[Cocci] "virtual rule patch not supported"

[Cocci] "virtual rule patch not supported"

[Brice Goglin]

Hello,
I am running the kernel 3.9-rc2 coccicheck script on some of my module,
and I get the following message:
  virtual rule patch not supported
Things seem to work fine aside of this message. Anything to be worried
about?
I already upgraded to coccinelle 1.0-rc16 (from Debian experimental)
since rc12 didn't accept orplus.cocci. Anything else to upgrade to avoid
the message?
Thanks
Brice

[Cocci] "virtual rule patch not supported"

[Julia Lawall]

On Mon, 11 Mar 2013, Brice Goglin wrote:

> Hello,
> I am running the kernel 3.9-rc2 coccicheck script on some of my module,
> and I get the following message:
>   virtual rule patch not supported

I guess it doesn't crash?  It is possible that some semantic patch just
doesn't support the patch option.  I don't remember if it is possible to
specify several options, and to have them tried in order?  That could be a
nice feature.  If not, I think that there is an option that tries all of
the options in a fixed order.  Nicolas?

julia

> Things seem to work fine aside of this message. Anything to be worried
> about?
> I already upgraded to coccinelle 1.0-rc16 (from Debian experimental)
> since rc12 didn't accept orplus.cocci. Anything else to upgrade to avoid
> the message?
> Thanks
> Brice
>
> _______________________________________________
> Cocci mailing list
> Cocci at systeme.lip6.fr
> https://systeme.lip6.fr/mailman/listinfo/cocci
>

[Cocci] "virtual rule patch not supported"

[Nicolas Palix]

Hi,

On Wed, Mar 13, 2013 at 9:47 AM, Julia Lawall <julia.lawall@lip6.fr> wrote:
> On Mon, 11 Mar 2013, Brice Goglin wrote:
>
>> Hello,
>> I am running the kernel 3.9-rc2 coccicheck script on some of my module,
>> and I get the following message:
>>   virtual rule patch not supported
>
> I guess it doesn't crash?  It is possible that some semantic patch just
> doesn't support the patch option.  I don't remember if it is possible to
> specify several options, and to have them tried in order?  That could be a
> nice feature.  If not, I think that there is an option that tries all of
> the options in a fixed order.  Nicolas?
>

According to the cocci files, every coccicheck mode could not be proposed.
It is indeed some time impossible to propose a bug correction because
several options are available for a single bug report and the semantic
patch is not designed to choose one.

The current basic modes are: patch, report, context, org
The 'chain' mode tries the following mode is order: patch || report ||
context || org.
Recently, the 'rep+ctxt' mode have been added for report and context modes.

If you haven't specify the mode, it defaults to chain. The patch mode
is thus tried first.
If it fails, coccicheck falls back to the report mode.

The following command will give you the available modes for each cocci file
distributed with the Linux kernel

      find scripts/coccinelle -name *.cocci -exec grep -H virtual \{} \;

Hope it helps.

Regards.

> julia
>
>> Things seem to work fine aside of this message. Anything to be worried
>> about?
>> I already upgraded to coccinelle 1.0-rc16 (from Debian experimental)
>> since rc12 didn't accept orplus.cocci. Anything else to upgrade to avoid
>> the message?
>> Thanks
>> Brice
>>
>> _______________________________________________
>> Cocci mailing list
>> Cocci at systeme.lip6.fr
>> https://systeme.lip6.fr/mailman/listinfo/cocci
>>
> _______________________________________________
> Cocci mailing list
> Cocci at systeme.lip6.fr
> https://systeme.lip6.fr/mailman/listinfo/cocci



-- 
Nicolas Palix

[Cocci] "virtual rule patch not supported"

[Brice Goglin]

