[bug#27431] Acl test investigation

[bug#27431] Acl test investigation

[Kei Kebreau]

[-- Attachment #1: Type: text/plain, Size: 518 bytes --]

I discovered that the Acl tests are failing partly because the coreutils
seemingly aren't compiled with Acl support before the tests are run
(even though acl is an input to coreutils). The tests are written with
the assumption that the coreutils are built with Acl support. An example
is in the sbits-restore.test file, where a sed command is expecting to
remove an Acl bit from ls output, but instead removes normal directory
permission information.

Any ideas on how to resolve this without disabling the Acl tests?

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]

Acl test investigation

[Kei Kebreau]

[-- Attachment #1: Type: text/plain, Size: 518 bytes --]

I discovered that the Acl tests are failing partly because the coreutils
seemingly aren't compiled with Acl support before the tests are run
(even though acl is an input to coreutils). The tests are written with
the assumption that the coreutils are built with Acl support. An example
is in the sbits-restore.test file, where a sed command is expecting to
remove an Acl bit from ls output, but instead removes normal directory
permission information.

Any ideas on how to resolve this without disabling the Acl tests?

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]

bug#27431: Acl test investigation

[Kei Kebreau]

[-- Attachment #1: Type: text/plain, Size: 601 bytes --]

Kei Kebreau <kei@openmailbox.org> writes:

> I discovered that the Acl tests are failing partly because the coreutils
> seemingly aren't compiled with Acl support before the tests are run
> (even though acl is an input to coreutils). The tests are written with
> the assumption that the coreutils are built with Acl support. An example
> is in the sbits-restore.test file, where a sed command is expecting to
> remove an Acl bit from ls output, but instead removes normal directory
> permission information.
>
> Any ideas on how to resolve this without disabling the Acl tests?

Wrong mailing list...

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]

[bug#27431] Acl test investigation

[Ludovic Courtès]

Hello Kei,

Kei Kebreau <kei@openmailbox.org> skribis:

> I discovered that the Acl tests are failing partly because the coreutils
> seemingly aren't compiled with Acl support before the tests are run
> (even though acl is an input to coreutils). The tests are written with
> the assumption that the coreutils are built with Acl support. An example
> is in the sbits-restore.test file, where a sed command is expecting to
> remove an Acl bit from ls output, but instead removes normal directory
> permission information.
>
> Any ideas on how to resolve this without disabling the Acl tests?

Could you give a bit more context?  On which branch does this failure
happen?

If the unit tests of the ‘acl’ package check the output of ‘ls’, then
there’s a circular dependency issue between Coreutils and ACL.  I would
argue that it’s up to Coreutils to test its ‘ls’ command, but in the
meantime, we could patch ACL to skip its ‘ls’ test.

Is this what you were asking?  :-)

Ludo’.

Re: Acl test investigation

[Kei Kebreau]

[-- Attachment #1: Type: text/plain, Size: 1433 bytes --]

ludo@gnu.org (Ludovic Courtès) writes:

> Hello Kei,
>
> Kei Kebreau <kei@openmailbox.org> skribis:
>
>> I discovered that the Acl tests are failing partly because the coreutils
>> seemingly aren't compiled with Acl support before the tests are run
>> (even though acl is an input to coreutils). The tests are written with
>> the assumption that the coreutils are built with Acl support. An example
>> is in the sbits-restore.test file, where a sed command is expecting to
>> remove an Acl bit from ls output, but instead removes normal directory
>> permission information.
>>
>> Any ideas on how to resolve this without disabling the Acl tests?
>
> Could you give a bit more context?  On which branch does this failure
> happen?
>

I discovered this on the master branch and there didn't seem to be fixes
on core-updates.

> If the unit tests of the ‘acl’ package check the output of ‘ls’, then
> there’s a circular dependency issue between Coreutils and ACL. I would
> argue that it’s up to Coreutils to test its ‘ls’ command, but in the
> meantime, we could patch ACL to skip its ‘ls’ test.
>

ACL tests still fail later on, but the 'ls' test is easily fixable (just
remove the sed statement that removes the last character). I'll craft
the 'ls' related patch and see if the others can be fixed as well.

> Is this what you were asking?  :-)
>

Yes. Thanks!

> Ludo’.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]