[Xenomai] panic macro interference with xerces api.

[Xenomai] panic macro interference with xerces api.

[Stéphane ANCELOT]

Hi,

I don't know if there might be a  big impact, but I noticed, depending 
on the order we include headers , that
there was a panic macro interfering with xercesc api headers.

#include <alchemy/task.h>
#include <xercesc/dom/DOMException.hpp>


/usr/xenomai-3.0-igb/include/boilerplate/ancillaries.h:64:10: error: 
expected identifier before '__func__'
   __panic(__func__, __fmt, ##__args)
           ^
Regards,
Steph

Re: [Xenomai] panic macro interference with xerces api.

[Gilles Chanteperdrix]

On Tue, Nov 10, 2015 at 02:26:54PM +0100, Stéphane ANCELOT wrote:
> Hi,
> 
> I don't know if there might be a  big impact, but I noticed, depending 
> on the order we include headers , that
> there was a panic macro interfering with xercesc api headers.
> 
> #include <alchemy/task.h>
> #include <xercesc/dom/DOMException.hpp>
> 
> 
> /usr/xenomai-3.0-igb/include/boilerplate/ancillaries.h:64:10: error: 
> expected identifier before '__func__'
>    __panic(__func__, __fmt, ##__args)

Well, if the two headers define the same macro, there are not many
solutions: arrange for not having to include them in the same file.
It should not be too hard, since a real-time thread using
alchemy/task.h has no business using an xml library, probably only
using linux services which a real-time task can not use without
switching to secondary mode.

-- 
					    Gilles.
https://click-hack.org

Re: [Xenomai] panic macro interference with xerces api.

[g.strobbe]

----- Original Message -----
> From: "Gilles Chanteperdrix" <gilles.chanteperdrix@xenomai.org>
> To: "Stéphane ANCELOT" <sancelot@free.fr>
> Cc: xenomai@xenomai.org
> Sent: Tuesday, November 10, 2015 3:28:14 PM
> Subject: Re: [Xenomai] panic macro interference with xerces api.
> 
> On Tue, Nov 10, 2015 at 02:26:54PM +0100, Stéphane ANCELOT wrote:
> > Hi,
> > 
> > I don't know if there might be a  big impact, but I noticed, depending
> > on the order we include headers , that
> > there was a panic macro interfering with xercesc api headers.
> > 
> > #include <alchemy/task.h>
> > #include <xercesc/dom/DOMException.hpp>
> > 
> > 
> > /usr/xenomai-3.0-igb/include/boilerplate/ancillaries.h:64:10: error:
> > expected identifier before '__func__'
> >    __panic(__func__, __fmt, ##__args)
> 
> Well, if the two headers define the same macro, there are not many
> solutions: arrange for not having to include them in the same file.
> It should not be too hard, since a real-time thread using
> alchemy/task.h has no business using an xml library, probably only
> using linux services which a real-time task can not use without
> switching to secondary mode.
> 
> --
> 					    Gilles.
> https://click-hack.org
> 
> _______________________________________________
> Xenomai mailing list
> Xenomai@xenomai.org
> http://xenomai.org/mailman/listinfo/xenomai
> 

I recently had a similar issue with the macro

#define panic(__fmt, __args...)			\
	__panic(__func__, __fmt, ##__args)

defined in ancillaries.h

The preprocessor modified virtual void panic(const PanicHandler::PanicReasons reason) = 0;

in the included xercesc/util/PanicHandler.hpp

resulting in a compilation error.

>From the gcc preprocessor manual https://gcc.gnu.org/onlinedocs/cpp/Undefining-and-Redefining-Macros.html#Undefining-and-Redefining-Macros

If a macro ceases to be useful, it may be undefined with the ‘#undef’ directive. ‘#undef’ takes a single argument, the name of the macro to undefine. You use the bare macro name, even if the macro is function-like. It is an error if anything appears on the line after the macro name. ‘#undef’ has no effect if the name is not a macro.  

So after the #include <alchemy/task.h> I placed
            
#undef panic

No problems since. Should also be fine as a solution, no?

Best regards,

Gino

Re: [Xenomai] panic macro interference with xerces api.

