsczaijs-curl-7.84.0.drv

$ guix time-machine --commit=81cf674              \
       -- build                                   \
       -e '(@@ (gnu packages curl) curl-7.84.0)' -d
/gnu/store/381k3irnlmb9iqqwlabsk1c3xsczaijs-curl-7.84.0.drv
--8<---------------cut here---------------end--------------->8---

If we compare
/gnu/store/v3kc0z1mx7zai6hky8kzlp171cxr2ccj-curl-7.84.0.drv and
/gnu/store/381k3irnlmb9iqqwlabsk1c3xsczaijs-curl-7.84.0.drv , even the
hashes of the inputs are different.  For instance,
/gnu/store/01mqx581dqhw947rzj2x83cb4c3aafq6-bash-minimal-5.1.8.drv vs
/gnu/store/3qaj00bq9h3wnpm5nxh02zfl0sg9vmqc-bash-minimal-5.1.8.drv.

Therefore, the twist is deep. :-)


Note that, the derivation which fails uses
/gnu/store/9p23m7lh18w21qij5s0ka2kd6s5vidr0-curl-7.84.0-builder and it
reads,

--8<---------------cut here---------------start------------->8---
("source" . "/gnu/store/0g24sh9v8hqf901aj7bp0hl5swf4hfr7-curl-7.84.0.tar.xz")
--8<---------------cut here---------------end--------------->8---

when the source from the correct builder reads,

--8<---------------cut here---------------start------------->8---
("source" . "/gnu/store/67w9zrlm1xahczf55f9rd15aaqadbixj-curl-7.84.0.tar.xz")
--8<---------------cut here---------------end--------------->8---

but both have the same hash:

--8<---------------cut here---------------start------------->8---
$ guix hash                                                        \
    /gnu/store/0g24sh9v8hqf901aj7bp0hl5swf4hfr7-curl-7.84.0.tar.xz \
    /gnu/store/67w9zrlm1xahczf55f9rd15aaqadbixj-curl-7.84.0.tar.xz
1wy3p1dz18sx0mv3yqpwd47k3026bwnc0giniy0ksaqfx77hspy0
1wy3p1dz18sx0mv3yqpwd47k3026bwnc0giniy0ksaqfx77hspy0
--8<---------------cut here---------------end--------------->8---

Well, note that,

--8<---------------cut here---------------start------------->8---
$ guix gc --derivers 
/gnu/store/0g24sh9v8hqf901aj7bp0hl5swf4hfr7-curl-7.84.0.tar.xz
/gnu/store/rhkhh9scip2j24mf5lx0xcawc0ykfd05-curl-7.84.0.tar.xz.drv
--8<---------------cut here---------------end--------------->8---

which reads,

--8<---------------cut here---------------start------------->8---
Derive
([("out","/gnu/store/b6h8v89cavb5zji6ys7bzlb193s0hzf2-curl-7.84.0.tar.xz","sha256","2d118b43f547bfe5bae806d8d47b4e596ea5b25a6c1f080aef49fbcd817c5db8")]
 ,[]
 
,["/gnu/store/0mxnx8l4fgigvd7gakwdk6hc6im4wnai-disarchive-mirrors","/gnu/store/lv7dc09895a2d7cn0m6ysrdlgkhx5ld8-content-addressed-mirrors","/gnu/store/nb73zsx418kbz2p3f15pvv9spi0afmz7-mirrors"]
 ,"i686-linux","builtin:download",[]
 
,[("content-addressed-mirrors","/gnu/store/lv7dc09895a2d7cn0m6ysrdlgkhx5ld8-content-addressed-mirrors")
   
,("disarchive-mirrors","/gnu/store/0mxnx8l4fgigvd7gakwdk6hc6im4wnai-disarchive-mirrors")
   ,("impureEnvVars","http_proxy https_proxy LC_ALL LC_MESSAGES LANG COLUMNS")
   ,("mirrors","/gnu/store/nb73zsx418kbz2p3f15pvv9spi0afmz7-mirrors")
   ,("out","/gnu/store/b6h8v89cavb5zji6ys7bzlb193s0hzf2-curl-7.84.0.tar.xz")
   ,("preferLocalBuild","1")
   ,("url","\"https://curl.se/download/curl-7.84.0.tar.xz\"";)])
--8<---------------cut here---------------end--------------->8---

