[Cocci] Coccinelle 1.0.0-rc21

[Cocci] Coccinelle 1.0.0-rc21

[Julia Lawall]

Version 1.0.0-rc21 has been released.  The main changes are as follows:

* To a limited extent, it is possible to make changes within strings.  The
change has to be either to a coplete format part, or to a complete
non-format part.  It is not eg possible to change a few characters of the
non-format part.  An example is as follows:

@@
@@

"foo
-%d
+%s
 bar"

* It is now possible to wrap code, eg braces, around a conditional branch
that ends in a return. (***)

* The performance problem reported by Ning Jiang seems to be solved. (***)

* The handling of variable declarations that declare more than one
variable is slightly more robust.

* Pretty printing is slightly improved.

* The demos are now compatible with the addition of metavariables for
initialize and finalize rules, as done in rc20.

The two changes marked with *** required some significant changes to the
control-flow graph.  Please let me know about any major degradations in
performance, correctness, etc.

julia

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[SF Markus Elfring]

Am 14.04.2014 14:05, schrieb Julia Lawall:
> Version 1.0.0-rc21 has been released.

Which release will contain improvements for open issues which we discussed for
the software "Coccinelle 1.0.0-rc20"?

Regards,
Markus

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[Julia Lawall]

On Mon, 14 Apr 2014, SF Markus Elfring wrote:

> Am 14.04.2014 14:05, schrieb Julia Lawall:
> > Version 1.0.0-rc21 has been released.
>
> Which release will contain improvements for open issues which we discussed for
> the software "Coccinelle 1.0.0-rc20"?

You had a problem with including some ocaml libraries, which will be
worked on at some point in the near future.

You had a problem with temporary file names, which is addressed (cf
changes.html).

Someone else had a problem with constructors, which will also be addressed
shortly.

I don't think anything else was discussed which was within the current
goals and resources for the development of Coccinelle.

julia

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[SF Markus Elfring]

> I don't think anything else was discussed which was within the current
> goals and resources for the development of Coccinelle.

I'm also curious on improvements for the handling of regular expressions.   ;-)

Regards,
Markus

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[Julia Lawall]

On Mon, 14 Apr 2014, SF Markus Elfring wrote:

> > I don't think anything else was discussed which was within the current
> > goals and resources for the development of Coccinelle.
>
> I'm also curious on improvements for the handling of regular expressions.   ;-)

Nothing is planned for the handling of regular expressions.

julia

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[SF Markus Elfring]

> Nothing is planned for the handling of regular expressions.

How does this feedback fit to your response for a "regexp problem" on March 7th?

Regards,
Markus

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[Julia Lawall]

On Mon, 14 Apr 2014, SF Markus Elfring wrote:

> > Nothing is planned for the handling of regular expressions.
>
> How does this feedback fit to your response for a "regexp problem" on March 7th?

OK, I will look into the abstract value problem as well.

There was some proble about really huge regular expressions, which I think
must be a library problem, not a Coccinelle problem, so I don't plan to do
anything about it.

julia

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[SF Markus Elfring]

> There was some proble about really huge regular expressions, which I think
> must be a library problem, not a Coccinelle problem, so I don't plan to do
> anything about it.

Did you eventually reproduce exceptions from the handling of a bit smaller
alternation in a SmPL constraint once more?

How are the chances that if this one would be fixed that the bigger filter
pattern will also work?

Regards,
Markus

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[Julia Lawall]

On Mon, 14 Apr 2014, SF Markus Elfring wrote:

> > There was some proble about really huge regular expressions, which I think
> > must be a library problem, not a Coccinelle problem, so I don't plan to do
> > anything about it.
>
> Did you eventually reproduce exceptions from the handling of a bit smaller
> alternation in a SmPL constraint once more?

Yes, I believe so.

> How are the chances that if this one would be fixed that the bigger filter
> pattern will also work?

No chance.  The problems are completely unrelated.  In OCaml it is not
possible to test equality of certain kinds of values.  PCRE regular
expressions are this kind of value.  Str regular expressions can be
compared, so this problem did not show up until we added PCRE support.

julia

[Cocci] Coccinelle 1.0.0-rc21: Clarification for "coccilib.elems"?

[SF Markus Elfring]

> Nothing is planned for the handling of regular expressions.

I have built the current release on my openSUSE system.

I have tried out the following command again.

elfring at Sonne:~/Projekte/Coccinelle/janitor> date && ../1.0.0-rc21/spatch.opt
--sp-file list_functions_with_unnecessary_checks1.cocci --dir
/usr/src/linux-stable > list_functions_with_unnecessary_checks1-rc21.txt 2>
list_functions_with_unnecessary_checks1-rc21-errors.txt && date
Mo 14. Apr 18:18:16 CEST 2014


Unfortunately, I stumble on an error message in the shown log file.

init_defs_builtins: /usr/local/share/coccinelle/standard.h
ImportError: No module named coccilib.elems
while importing module coccilib.elems:
Failure in rule starting on line 1
Fatal error: exception Yes_pycocci.Pycocciexception


Why does it not work as expected?

Regards,
Markus

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[SF Markus Elfring]

>> Did you eventually reproduce exceptions from the handling of a bit smaller
>> alternation in a SmPL constraint once more?
> 
> Yes, I believe so.

How are the chances to get the filter pattern to work which was created with an
alternation of 893 function names?

Would you like to reduce the corresponding list of 217 failure situations around
the message 'EXN:Invalid_argument("equal: abstract value")'?

