Re: [PATCH 3/5] gnu: Enable tests in Python 3.

[ludo@gnu.org (Ludovic Courtès)]

Cyril Roelandt <tipecaml@gmail.com> skribis:

> On 03/21/2014 02:15 PM, Ludovic Courtès wrote:
>> Cyril Roelandt <tipecaml@gmail.com> skribis:
>>
>>> On 03/07/2014 10:40 AM, Ludovic Courtès wrote:
>>>> Cyril Roelandt <tipecaml@gmail.com> skribis:
>>
>> [...]
>>
>>>>> +--- Lib/test/test_posixpath.py	2014-03-01 05:46:56.984311000 +0100
>>>>> ++++ Lib/test/test_posixpath.py	2014-03-01 06:20:50.704311000 +0100
>>>>> +@@ -319,7 +319,11 @@
>>>>> +                 del env['HOME']
>>>>> +                 home = pwd.getpwuid(os.getuid()).pw_dir
>>>>> +                 # $HOME can end with a trailing /, so strip it (see #17809)
>>>>> +-                self.assertEqual(posixpath.expanduser("~"), home.rstrip("/"))
>>>>> ++		# The Guix builders have '/' as a home directory, so
>>>>> ++		# home.rstrip("/") will be an empty string and the test will
>>>>> ++		# fail. Let's just disable it since it does not really make
>>>>> ++		# sense with such a bizarre setup.
>>>>> ++                # self.assertEqual(posixpath.expanduser("~"), home.rstrip("/"))
>>>>
>>>> I see that the recipe’s ‘pre-check’ phase does:
>>>>
>>>>     (setenv "HOME" (getcwd))
>>>>
>>>> so $HOME should actually be /tmp/nix-build-xxx, not just /, no?
>>>>
>>>
>>> Yes, but I think it uses getpwent() or something that does not use
>>> $HOME to retrieve the home directory in this case.
>>
>> Here’s what I see, with nscd turned off (this uses getpwnam(2)):
>>
>> --8<---------------cut here---------------start------------->8---
>> $ strace guile -c '(getpw "ludo")'
>> [...]
>> open("/etc/passwd", O_RDONLY|O_CLOEXEC) = 5
>> fstat(5, {st_mode=S_IFREG|0644, st_size=2072, ...}) = 0
>> mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f264dea5000
>> read(5, "root:x:0:0:System administrator:"..., 4096) = 2072
>> --8<---------------cut here---------------end--------------->8---
>>
>> Normally the chroot environment has all that too.  And indeed, the
>> equivalent of the above Python snippet seems to work:
>>
>> --8<---------------cut here---------------start------------->8---
>> scheme@(guile-user)> ,use(guix)
>> scheme@(guile-user)> ,use(guix monads)
>> scheme@(guile-user)> (define s (open-connection))
>> scheme@(guile-user)> (derivation-expression "getpw" '(pk (getpw (getuid))))
>> $11 = #<procedure 3da30f0 at guix/monads.scm:297:6 (store)>
>> scheme@(guile-user)> (run-with-store s $11)
>> $12 = #<derivation /gnu/store/bhmd5bg9s64an3s8ssmwx6l90dkr56hx-getpw.drv => /gnu/store/z68rzri3myxx1dg931fb1ksxdg7c29zg-getpw 33dfa00>
>> scheme@(guile-user)> (build-derivations s (list $12))
>> building path(s) `/gnu/store/z68rzri3myxx1dg931fb1ksxdg7c29zg-getpw'
>>
>> ;;; (#("nixbld" "x" 30001 30000 "Nix build user" "/" "/noshell"))
>
> Am I misteading this, or is the home directory '/' for the nixbld user
> ?

Correct, as per build.cc:

--8<---------------cut here---------------start------------->8---
        /* Create a /etc/passwd with entries for the build user and the
           nobody account.  The latter is kind of a hack to support
           Samba-in-QEMU. */
        createDirs(chrootRootDir + "/etc");

        writeFile(chrootRootDir + "/etc/passwd",
            (format(
                "nixbld:x:%1%:%2%:Nix build user:/:/noshell\n"
                "nobody:x:65534:65534:Nobody:/:/noshell\n")
                % (buildUser.enabled() ? buildUser.getUID() : getuid())
                % (buildUser.enabled() ? buildUser.getGID() : getgid())).str());
--8<---------------cut here---------------end--------------->8---

> That's exactly what happens in Python (that uses getwpent). Hence the
> bug. I think skipping this assertion is the right thing to do here.

Oh right.  (Somehow I ended up thinking that getpwent was not working or
something like that; sorry for the misunderstanding!)

So OK to push as is!

Thanks,
Ludo’.