Le 15/03/2013 15:49, Nicolas Palix a ?crit :
> Hi,
>
> On Wed, Mar 13, 2013 at 9:47 AM, Julia Lawall <julia.lawall@lip6.fr> wrote:
>> On Mon, 11 Mar 2013, Brice Goglin wrote:
>>
>>> Hello,
>>> I am running the kernel 3.9-rc2 coccicheck script on some of my module,
>>> and I get the following message:
>>>   virtual rule patch not supported
>> I guess it doesn't crash?  It is possible that some semantic patch just
>> doesn't support the patch option.  I don't remember if it is possible to
>> specify several options, and to have them tried in order?  That could be a
>> nice feature.  If not, I think that there is an option that tries all of
>> the options in a fixed order.  Nicolas?
>>
> According to the cocci files, every coccicheck mode could not be proposed.
> It is indeed some time impossible to propose a bug correction because
> several options are available for a single bug report and the semantic
> patch is not designed to choose one.
>
> The current basic modes are: patch, report, context, org
> The 'chain' mode tries the following mode is order: patch || report ||
> context || org.
> Recently, the 'rep+ctxt' mode have been added for report and context modes.
>
> If you haven't specify the mode, it defaults to chain. The patch mode
> is thus tried first.
> If it fails, coccicheck falls back to the report mode.
>

In short, this warning is harmless ?

Should it be disabled in the kernel by default ? Most people that uses
coccicheck in the kernel don't know much about it, and they don't write
the rules directly, so they may be afraid by the warning just like me.

Brice

[Cocci] "virtual rule patch not supported"

[Nicolas Palix]

On Fri, Mar 15, 2013 at 4:03 PM, Brice Goglin <Brice.Goglin@inria.fr> wrote:
> Le 15/03/2013 15:49, Nicolas Palix a ?crit :
>> Hi,
>>
>> On Wed, Mar 13, 2013 at 9:47 AM, Julia Lawall <julia.lawall@lip6.fr> wrote:
>>> On Mon, 11 Mar 2013, Brice Goglin wrote:
>>>
>>>> Hello,
>>>> I am running the kernel 3.9-rc2 coccicheck script on some of my module,
>>>> and I get the following message:
>>>>   virtual rule patch not supported
>>> I guess it doesn't crash?  It is possible that some semantic patch just
>>> doesn't support the patch option.  I don't remember if it is possible to
>>> specify several options, and to have them tried in order?  That could be a
>>> nice feature.  If not, I think that there is an option that tries all of
>>> the options in a fixed order.  Nicolas?
>>>
>> According to the cocci files, every coccicheck mode could not be proposed.
>> It is indeed some time impossible to propose a bug correction because
>> several options are available for a single bug report and the semantic
>> patch is not designed to choose one.
>>
>> The current basic modes are: patch, report, context, org
>> The 'chain' mode tries the following mode is order: patch || report ||
>> context || org.
>> Recently, the 'rep+ctxt' mode have been added for report and context modes.
>>
>> If you haven't specify the mode, it defaults to chain. The patch mode
>> is thus tried first.
>> If it fails, coccicheck falls back to the report mode.
>>
>
> In short, this warning is harmless ?

Indeed. But you will have various kind of output.

>
> Should it be disabled in the kernel by default ? Most people that uses
> coccicheck in the kernel don't know much about it, and they don't write
> the rules directly, so they may be afraid by the warning just like me.

One may improve the handling of the chain mode, and skip unavailable
modes instead of falling back to the next one in case of error...

Do you volunteer ?

By the way, thank you for the report.

>
> Brice
>



-- 
Nicolas Palix

[Cocci] "virtual rule patch not supported"

[Nicolas Palix]

Alternatively, you could use 'make coccicheck MODE=report' as every
cocci file currently provide at least this mode
or we could swap patch and report for the chain mode.

 linux $  find scripts/coccinelle -name *.cocci | wc -l
      42
npalix at lig-yttrium linux $ find scripts/coccinelle -name *.cocci -exec
grep -H 'virtual context' \{} \; | wc -l
      36
npalix at lig-yttrium linux $ find scripts/coccinelle -name *.cocci -exec
grep -H 'virtual report' \{} \; | wc -l
      42
npalix at lig-yttrium linux $ find scripts/coccinelle -name *.cocci -exec
grep -H 'virtual org' \{} \; | wc -l
      40
npalix at lig-yttrium linux $ find scripts/coccinelle -name *.cocci -exec
grep -H 'virtual patch' \{} \; | wc -l
      24