Regards,
Markus

[Cocci] Coccinelle 1.0.0-rc21

[Derek M Jones]

Julia,

> Version 1.0.0-rc21 has been released.  The main changes are as follows:

Have to take my shoes and socks off to count the release candidates!

Time for a real release :-)

-- 
Derek M. Jones                  tel: +44 (0) 1252 520 667
Knowledge Software Ltd          blog:shape-of-code.coding-guidelines.com
Software analysis               http://www.knosof.co.uk

[Cocci] Coccinelle 1.0.0-rc21: Clarification for "coccilib.elems"?

[SF Markus Elfring]

> Why does it not work as expected?

It seems that an installation is always needed. (No "shortcut" so far ... ?)

It would be nice if the desired command will also work from a test directory
directly.

Regards,
Markus

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[Julia Lawall]

On Mon, 14 Apr 2014, SF Markus Elfring wrote:

> >> Did you eventually reproduce exceptions from the handling of a bit smaller
> >> alternation in a SmPL constraint once more?
> >
> > Yes, I believe so.
>
> How are the chances to get the filter pattern to work which was created with an
> alternation of 893 function names?

I have already stated that this is not a priority.  First, it is not a
reasonable use of Cocinelle.  Second, the problem is likely not in
Coccinelle but in the regular expression library that is being used.  If
you want that fixed, you can contact the maintainers of that library.

> Would you like to reduce the corresponding list of 217 failure situations around
> the message 'EXN:Invalid_argument("equal: abstract value")'?

This I will try to address.

julia

[Cocci] Coccinelle 1.0.0-rc21

[Julia Lawall]

On Mon, 14 Apr 2014, Derek M Jones wrote:

> Julia,
>
> > Version 1.0.0-rc21 has been released.  The main changes are as follows:
>
> Have to take my shoes and socks off to count the release candidates!
>
> Time for a real release :-)

Actually, I have a small list of significant bugs from Nic that have to be
addressed first...

julia

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[SF Markus Elfring]

>> How are the chances to get the filter pattern to work which was created with an
>> alternation of 893 function names?
> 
> I have already stated that this is not a priority.

I can understand that.


> First, it is not a reasonable use of Cocinelle.

I find this feedback strange.

Static source code analysis on Linux files can result in higher and interesting
numbers for further consideration of corresponding transformations.


> Second, the problem is likely not in Coccinelle but in the regular expression library
> that is being used.

Which source code places are affected by this software development challenge?

Regards,
Markus

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[Julia Lawall]

On Tue, 15 Apr 2014, SF Markus Elfring wrote:

> >> How are the chances to get the filter pattern to work which was created with an
> >> alternation of 893 function names?
> >
> > I have already stated that this is not a priority.
>
> I can understand that.
>
>
> > First, it is not a reasonable use of Cocinelle.
>
> I find this feedback strange.
>
> Static source code analysis on Linux files can result in higher and interesting
> numbers for further consideration of corresponding transformations.

Yu should use virtual identifiers, not huge regular expressions.

> > Second, the problem is likely not in Coccinelle but in the regular expression library
> > that is being used.
>
> Which source code places are affected by this software development challenge?

I have no idea.   It is the PCRE library.

julia

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[SF Markus Elfring]

> Yu should use virtual identifiers, not huge regular expressions.

Unfortunately, "big" regular expressions can not be passed by virtual
identifiers to SmPL constraints so far.

Regards,
Markus

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[Julia Lawall]

On Tue, 15 Apr 2014, SF Markus Elfring wrote:

> > Yu should use virtual identifiers, not huge regular expressions.
>
> Unfortunately, "big" regular expressions can not be passed by virtual
> identifiers to SmPL constraints so far.

Th eidea is to pass the collected information one by on, eg using a make
file.  Not to use a regular expression at all.

julia

>
> Regards,
> Markus
>

[Cocci] Relevance of some regular expressions in SmPL

[SF Markus Elfring]

> Th eidea is to pass the collected information one by on, eg using a make
> file.  Not to use a regular expression at all.

Regular expressions can be used to some degree in semantic patches for a while.
I like that functionality in principle!

I do not want to pass small data items by virtual identifiers when I need a
complete information package.


I imagine that it would be nice if the discussed bigger SmPL constraints could
be included from an external file or could be implemented by a data base query.
I assume that such ideas for further software extensions will not help in the
near future.

Regards,
Markus

[Cocci] Relevance of some regular expressions in SmPL

[Julia Lawall]

> I imagine that it would be nice if the discussed bigger SmPL constraints could
> be included from an external file or could be implemented by a data base query.
> I assume that such ideas for further software extensions will not help in the
> near future.

You can do whatever you want in python or ocaml code. You can get
information from your database, and then call coccinelle from coccinelle
using the iteration mechanism (cf demos/iteration.cocci).

julia

[Cocci] Relevance of some regular expressions in SmPL

[SF Markus Elfring]

> You can get information from your database, and then call coccinelle from coccinelle
> using the iteration mechanism (cf demos/iteration.cocci).

I do not want to iterate over smaller elements when a SmPL constraint needs all
relevant information at once. The current SmPL constraint implementation will
not consult an external data base.

Regards,
Markus

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[Julia Lawall]

The problem of the exception due to comparing regular expressions has been 
resolved.  A new version with the fix will be available shortly.

julia

[Cocci] Road map after Coccinelle 1.0.0-rc21?

[SF Markus Elfring]

> The problem of the exception due to comparing regular expressions has been resolved.

That is a nice surprise.

Now I am curious when I can retry my list of update candidates ... ;-)

Regards,
Markus