[Jan Kiszka]

On 2015-11-10 15:28, Gilles Chanteperdrix wrote:
> On Tue, Nov 10, 2015 at 02:26:54PM +0100, Stéphane ANCELOT wrote:
>> Hi,
>>
>> I don't know if there might be a  big impact, but I noticed, depending 
>> on the order we include headers , that
>> there was a panic macro interfering with xercesc api headers.
>>
>> #include <alchemy/task.h>
>> #include <xercesc/dom/DOMException.hpp>
>>
>>
>> /usr/xenomai-3.0-igb/include/boilerplate/ancillaries.h:64:10: error: 
>> expected identifier before '__func__'
>>    __panic(__func__, __fmt, ##__args)
> 
> Well, if the two headers define the same macro, there are not many
> solutions: arrange for not having to include them in the same file.
> It should not be too hard, since a real-time thread using
> alchemy/task.h has no business using an xml library, probably only
> using linux services which a real-time task can not use without
> switching to secondary mode.

Isn't *panic an internal service? Then we can maybe avoid that it is
exposed to applications. If it's not internal, it should probably carry
some "xenomai" prefix in its (fairly generic) name.

Jan

-- 
Siemens AG, Corporate Technology, CT RTC ITP SES-DE
Corporate Competence Center Embedded Linux

Re: [Xenomai] panic macro interference with xerces api.

[Philippe Gerum]

On 11/11/2015 11:31 AM, Jan Kiszka wrote:
> On 2015-11-10 15:28, Gilles Chanteperdrix wrote:
>> On Tue, Nov 10, 2015 at 02:26:54PM +0100, Stéphane ANCELOT wrote:
>>> Hi,
>>>
>>> I don't know if there might be a  big impact, but I noticed, depending 
>>> on the order we include headers , that
>>> there was a panic macro interfering with xercesc api headers.
>>>
>>> #include <alchemy/task.h>
>>> #include <xercesc/dom/DOMException.hpp>
>>>
>>>
>>> /usr/xenomai-3.0-igb/include/boilerplate/ancillaries.h:64:10: error: 
>>> expected identifier before '__func__'
>>>    __panic(__func__, __fmt, ##__args)
>>
>> Well, if the two headers define the same macro, there are not many
>> solutions: arrange for not having to include them in the same file.
>> It should not be too hard, since a real-time thread using
>> alchemy/task.h has no business using an xml library, probably only
>> using linux services which a real-time task can not use without
>> switching to secondary mode.
> 
> Isn't *panic an internal service? Then we can maybe avoid that it is
> exposed to applications. If it's not internal, it should probably carry
> some "xenomai" prefix in its (fairly generic) name.
> 

Sure, this service is aimed at being fairly generic and general-purpose,
with backtracing support, timing data, error code translation and such.
With respect to layering, I would rather expect an XML lib not to mess
with the namespace of core services, instead of asking the xenomai libs
which do deliver core services to apps to prefix its own services.

I may move those helpers in some specific file to hide them, but no, I
really don't want to go for xenomai_panic(), because it really should
work the other way around (although nobody is likely to care about this
on the apache side).

-- 
Philippe.

Re: [Xenomai] panic macro interference with xerces api.

[Philippe Gerum]

On 11/11/2015 12:52 PM, Philippe Gerum wrote:
> On 11/11/2015 11:31 AM, Jan Kiszka wrote:
>> On 2015-11-10 15:28, Gilles Chanteperdrix wrote:
>>> On Tue, Nov 10, 2015 at 02:26:54PM +0100, Stéphane ANCELOT wrote:
>>>> Hi,
>>>>
>>>> I don't know if there might be a  big impact, but I noticed, depending 
>>>> on the order we include headers , that
>>>> there was a panic macro interfering with xercesc api headers.
>>>>
>>>> #include <alchemy/task.h>
>>>> #include <xercesc/dom/DOMException.hpp>
>>>>
>>>>
>>>> /usr/xenomai-3.0-igb/include/boilerplate/ancillaries.h:64:10: error: 
>>>> expected identifier before '__func__'
>>>>    __panic(__func__, __fmt, ##__args)
>>>
>>> Well, if the two headers define the same macro, there are not many
>>> solutions: arrange for not having to include them in the same file.
>>> It should not be too hard, since a real-time thread using
>>> alchemy/task.h has no business using an xml library, probably only
>>> using linux services which a real-time task can not use without
>>> switching to secondary mode.
>>
>> Isn't *panic an internal service? Then we can maybe avoid that it is
>> exposed to applications. If it's not internal, it should probably carry
>> some "xenomai" prefix in its (fairly generic) name.
>>
> 
> Sure, this service is aimed at being fairly generic and general-purpose,
> with backtracing support, timing data, error code translation and such.
> With respect to layering, I would rather expect an XML lib not to mess
> with the namespace of core services, instead of asking the xenomai libs
> which do deliver core services to apps to prefix its own services.
> 
> I may move those helpers in some specific file to hide them, but no, I
> really don't want to go for xenomai_panic(), because it really should
> work the other way around (although nobody is likely to care about this
> on the apache side).
> 

Actually, the only issue is about defining the panic helpers as macros,
so that we can pull the source/line information automatically. The
namespace issue should not affect C++.

