Rick Hillegas <[email protected]> writes:
> On 11/23/17 3:39 AM, Knut Anders Hatlen wrote:
>> Knut Anders Hatlen <[email protected]> writes:
>>
>>> Rick Hillegas <[email protected]> writes:
>>>
>>>> On 11/22/17 12:56 AM, Knut Anders Hatlen wrote:
>>>>> Rick Hillegas <[email protected]> writes:
>>>>>
>>>>>> I have updated the Derby-trunk-JaCoCo configuration to use the 0.7.9
>>>>>> JaCoCo jars at
>>>>>> https://urldefense.proofpoint.com/v2/url?u=http-3A__people.apache.org_-7Erhillegas_derby_jacoco-2Djars_&d=DwIDaQ&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=mcFajTlbkIJ-DV-XHLk2TiSMDFfBhxQNXM5OBfEvYQ0&m=Otq7uQ72PfDMCJ4CD06LeG6bH0QqiepAX8rdSDsPTsM&s=ZhaPG5oM4BGrJfVvjLS03y6iz4nvtBip_vUHOgTHkhI&e=.
>>>>>> Let's see if that helps.
>>>>>>
>>>>>> Does anyone know why we also have a Derby-JaCoCo job (also broken)?
>>>>>> Why are we running JaCoCo twice and with different scripts?
>>>>> Hi Rick,
>>>>>
>>>>> I did some maintenance on the Derby-JaCoCo job some years ago, after it
>>>>> had broken when we had moved to a newer Java version which was not
>>>>> supported by the JaCoCo version we were using.
>>>>>
>>>>> I remember that we had two jobs at that time too. Don't know why we had
>>>>> two, though. I left the Derby-trunk-JaCoCo job alone, because it
>>>>> appeared to be abandoned (it was disabled, I think), and also because it
>>>>> fetched jar files from a committer's home directory. The Derby-JaCoCo
>>>>> job fetched the jar files from Maven, which sounded more maintainable as
>>>>> it allowed others than that one committer to make updates, if required.
>>>>>
>>>>> I was actually planning to delete the Derby-trunk-JaCoCo job after a
>>>>> while if the Derby-JaCoCo proved to be stable. But apparently I forgot.
>>>> Thanks, Knut. Sounds like we should retire Derby-trunk-JaCoCo and
>>>> figure out how to get Derby-JaCoCo to fetch the correct version of the
>>>> JaCoCo jars.
>>> As a first step, I've updated Derby-JaCoCo to use the latest version of
>>> JaCoCo (0.7.9). I did this by updating the following variables in the
>>> build section of the configuration:
>>>
>>> jacoco_version=0.7.9
>>> jacoco_zip_sha1=1d56a66be0f4a6b529d8f983ab65cc77a52cccb6
>>> jacocoant_sha1=87aa22de3854fd5a43e37a17c4b245b01a46de93
>>> jacocoagent_sha1=a6ac9cca89d889222a40dab9dd5039bfd22a4cff
>>>
>>> I had to download the JaCoCo zip file locally and extract it to
>>> calculate the SHA1 checksums of the zip file and the two jar files we
>>> need from it. The checksums are used by the build script to check if we
>>> already have the right version, or if we have to download it from Maven.
>>>
>>> I've kicked off a build, and it has already passed the point where it
>>> failed before, so maybe that's all that was needed. Let's see what
>>> happens...
>> The build job completed and produced a coverage report, so there is
>> improvement. 10 tests failed, though, so the job is still marked as
>> "unstable".
> Thanks, Knut. The failures all result from a missing FilePermission
> under DropDatabaseSetup while checking for the existence of a
> directory before deleting it. The tests work standalone on my machine.
> Do you remember if the JaCoCo test runs have their own security
> policy, which might need to be updated?
Hi Rick,
I think they use the ordinary policy files, but the extra permissions
needed for JaCoCo are guarded by the jacoco.active system property. For
example:
permission java.io.FilePermission "${jacoco.active}${user.dir}${/}*", "read,
write";
When JaCoCo is active, the jacoco.active property is set to the empty
string, so that the above rule becomes equivalent to
permission java.io.FilePermission "${user.dir}${/}*", "read, write";
and thereby becomes active.
The code that sets the jacoco.active property seems to be located in
BaseTestCase and SecurityManagerSetup.
--
Knut Anders