Re: Weak globals

Yes, I do.  It's going to take a little while because I want the whole 
package.  Another anomaly is that the syntax does NOT work with the C 
compiler, including the alternate syntax described in the manual:

https://www.ibm.com/support/knowledgecenter/en/ssw_ibm_i_73/rzarg/function_attributes.htm

So I'll put together my test cases and reach out to the fine folks.



On 4/2/2019 10:00 AM, Mark Waterbury wrote:
>   Joe,
> Sounds like it may be time to open a Support incident with IBM (PMR) to get some clarification -- maybe it's a defect and they might accept it as an APAR that may eventually result in PTFs?  :)  And, you already have a nice simple little "test case" to submit with your PMR ... ;-)
>
> Good luck,
> Mark S. Waterbury
>
>      On Tuesday, April 2, 2019, 10:54:31 AM EDT, Joe Pluta <[email protected]> wrote:
>   
>   Sadly, it's quite inconsistent.
>
> This works in CRTSRVPGM:
>
> // Compilation unit 2:
>
> #include <stdio.h>
>
> void foo() __attribute__((weak));
> void foo()
> {
> printf("Foo called from compilation unit 2.\n");
> }
>
> I am able to create the service program and it looks like the symbols
> resolved correctly.  But before I test that, I tried to do CRTPGM with
> the same modules (remember that TFOO1 has a proper main).  That failed
> with CPD5D03 and the listing showed that the TFOO2 definition was being
> treated as a strong global.
>
> So:
>
> #pragma weak doesn't work in CRTSRVPGM or CRTPGM
> __attribute__((weak)) works in CRTSRVPGM
> Neither works in CRTPGM
>
> Ugh.  I'll try the service program in a moment, just to see if calling
> routines resolve correctly.
>
>
>
>
> On 4/2/2019 9:20 AM, Joe Pluta wrote:
>> It looks like the #pragma weak is just not working.  Here's the
>> relevant CRTSRVPGM output:
>>
>> 5770SS1 V7R3M0 160422 MVXJP/TFOO    DEV       04/02
>>   Module . . . . . . . . . . . : TFOO2
>>     Library  . . . . . . . . . : MVXJP
>>     Bound  . . . . . . . . . . : *YES
>>     Change date/time . . . . . :     04/02/19 07:39:13
>>     Teraspace storage enabled  : *YES
>>     Storage model  . . . . . . : *SNGLVL
>>         Number    Symbol      Ref
>> Identifier                                   Type      Scope Export
>>         00000025  Def foo__Fv Proc      Module Strong
>>         00000026  Ref         0000011C _C_SIGABRT_ctl_action Data
>>
>> Note that foo__Fv in TFOO2 is identified as a strong export. That's
>> just not correct.  It later conflicts with the foo__Fv from TFOO3 and
>> the service program is not created.
>>
>>
>>
>> On 4/2/2019 9:12 AM, Joe Pluta wrote:
>>> Nope - still getting CPD5D03.
>>>
>>> On 4/2/2019 9:08 AM, Joe Pluta wrote:
>>>> Thanks Jon, that's the way I'll go if I can't figure anything else
>>>> out. As you probably guess, I don't like DUPPROC, especially since
>>>> as I read the documentation, weak should work.
>>>>
>>>> The EXPORT(*ALL) is simply for testing, but I'm not sure how the
>>>> binder source would help.  I'm not very versed in binder source, but
>>>> as far as I can tell all binder source does is order the exported
>>>> modules.  How would it help when resolving duplicate symbols?
>>>>
>>>> Oh wait.  Maybe that's the problem! Maybe it's trying to export both
>>>> the weak and the strong.  Ah, let me try that.  If I add binder
>>>> source to only export main maybe that will work.
>>>>
>>>>
>>>> On 4/2/2019 8:53 AM, Jon Paris wrote:
>>>>> Joe - I think dup proc is the only way to go. As to your comment
>>>>> "and even then the order isn't _guaranteed_" that should be dealt
>>>>> with by binder language/source instead of doing Export *All.  That
>>>>> guarantees the slot sequence.
>>>>>
>>>>>
>>>>> Jon Paris
>>>>>
>>>>> www.partner400.com
>>>>> www.SystemiDeveloper.com
>>>>>
>>>>>> On Apr 2, 2019, at 8:43 AM, Joe Pluta <[email protected]>
>>>>>> wrote:
>>>>>>
>>>>>> Has anybody successfully used the weak #pragma for a function?
>>>>>> I'm trying to create a service program with three modules: one
>>>>>> that references a function, one with that function defined weakly,
>>>>>> and one with the function defined strongly.  In that scenario, I
>>>>>> should be able to combine the following:
>>>>>>
>>>>>> Reference, weak: calls weak
>>>>>> Reference, strong: calls strong
>>>>>> Reference, weak, strong: calls strong
>>>>>> Reference only: won't compile
>>>>>>
>>>>>> The best example code I found was in the AIX reference:
>>>>>>
>>>>>> https://www.ibm.com/support/knowledgecenter/SSGH3R_13.1.0/com.ibm.xlcpp131.aix.doc/compiler_ref/prag_weak.html
>>>>>>
>>>>>>
>>>>>>   From what I can tell, that same syntax should work in the ILE
>>>>>> compiler, so I created the following:
>>>>>>
>>>>>> |// TFOO1.CPP #include <stdio.h> void foo(); int main() { foo(); }
>>>>>> // TFOO2.CPP #include <stdio.h> #pragma weak foo__Fv void foo() {
>>>>>> printf("Foo called from compilation unit 2\n"); } // TFOO3.CPP
>>>>>> #include <stdio.h> void foo() { printf("Foo called from
>>>>>> compilation unit 3\n"); } |
>>>>>>
>>>>>>
>>>>>> CRTSRVPGM TFOO MODULES(TFOO1 TFOO2 TFOO3) EXPORT(*ALL) fails with
>>>>>> CPD5D03, Definition supplied multiple times for symbol 'foo__Fv'.
>>>>>>
>>>>>> I can get around it with OPTION(*DUPPROC) but then I have to be
>>>>>> very careful of the order of my modules in the MODULES keyword of
>>>>>> CRTSRVPGM (and even then the order isn't _guaranteed_).  That also
>>>>>> would hide any cases of unintentional procedure duplication, so
>>>>>> I'd really like to get #pragma weak working.
>>>>>>
>>>>>>
>>>>>> -- 
>>>>>> This is the Bare Metal Programming IBM i (AS/400 and iSeries)
>>>>>> (C400-L) mailing list
>>>>>> To post a message email: [email protected]
>>>>>> To subscribe, unsubscribe, or change list options,
>>>>>> visit: https://lists.midrange.com/mailman/listinfo/c400-l
>>>>>> or email: [email protected]
>>>>>> Before posting, please take a moment to review the archives
>>>>>> at https://archive.midrange.com/c400-l.
>>>>>>
>>>>>> Help support midrange.com by shopping at amazon.com with our
>>>>>> affiliate
>>>>>> link: https://amazon.midrange.com
>>>>

-- 
This is the Bare Metal Programming IBM i (AS/400 and iSeries) (C400-L) mailing list
To post a message email: [email protected]
To subscribe, unsubscribe, or change list options,
visit: https://lists.midrange.com/mailman/listinfo/c400-l
or email: [email protected]
Before posting, please take a moment to review the archives
at https://archive.midrange.com/c400-l.

Help support midrange.com by shopping at amazon.com with our affiliate
link: https://amazon.midrange.com