Tandan! :-)  The other derivation (for x86_64) reads,

--8<---------------cut here---------------start------------->8---
Derive
([("out","/gnu/store/b6h8v89cavb5zji6ys7bzlb193s0hzf2-curl-7.84.0.tar.xz","sha256","2d118b43f547bfe5bae806d8d47b4e596ea5b25a6c1f080aef49fbcd817c5db8")]
 ,[]
 
,["/gnu/store/0mxnx8l4fgigvd7gakwdk6hc6im4wnai-disarchive-mirrors","/gnu/store/lv7dc09895a2d7cn0m6ysrdlgkhx5ld8-content-addressed-mirrors","/gnu/store/nb73zsx418kbz2p3f15pvv9spi0afmz7-mirrors"]
 ,"x86_64-linux","builtin:download",[]
 
,[("content-addressed-mirrors","/gnu/store/lv7dc09895a2d7cn0m6ysrdlgkhx5ld8-content-addressed-mirrors")
   
,("disarchive-mirrors","/gnu/store/0mxnx8l4fgigvd7gakwdk6hc6im4wnai-disarchive-mirrors")
   ,("impureEnvVars","http_proxy https_proxy LC_ALL LC_MESSAGES LANG COLUMNS")
   ,("mirrors","/gnu/store/nb73zsx418kbz2p3f15pvv9spi0afmz7-mirrors")
   ,("out","/gnu/store/b6h8v89cavb5zji6ys7bzlb193s0hzf2-curl-7.84.0.tar.xz")
   ,("preferLocalBuild","1")
   ,("url","\"https://curl.se/download/curl-7.84.0.tar.xz\"";)])
--8<---------------cut here---------------end--------------->8---
   
That’s why the prefix hash is different but not the content hash.  And
the fixed-output corresponding to the checksum hash in the source is the
expected one:

--8<---------------cut here---------------start------------->8---
$ guix hash /gnu/store/b6h8v89cavb5zji6ys7bzlb193s0hzf2-curl-7.84.0.tar.xz
1f2xgj0wvys9xw50h7vcbaraavjr9rxx9n06x2xfbgs7ym1qn49d
--8<---------------cut here---------------end--------------->8---


Last to be complete, if like me, you are asking why there is hashes
mismatch, some explanations.

These two items

  /gnu/store/0g24sh9v8hqf901aj7bp0hl5swf4hfr7-curl-7.84.0.tar.xz 
  /gnu/store/67w9zrlm1xahczf55f9rd15aaqadbixj-curl-7.84.0.tar.xz

have the same checksum hash but it is different from

  /gnu/store/b6h8v89cavb5zji6ys7bzlb193s0hzf2-curl-7.84.0.tar.xz

because the ’origin’ contains a patch:

    (origin
      (method url-fetch)
      (uri (string-append "https://curl.se/download/curl-";
                          version ".tar.xz"))
      (sha256
       (base32
        "129n9hi7rbg3s112chyadhp4y27ppb5i65n12wm77aw2255zf1h6"))
      (patches (search-patches "curl-use-ssl-cert-env.patch")))

Therefore, it means that the output of “guix build -S curl” contains
this patch and it does not correspond to what upstream distributes with
the checksum 129n9hi7rbg3s112chyadhp4y27ppb5i65n12wm77aw2255zf1h6.


The prefix hash /gnu/store/<prefix-hash>-curl-7.84.0.tar.xz is different
because it is computed using all the inputs and the derivation itself.
Since the derivation here contains the architecture i686-linux or
x86_64-linux then the output (prefix-hash) is thus different.


1: <87r134n66m.fsf@disroot.org">https://yhetil.org/guix/87r134n66m.fsf@disroot.org>


Cheers,
simon