Michał Górny <[email protected]> writes:

> On Fri, 2025-08-08 at 17:26 -0400, Eli Schwartz wrote:
>> Apparently this PROPERTIES has the insidious effect of requiring
>> multiple moving parts to actually have the desired effect:
>> 
>> - PROPERTIES="test_network"
>> 
>>   Add network access, always, and run tests without comment.
>> 
>> - PROPERTIES="test_network" RESTRICT="test"
>> 
>>   Don't run tests, but if tests ran, they would have network.
>> 
>> - PROPERTIES="test_network" RESTRICT="test" ALLOW_TEST="network"
>> 
>>   Disable RESTRICT, and run tests.
>> 
>> - PROPERTIES="test_network" RESTRICT="!test? ( test )" ALLOW_TEST="network"
>> 
>>   Run tests without build dependencies installed, if USE="-test" is set.
>> 
>> - eclass:
>>       PROPERTIES="test_network" RESTRICT="!test? ( test ) test" 
>> ALLOW_TEST="network"
>>   ebuild:
>>       RESTRICT="test" # tests broken, see bug #xxxxxx
>> 
>>   Run broken tests, which fail. Sorry, you're not allowed to communicate
>>   the desired intent. Don't use eclasses.
>> 
>> This is a total mess, but apparently it is what we must do anyways in
>> order to have the eclass behave as intended. So, switch from the first
>> configuration to the fifth, and let's simply pray nobody ever wants to
>> restrict tests for *any* reason for the package subset known as
>> "sec-keys/*" and we can safely co-opt its meaning.
>> 
>> Signed-off-by: Eli Schwartz <[email protected]>
>> ---
>>  eclass/sec-keys.eclass | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>> 
>> diff --git a/eclass/sec-keys.eclass b/eclass/sec-keys.eclass
>> index ce0396be414a..ed53adb5dbae 100644
>> --- a/eclass/sec-keys.eclass
>> +++ b/eclass/sec-keys.eclass
>> @@ -96,7 +96,7 @@ SLOT="0"
>>  
>>  IUSE="test"
>>  PROPERTIES="test_network"
>> -RESTRICT="!test? ( test )"
>> +RESTRICT="!test? ( test ) test"
>> 
>
> You do realize that this is equivalent to RESTRICT="test", right?

I often write both when I expect to drop the 'test' in the
near-future. Not that it really matters.

Reply via email to