Mailing List Archive

On 4 January 2017 at 13:27, Rainer Jung <rainer.jung@kippdata.de> wrote:
> Hi there,
>
> this commit broke compatibility for mod_perl 2.0.10 with perl before version
> 5.14. AFAIK API function mg_free_type() used here was only introduced in
> 5.14.0. Since the above commit was titled "Add support for Perl 5.22.x.", it
> might be possible to fix this using a version dependent code path. But I
> don't know enough about these internals to suggest an appropriate patch,
> sorry.
>

Thanks for the report.

Are you in a position to try the attached patch, which I think should
be sufficient to fix it? (mg_free() removes magic of all types, but I
wouldn't expect any magic other than the type targetted by the
mg_free_type() call to be present anyway.)

If not then I'll build a 5.14 and try it myself, but I don't have one
to hand at the moment.

Am 04.01.2017 um 14:54 schrieb Steve Hay:
> On 4 January 2017 at 13:27, Rainer Jung <rainer.jung@kippdata.de> wrote:
>> Hi there,
>>
>> this commit broke compatibility for mod_perl 2.0.10 with perl before version
>> 5.14. AFAIK API function mg_free_type() used here was only introduced in
>> 5.14.0. Since the above commit was titled "Add support for Perl 5.22.x.", it
>> might be possible to fix this using a version dependent code path. But I
>> don't know enough about these internals to suggest an appropriate patch,
>> sorry.
>>
>
> Thanks for the report.
>
> Are you in a position to try the attached patch, which I think should
> be sufficient to fix it? (mg_free() removes magic of all types, but I
> wouldn't expect any magic other than the type targetted by the
> mg_free_type() call to be present anyway.)
>
> If not then I'll build a 5.14 and try it myself, but I don't have one
> to hand at the moment.

Wow, thanks for the quick patch suggestion. Yes I can test, e.g. using
perl 5.10.1 on RHEL6. Any specific test scenario, that you would suggest
so that the code path in question is triggered?

Regards,

Rainer

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@perl.apache.org
For additional commands, e-mail: dev-help@perl.apache.org

On 4 January 2017 at 14:21, Rainer Jung <rainer.jung@kippdata.de> wrote:
> Am 04.01.2017 um 14:54 schrieb Steve Hay:
>>
>> On 4 January 2017 at 13:27, Rainer Jung <rainer.jung@kippdata.de> wrote:
>>>
>>> Hi there,
>>>
>>> this commit broke compatibility for mod_perl 2.0.10 with perl before
>>> version
>>> 5.14. AFAIK API function mg_free_type() used here was only introduced in
>>> 5.14.0. Since the above commit was titled "Add support for Perl 5.22.x.",
>>> it
>>> might be possible to fix this using a version dependent code path. But I
>>> don't know enough about these internals to suggest an appropriate patch,
>>> sorry.
>>>
>>
>> Thanks for the report.
>>
>> Are you in a position to try the attached patch, which I think should
>> be sufficient to fix it? (mg_free() removes magic of all types, but I
>> wouldn't expect any magic other than the type targetted by the
>> mg_free_type() call to be present anyway.)
>>
>> If not then I'll build a 5.14 and try it myself, but I don't have one
>> to hand at the moment.
>
>
> Wow, thanks for the quick patch suggestion. Yes I can test, e.g. using perl
> 5.10.1 on RHEL6. Any specific test scenario, that you would suggest so that
> the code path in question is triggered?
>

Anything will do - the code is used in start-up and shutdown, so
you'll know straight away if it's failing because it probably won't
even start up.

If you're able to run the test suite as well then that will exercise
all the ENV-related code in-between as well, but I think that should
all be fine.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@perl.apache.org
For additional commands, e-mail: dev-help@perl.apache.org

