| 1 |
On 08/26/2018 01:08 AM, Michał Górny wrote: |
| 2 |
> On Sun, 2018-08-26 at 00:58 -0700, Zac Medico wrote: |
| 3 |
>> On 08/26/2018 12:39 AM, Michał Górny wrote: |
| 4 |
>>> On Sun, 2018-08-26 at 00:28 -0700, Zac Medico wrote: |
| 5 |
>>>> On 08/25/2018 07:11 PM, Andrew Savchenko wrote: |
| 6 |
>>>>> On Sat, 25 Aug 2018 14:24:02 -0400 Mike Gilbert wrote: |
| 7 |
>>>>>> On Sat, Aug 25, 2018 at 1:41 AM Andrew Savchenko <bircoph@g.o> wrote: |
| 8 |
>>>>>>> |
| 9 |
>>>>>>> On Fri, 24 Aug 2018 14:24:06 -0400 Mike Gilbert wrote: |
| 10 |
>>>>>>>> --- |
| 11 |
>>>>>>>> general-concepts/dependencies/text.xml | 38 ++++++++++++++++++++++++++ |
| 12 |
>>>>>>>> 1 file changed, 38 insertions(+) |
| 13 |
>>>>>>>> |
| 14 |
>>>>>>>> diff --git a/general-concepts/dependencies/text.xml b/general-concepts/dependencies/text.xml |
| 15 |
>>>>>>>> index 2f10380..64be9dc 100644 |
| 16 |
>>>>>>>> --- a/general-concepts/dependencies/text.xml |
| 17 |
>>>>>>>> +++ b/general-concepts/dependencies/text.xml |
| 18 |
>>>>>>>> @@ -578,6 +578,44 @@ valid. |
| 19 |
>>>>>>>> </body> |
| 20 |
>>>>>>>> </section> |
| 21 |
>>>>>>>> |
| 22 |
>>>>>>>> +<section> |
| 23 |
>>>>>>>> +<title>Test Dependencies</title> |
| 24 |
>>>>>>>> +<body> |
| 25 |
>>>>>>>> + |
| 26 |
>>>>>>>> +<p> |
| 27 |
>>>>>>>> +Packages often have optional dependencies that are needed only when running |
| 28 |
>>>>>>>> +tests. These should be specified in DEPEND behind a USE flag. Often, the |
| 29 |
>>>>>>>> +'test' USE flag is used for this purpose. |
| 30 |
>>>>>>>> +</p> |
| 31 |
>>>>>>>> + |
| 32 |
>>>>>>>> +<p> |
| 33 |
>>>>>>>> +Since testing will likely fail when test dependencies are not installed, the |
| 34 |
>>>>>>>> +test phase should be disabled in this case. This may be accomplished via USE |
| 35 |
>>>>>>>> +conditionals in the RESTRICT variable. |
| 36 |
>>>>>>>> +</p> |
| 37 |
>>>>>>>> + |
| 38 |
>>>>>>>> +<p> |
| 39 |
>>>>>>>> +If other optional features must be enabled/disabled when testing, REQUIRED_USE |
| 40 |
>>>>>>>> +may be set to express this. |
| 41 |
>>>>>>>> +</p> |
| 42 |
>>>>>>>> + |
| 43 |
>>>>>>>> +<codesample lang="ebuild"> |
| 44 |
>>>>>>>> +# Define some USE flags |
| 45 |
>>>>>>>> +IUSE="debug test" |
| 46 |
>>>>>>>> + |
| 47 |
>>>>>>>> +# Disable test phase when test USE flag is disabled |
| 48 |
>>>>>>>> +RESTRICT="!test? ( test )" |
| 49 |
>>>>>>> |
| 50 |
>>>>>>> I do not understand why we need this useless code. If test USE flag |
| 51 |
>>>>>>> is disabled, tests must be disabled as well. It is PM's job and |
| 52 |
>>>>>>> there is no need to put this obvious stuff into each ebuild with |
| 53 |
>>>>>>> tests and extra deps. I see no reason to support running src_test() |
| 54 |
>>>>>>> with USE="-test". |
| 55 |
>>>>>> |
| 56 |
>>>>>> PMS does not specify that behavior (skipping src_test with USE=-test). |
| 57 |
>>>>>> It is better to define the requrement explicitly rather than relying |
| 58 |
>>>>>> on a Portage-specific behavior. |
| 59 |
>>>>> |
| 60 |
>>>>> Then PMS should be fixed. Putting useless code in thousands |
| 61 |
>>>>> of ebuilds due to bureaucratic reasons is ridiculous. Having strict |
| 62 |
>>>>> conformance to the PMS is good, but common sense should still be |
| 63 |
>>>>> considered. |
| 64 |
>>>> |
| 65 |
>>>> Since PMS doesn't specify the behavior of FEATURES, I suppose we could |
| 66 |
>>>> make FEATURES=test imply RESTRICT="!test? ( test )". Would there be any |
| 67 |
>>>> drawbacks to that? |
| 68 |
>>> |
| 69 |
>>> Except for Portage once again silently letting developers pass with non- |
| 70 |
>>> PMS behavior, and making other PMs fail with the ebuilds? |
| 71 |
>> |
| 72 |
>> Is it really sane for other PMs to behave that way though? |
| 73 |
>> |
| 74 |
>> In PMS it says src_test "may be disabled by user too, using a |
| 75 |
>> PM-specific mechanism", which doesn't explicitly rule out behavior based |
| 76 |
>> on USE settings. |
| 77 |
> |
| 78 |
> It doesn't rule out disabling based on the phase of the moon either, or |
| 79 |
> my mood which is degrading badly with this discussion. |
| 80 |
|
| 81 |
Well, using package.use for this doesn't really make sense within the |
| 82 |
context of ebuilds that don't have IUSE=test, which is an argument for |
| 83 |
people to use package.env as a consistent means to selectively disable |
| 84 |
unit tests. |
| 85 |
-- |
| 86 |
Thanks, |
| 87 |
Zac |
Archives