On Fri, Mar 15, 2013 at 4:11 PM, Nicolas Palix <nicolas.palix@imag.fr> wrote:
> On Fri, Mar 15, 2013 at 4:03 PM, Brice Goglin <Brice.Goglin@inria.fr> wrote:
>> Le 15/03/2013 15:49, Nicolas Palix a ?crit :
>>> Hi,
>>>
>>> On Wed, Mar 13, 2013 at 9:47 AM, Julia Lawall <julia.lawall@lip6.fr> wrote:
>>>> On Mon, 11 Mar 2013, Brice Goglin wrote:
>>>>
>>>>> Hello,
>>>>> I am running the kernel 3.9-rc2 coccicheck script on some of my module,
>>>>> and I get the following message:
>>>>>   virtual rule patch not supported
>>>> I guess it doesn't crash?  It is possible that some semantic patch just
>>>> doesn't support the patch option.  I don't remember if it is possible to
>>>> specify several options, and to have them tried in order?  That could be a
>>>> nice feature.  If not, I think that there is an option that tries all of
>>>> the options in a fixed order.  Nicolas?
>>>>
>>> According to the cocci files, every coccicheck mode could not be proposed.
>>> It is indeed some time impossible to propose a bug correction because
>>> several options are available for a single bug report and the semantic
>>> patch is not designed to choose one.
>>>
>>> The current basic modes are: patch, report, context, org
>>> The 'chain' mode tries the following mode is order: patch || report ||
>>> context || org.
>>> Recently, the 'rep+ctxt' mode have been added for report and context modes.
>>>
>>> If you haven't specify the mode, it defaults to chain. The patch mode
>>> is thus tried first.
>>> If it fails, coccicheck falls back to the report mode.
>>>
>>
>> In short, this warning is harmless ?
>
> Indeed. But you will have various kind of output.
>
>>
>> Should it be disabled in the kernel by default ? Most people that uses
>> coccicheck in the kernel don't know much about it, and they don't write
>> the rules directly, so they may be afraid by the warning just like me.
>
> One may improve the handling of the chain mode, and skip unavailable
> modes instead of falling back to the next one in case of error...
>
> Do you volunteer ?
>
> By the way, thank you for the report.
>
>>
>> Brice
>>
>
>
>
> --
> Nicolas Palix



-- 
Nicolas Palix

[Cocci] "virtual rule patch not supported"

[Brice Goglin]

Le 15/03/2013 16:14, Nicolas Palix a ?crit :
> Alternatively, you could use 'make coccicheck MODE=report' as every
> cocci file currently provide at least this mode

Yes that works thanks.

> or we could swap patch and report for the chain mode.

[...]

> One may improve the handling of the chain mode, and skip unavailable
> modes instead of falling back to the next one in case of error...
>
> Do you volunteer ?
>

If it's just a matter of adding grep before run_cmd, the attached patch
seems to work.

But I don't have enough knowledge about all this to decide what's the
best thing to do among everything you said above.

Brice
 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: coccichain.patch
Type: text/x-patch
Size: 1854 bytes
Desc: not available
URL: <https://systeme.lip6.fr/pipermail/cocci/attachments/20130315/a868b193/attachment-0001.bin>

[Cocci] "virtual rule patch not supported"

[Julia Lawall]

On Fri, 15 Mar 2013, Nicolas Palix wrote:

> Alternatively, you could use 'make coccicheck MODE=report' as every
> cocci file currently provide at least this mode
> or we could swap patch and report for the chain mode.

If you do that, it should be the same as report, because every semantic
patch should provide report...