Am 04.01.2017 um 15:51 schrieb Steve Hay:
> On 4 January 2017 at 14:21, Rainer Jung <rainer.jung@kippdata.de> wrote:
>> Am 04.01.2017 um 14:54 schrieb Steve Hay:
>>>
>>> On 4 January 2017 at 13:27, Rainer Jung <rainer.jung@kippdata.de> wrote:
>>>>
>>>> Hi there,
>>>>
>>>> this commit broke compatibility for mod_perl 2.0.10 with perl before
>>>> version
>>>> 5.14. AFAIK API function mg_free_type() used here was only introduced in
>>>> 5.14.0. Since the above commit was titled "Add support for Perl 5.22.x.",
>>>> it
>>>> might be possible to fix this using a version dependent code path. But I
>>>> don't know enough about these internals to suggest an appropriate patch,
>>>> sorry.
>>>>
>>>
>>> Thanks for the report.
>>>
>>> Are you in a position to try the attached patch, which I think should
>>> be sufficient to fix it? (mg_free() removes magic of all types, but I
>>> wouldn't expect any magic other than the type targetted by the
>>> mg_free_type() call to be present anyway.)
>>>
>>> If not then I'll build a 5.14 and try it myself, but I don't have one
>>> to hand at the moment.
>>
>>
>> Wow, thanks for the quick patch suggestion. Yes I can test, e.g. using perl
>> 5.10.1 on RHEL6. Any specific test scenario, that you would suggest so that
>> the code path in question is triggered?
>>
>
> Anything will do - the code is used in start-up and shutdown, so
> you'll know straight away if it's failing because it probably won't
> even start up.
>
> If you're able to run the test suite as well then that will exercise
> all the ENV-related code in-between as well, but I think that should
> all be fine.

built and tested on RHEL 6 using platform tools gcc 4.4.5-6 and perl
v5.10.1. but Apache 2.4.4 (custom build).

Build succeeds without mg_free_type warning (as expected), make test
also succeeds:

...
All tests successful.
Files=245, Tests=4690, 401 wallclock secs ( 2.34 usr 7.62 sys + 174.34
cusr 157.29 csys = 341.59 CPU)
Result: PASS
...
cd ModPerl-Registry && make test
...
All tests successful.
Files=21, Tests=81, 35 wallclock secs ( 0.19 usr 0.64 sys + 14.21 cusr
12.91 csys = 27.95 CPU)
Result: PASS
...
cd Apache-Reload && make test
...
All tests successful.
Files=1, Tests=3, 1 wallclock secs ( 0.01 usr 0.06 sys + 0.65 cusr
0.57 csys = 1.29 CPU)
Result: PASS

So looks good!

Thanks a bunch,

Rainer

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@perl.apache.org
For additional commands, e-mail: dev-help@perl.apache.org

On 4 January 2017 at 15:33, Rainer Jung <rainer.jung@kippdata.de> wrote:
> Am 04.01.2017 um 15:51 schrieb Steve Hay:
>>
>> On 4 January 2017 at 14:21, Rainer Jung <rainer.jung@kippdata.de> wrote:
>>>
>>> Am 04.01.2017 um 14:54 schrieb Steve Hay:
>>>>
>>>>
>>>> On 4 January 2017 at 13:27, Rainer Jung <rainer.jung@kippdata.de> wrote:
>>>>>
>>>>>
>>>>> Hi there,
>>>>>
>>>>> this commit broke compatibility for mod_perl 2.0.10 with perl before
>>>>> version
>>>>> 5.14. AFAIK API function mg_free_type() used here was only introduced
>>>>> in
>>>>> 5.14.0. Since the above commit was titled "Add support for Perl
>>>>> 5.22.x.",
>>>>> it
>>>>> might be possible to fix this using a version dependent code path. But
>>>>> I
>>>>> don't know enough about these internals to suggest an appropriate
>>>>> patch,
>>>>> sorry.
>>>>>
>>>>
>>>> Thanks for the report.
>>>>
>>>> Are you in a position to try the attached patch, which I think should
>>>> be sufficient to fix it? (mg_free() removes magic of all types, but I
>>>> wouldn't expect any magic other than the type targetted by the
>>>> mg_free_type() call to be present anyway.)
>>>>
>>>> If not then I'll build a 5.14 and try it myself, but I don't have one
>>>> to hand at the moment.
>>>
>>>
>>>
>>> Wow, thanks for the quick patch suggestion. Yes I can test, e.g. using
>>> perl
>>> 5.10.1 on RHEL6. Any specific test scenario, that you would suggest so
>>> that
>>> the code path in question is triggered?
>>>
>>
>> Anything will do - the code is used in start-up and shutdown, so
>> you'll know straight away if it's failing because it probably won't
>> even start up.
>>
>> If you're able to run the test suite as well then that will exercise
>> all the ENV-related code in-between as well, but I think that should
>> all be fine.
>
>
> built and tested on RHEL 6 using platform tools gcc 4.4.5-6 and perl
> v5.10.1. but Apache 2.4.4 (custom build).
>
> Build succeeds without mg_free_type warning (as expected), make test also
> succeeds:
>

Great! Thanks for testing. Now committed in r1777355.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@perl.apache.org
For additional commands, e-mail: dev-help@perl.apache.org