-- 
Philippe.

Re: [Xenomai] panic macro interference with xerces api.

[Philippe Gerum]

On 11/11/2015 01:28 PM, Philippe Gerum wrote:
> On 11/11/2015 12:52 PM, Philippe Gerum wrote:
>> On 11/11/2015 11:31 AM, Jan Kiszka wrote:
>>> On 2015-11-10 15:28, Gilles Chanteperdrix wrote:
>>>> On Tue, Nov 10, 2015 at 02:26:54PM +0100, Stéphane ANCELOT wrote:
>>>>> Hi,
>>>>>
>>>>> I don't know if there might be a  big impact, but I noticed, depending 
>>>>> on the order we include headers , that
>>>>> there was a panic macro interfering with xercesc api headers.
>>>>>
>>>>> #include <alchemy/task.h>
>>>>> #include <xercesc/dom/DOMException.hpp>
>>>>>
>>>>>
>>>>> /usr/xenomai-3.0-igb/include/boilerplate/ancillaries.h:64:10: error: 
>>>>> expected identifier before '__func__'
>>>>>    __panic(__func__, __fmt, ##__args)
>>>>
>>>> Well, if the two headers define the same macro, there are not many
>>>> solutions: arrange for not having to include them in the same file.
>>>> It should not be too hard, since a real-time thread using
>>>> alchemy/task.h has no business using an xml library, probably only
>>>> using linux services which a real-time task can not use without
>>>> switching to secondary mode.
>>>
>>> Isn't *panic an internal service? Then we can maybe avoid that it is
>>> exposed to applications. If it's not internal, it should probably carry
>>> some "xenomai" prefix in its (fairly generic) name.
>>>
>>
>> Sure, this service is aimed at being fairly generic and general-purpose,
>> with backtracing support, timing data, error code translation and such.
>> With respect to layering, I would rather expect an XML lib not to mess
>> with the namespace of core services, instead of asking the xenomai libs
>> which do deliver core services to apps to prefix its own services.
>>
>> I may move those helpers in some specific file to hide them, but no, I
>> really don't want to go for xenomai_panic(), because it really should
>> work the other way around (although nobody is likely to care about this
>> on the apache side).
>>
> 
> Actually, the only issue is about defining the panic helpers as macros,
> so that we can pull the source/line information automatically.
> 

We are not even doing that anymore, so this is eventually about moving
these helpers in some header that would not be pulled indirectly, or
using a different internal identifier from the reserved namespace.

-- 
Philippe.

Re: [Xenomai] panic macro interference with xerces api.

[Stéphane ANCELOT]

On 11/11/2015 13:35, Philippe Gerum wrote:
> On 11/11/2015 01:28 PM, Philippe Gerum wrote:
>> On 11/11/2015 12:52 PM, Philippe Gerum wrote:
>>> On 11/11/2015 11:31 AM, Jan Kiszka wrote:
>>>> On 2015-11-10 15:28, Gilles Chanteperdrix wrote:
>>>>> On Tue, Nov 10, 2015 at 02:26:54PM +0100, Stéphane ANCELOT wrote:
>>>>>> Hi,
>>>>>>
>>>>>> I don't know if there might be a  big impact, but I noticed, depending
>>>>>> on the order we include headers , that
>>>>>> there was a panic macro interfering with xercesc api headers.
>>>>>>
>>>>>> #include <alchemy/task.h>
>>>>>> #include <xercesc/dom/DOMException.hpp>
>>>>>>
>>>>>>
>>>>>> /usr/xenomai-3.0-igb/include/boilerplate/ancillaries.h:64:10: error:
>>>>>> expected identifier before '__func__'
>>>>>>     __panic(__func__, __fmt, ##__args)
>>>>> Well, if the two headers define the same macro, there are not many
>>>>> solutions: arrange for not having to include them in the same file.
>>>>> It should not be too hard, since a real-time thread using
>>>>> alchemy/task.h has no business using an xml library, probably only
>>>>> using linux services which a real-time task can not use without
>>>>> switching to secondary mode.
>>>> Isn't *panic an internal service? Then we can maybe avoid that it is
>>>> exposed to applications. If it's not internal, it should probably carry
>>>> some "xenomai" prefix in its (fairly generic) name.
>>>>
>>> Sure, this service is aimed at being fairly generic and general-purpose,
>>> with backtracing support, timing data, error code translation and such.
>>> With respect to layering, I would rather expect an XML lib not to mess
>>> with the namespace of core services, instead of asking the xenomai libs
>>> which do deliver core services to apps to prefix its own services.
>>>
>>> I may move those helpers in some specific file to hide them, but no, I
>>> really don't want to go for xenomai_panic(), because it really should
>>> work the other way around (although nobody is likely to care about this
>>> on the apache side).
>>>
>> Actually, the only issue is about defining the panic helpers as macros,
>> so that we can pull the source/line information automatically.
>>
> We are not even doing that anymore, so this is eventually about moving
> these helpers in some header that would not be pulled indirectly, or
> using a different internal identifier from the reserved namespace.
>
I ordered differently the headers to enable compilation.
I may arrange source code , since the xerces lib is needed only for data 
parsing  during the setup process.
Regards,
Steph