julia

>  linux $  find scripts/coccinelle -name *.cocci | wc -l
>       42
> npalix at lig-yttrium linux $ find scripts/coccinelle -name *.cocci -exec
> grep -H 'virtual context' \{} \; | wc -l
>       36
> npalix at lig-yttrium linux $ find scripts/coccinelle -name *.cocci -exec
> grep -H 'virtual report' \{} \; | wc -l
>       42
> npalix at lig-yttrium linux $ find scripts/coccinelle -name *.cocci -exec
> grep -H 'virtual org' \{} \; | wc -l
>       40
> npalix at lig-yttrium linux $ find scripts/coccinelle -name *.cocci -exec
> grep -H 'virtual patch' \{} \; | wc -l
>       24
>
> On Fri, Mar 15, 2013 at 4:11 PM, Nicolas Palix <nicolas.palix@imag.fr> wrote:
> > On Fri, Mar 15, 2013 at 4:03 PM, Brice Goglin <Brice.Goglin@inria.fr> wrote:
> >> Le 15/03/2013 15:49, Nicolas Palix a ?crit :
> >>> Hi,
> >>>
> >>> On Wed, Mar 13, 2013 at 9:47 AM, Julia Lawall <julia.lawall@lip6.fr> wrote:
> >>>> On Mon, 11 Mar 2013, Brice Goglin wrote:
> >>>>
> >>>>> Hello,
> >>>>> I am running the kernel 3.9-rc2 coccicheck script on some of my module,
> >>>>> and I get the following message:
> >>>>>   virtual rule patch not supported
> >>>> I guess it doesn't crash?  It is possible that some semantic patch just
> >>>> doesn't support the patch option.  I don't remember if it is possible to
> >>>> specify several options, and to have them tried in order?  That could be a
> >>>> nice feature.  If not, I think that there is an option that tries all of
> >>>> the options in a fixed order.  Nicolas?
> >>>>
> >>> According to the cocci files, every coccicheck mode could not be proposed.
> >>> It is indeed some time impossible to propose a bug correction because
> >>> several options are available for a single bug report and the semantic
> >>> patch is not designed to choose one.
> >>>
> >>> The current basic modes are: patch, report, context, org
> >>> The 'chain' mode tries the following mode is order: patch || report ||
> >>> context || org.
> >>> Recently, the 'rep+ctxt' mode have been added for report and context modes.
> >>>
> >>> If you haven't specify the mode, it defaults to chain. The patch mode
> >>> is thus tried first.
> >>> If it fails, coccicheck falls back to the report mode.
> >>>
> >>
> >> In short, this warning is harmless ?
> >
> > Indeed. But you will have various kind of output.
> >
> >>
> >> Should it be disabled in the kernel by default ? Most people that uses
> >> coccicheck in the kernel don't know much about it, and they don't write
> >> the rules directly, so they may be afraid by the warning just like me.
> >
> > One may improve the handling of the chain mode, and skip unavailable
> > modes instead of falling back to the next one in case of error...
> >
> > Do you volunteer ?
> >
> > By the way, thank you for the report.
> >
> >>
> >> Brice
> >>
> >
> >
> >
> > --
> > Nicolas Palix
>
>
>
> --
> Nicolas Palix
>

[Cocci] "virtual rule patch not supported"

[Julia Lawall]

Maybe the kernel should just use report mode by default?

julia

On Fri, 15 Mar 2013, Brice Goglin wrote:

> Le 15/03/2013 15:49, Nicolas Palix a ?crit :
> > Hi,
> >
> > On Wed, Mar 13, 2013 at 9:47 AM, Julia Lawall <julia.lawall@lip6.fr> wrote:
> >> On Mon, 11 Mar 2013, Brice Goglin wrote:
> >>
> >>> Hello,
> >>> I am running the kernel 3.9-rc2 coccicheck script on some of my module,
> >>> and I get the following message:
> >>>   virtual rule patch not supported
> >> I guess it doesn't crash?  It is possible that some semantic patch just
> >> doesn't support the patch option.  I don't remember if it is possible to
> >> specify several options, and to have them tried in order?  That could be a
> >> nice feature.  If not, I think that there is an option that tries all of
> >> the options in a fixed order.  Nicolas?
> >>
> > According to the cocci files, every coccicheck mode could not be proposed.
> > It is indeed some time impossible to propose a bug correction because
> > several options are available for a single bug report and the semantic
> > patch is not designed to choose one.
> >
> > The current basic modes are: patch, report, context, org
> > The 'chain' mode tries the following mode is order: patch || report ||
> > context || org.
> > Recently, the 'rep+ctxt' mode have been added for report and context modes.
> >
> > If you haven't specify the mode, it defaults to chain. The patch mode
> > is thus tried first.
> > If it fails, coccicheck falls back to the report mode.
> >
>
> In short, this warning is harmless ?
>
> Should it be disabled in the kernel by default ? Most people that uses
> coccicheck in the kernel don't know much about it, and they don't write
> the rules directly, so they may be afraid by the warning just like me.
>
> Brice
>
>