Re: [Xenomai] panic macro interference with xerces api.

[Philippe Gerum]

On 11/12/2015 08:37 AM, Stéphane ANCELOT wrote:
> On 11/11/2015 13:35, Philippe Gerum wrote:
>> On 11/11/2015 01:28 PM, Philippe Gerum wrote:
>>> On 11/11/2015 12:52 PM, Philippe Gerum wrote:
>>>> On 11/11/2015 11:31 AM, Jan Kiszka wrote:
>>>>> On 2015-11-10 15:28, Gilles Chanteperdrix wrote:
>>>>>> On Tue, Nov 10, 2015 at 02:26:54PM +0100, Stéphane ANCELOT wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> I don't know if there might be a  big impact, but I noticed,
>>>>>>> depending
>>>>>>> on the order we include headers , that
>>>>>>> there was a panic macro interfering with xercesc api headers.
>>>>>>>
>>>>>>> #include <alchemy/task.h>
>>>>>>> #include <xercesc/dom/DOMException.hpp>
>>>>>>>
>>>>>>>
>>>>>>> /usr/xenomai-3.0-igb/include/boilerplate/ancillaries.h:64:10: error:
>>>>>>> expected identifier before '__func__'
>>>>>>>     __panic(__func__, __fmt, ##__args)
>>>>>> Well, if the two headers define the same macro, there are not many
>>>>>> solutions: arrange for not having to include them in the same file.
>>>>>> It should not be too hard, since a real-time thread using
>>>>>> alchemy/task.h has no business using an xml library, probably only
>>>>>> using linux services which a real-time task can not use without
>>>>>> switching to secondary mode.
>>>>> Isn't *panic an internal service? Then we can maybe avoid that it is
>>>>> exposed to applications. If it's not internal, it should probably
>>>>> carry
>>>>> some "xenomai" prefix in its (fairly generic) name.
>>>>>
>>>> Sure, this service is aimed at being fairly generic and
>>>> general-purpose,
>>>> with backtracing support, timing data, error code translation and such.
>>>> With respect to layering, I would rather expect an XML lib not to mess
>>>> with the namespace of core services, instead of asking the xenomai libs
>>>> which do deliver core services to apps to prefix its own services.
>>>>
>>>> I may move those helpers in some specific file to hide them, but no, I
>>>> really don't want to go for xenomai_panic(), because it really should
>>>> work the other way around (although nobody is likely to care about this
>>>> on the apache side).
>>>>
>>> Actually, the only issue is about defining the panic helpers as macros,
>>> so that we can pull the source/line information automatically.
>>>
>> We are not even doing that anymore, so this is eventually about moving
>> these helpers in some header that would not be pulled indirectly, or
>> using a different internal identifier from the reserved namespace.
>>
> I ordered differently the headers to enable compilation.
> I may arrange source code , since the xerces lib is needed only for data
> parsing  during the setup process.

This change from the stable branch should prevent the conflict from
arising. Including boilerplate/ancillaries.h is now required to pull the
related definitions (no more indirect inclusion from the API headers),
which should limit the scope of conflict to direct users of the
ancillaries services.

http://git.xenomai.org/xenomai-3.git/commit/?h=stable-3.0.x&id=5c88a11c11e9a2b13be4da91624dc3f3ff323e46

-- 
Philippe.