[cocci] Bug with '*' to display a match on the subsequent rule.

[cocci] Bug with '*' to display a match on the subsequent rule.

[Eric Wheeler]

Hello,

I may have found a bug:

	@ r1 @
	type T;
	identifier I;
	@@
	*T I;

	@ r2 @
	@@
	f()
	{
	}
	+void a(){}

which prints this error:

	init_defs_builtins: /usr/local/coccinelle-git/bin/lib/coccinelle/standard.h
	File "cocci/broken.cocci", line 12, column 0, charpos = 57
	  around = '+',
	  whole content = +void a(){}

but if I remove the '*' or turn it to a '-' then it works:

	@ r1 @
	type T;
	identifier I;
	@@
	-T I;

	@ r2 @
	@@
	f()
	{
	}
	+void a(){}

with the diff:

	--- t.c
	+++ /tmp/cocci-output-6974-08cd95-t.c
	@@ -5,11 +5,13 @@ typedef struct
		int w;
	 } A;
	 
	-A a, b;
	-
	 #define NOMATCH(s1, s2, member) printf(#s1  "." #member "!=" #s2 "." #member);
	 f()
	 {
	 }
	 
	+void a()
	+{
	+}
	+
	 // functions here


Here is the t.c file:

	typedef struct 
	{
		int x, y, z;
		int u, v;
		int w;
	} A;

	A a, b;

	#define NOMATCH(s1, s2, member) printf(#s1  "." #member "!=" #s2 "." #member);
	f()
	{
	}

	// functions here

--
Eric Wheeler

Re: [cocci] Bug with '*' to display a match on the subsequent rule.

[Julia Lawall]

On Fri, 18 Mar 2022, Eric Wheeler wrote:

> Hello,
>
> I may have found a bug:
>
> 	@ r1 @
> 	type T;
> 	identifier I;
> 	@@
> 	*T I;
>
> 	@ r2 @
> 	@@
> 	f()
> 	{
> 	}
> 	+void a(){}
>
> which prints this error:
>
> 	init_defs_builtins: /usr/local/coccinelle-git/bin/lib/coccinelle/standard.h
> 	File "cocci/broken.cocci", line 12, column 0, charpos = 57
> 	  around = '+',
> 	  whole content = +void a(){}

You can't mix * and + in a single semantic patch.  The patch you get would
be strange.  There would be a + for the code that you want to add and a -
for the code that you want to highlight.  If you want to make changes and
find out about some things, then you can use some python code to print out
whatever you wnt.

julia

>
> but if I remove the '*' or turn it to a '-' then it works:
>
> 	@ r1 @
> 	type T;
> 	identifier I;
> 	@@
> 	-T I;
>
> 	@ r2 @
> 	@@
> 	f()
> 	{
> 	}
> 	+void a(){}
>
> with the diff:
>
> 	--- t.c
> 	+++ /tmp/cocci-output-6974-08cd95-t.c
> 	@@ -5,11 +5,13 @@ typedef struct
> 		int w;
> 	 } A;
>
> 	-A a, b;
> 	-
> 	 #define NOMATCH(s1, s2, member) printf(#s1  "." #member "!=" #s2 "." #member);
> 	 f()
> 	 {
> 	 }
>
> 	+void a()
> 	+{
> 	+}
> 	+
> 	 // functions here
>
>
> Here is the t.c file:
>
> 	typedef struct
> 	{
> 		int x, y, z;
> 		int u, v;
> 		int w;
> 	} A;
>
> 	A a, b;
>
> 	#define NOMATCH(s1, s2, member) printf(#s1  "." #member "!=" #s2 "." #member);
> 	f()
> 	{
> 	}
>
> 	// functions here
>
> --
> Eric Wheeler
>

Re: [cocci] Bug with '*' to display a match on the subsequent rule.

[Markus Elfring]

> You can't mix * and + in a single semantic patch.  The patch you get would
> be strange.  There would be a + for the code that you want to add and a -
> for the code that you want to highlight.


How do you think to provide a better error message from the Coccinelle software
for a questionable mixture of SmPL data processing indicators?

Regards,
Markus

Re: [cocci] Bug with '*' to display a match on the subsequent rule.

[Eric Wheeler]

On Sat, 19 Mar 2022, Markus Elfring wrote: 
> > You can't mix * and + in a single semantic patch.  The patch you get would
> > be strange.  There would be a + for the code that you want to add and a -
> > for the code that you want to highlight.
> 
> 
> How do you think to provide a better error message from the Coccinelle software
> for a questionable mixture of SmPL data processing indicators?

Its true, that would be nice.  I spent 20 minutes trying to figure out 
what I had wrong with my SmPL.  I had added the * for debugging and forgot 
about it.  After reducing it down to nothing it became clear that '*' was 
the issue.

A friendly error like "You can't mix * and + in a single semantic patch.  
The patch you get would be strange" would be useful ;)

--
Eric Wheeler



> 
> Regards,
> Markus
> 
> 

Re: [cocci] Bug with '*' to display a match on the subsequent rule.

[Julia Lawall]

On Sat, 19 Mar 2022, Eric Wheeler wrote:

> On Sat, 19 Mar 2022, Markus Elfring wrote:
> > > You can't mix * and + in a single semantic patch.  The patch you get would
> > > be strange.  There would be a + for the code that you want to add and a -
> > > for the code that you want to highlight.
> >
> >
> > How do you think to provide a better error message from the Coccinelle software
> > for a questionable mixture of SmPL data processing indicators?
>
> Its true, that would be nice.  I spent 20 minutes trying to figure out
> what I had wrong with my SmPL.  I had added the * for debugging and forgot
> about it.  After reducing it down to nothing it became clear that '*' was
> the issue.
>
> A friendly error like "You can't mix * and + in a single semantic patch.
> The patch you get would be strange" would be useful ;)

This has been fixed.  It resulted from a well intentioned but incorrect
attempt to provide file names and line numbers for this kind of problem,
which resulted in discarding the actual error message.

julia

Re: [cocci] Improved error handling for mixed SmPL data processing indicators

[Markus Elfring]

>> A friendly error like "You can't mix * and + in a single semantic patch.
>> The patch you get would be strange" would be useful ;)
> This has been fixed.  It resulted from a well intentioned but incorrect
> attempt to provide file names and line numbers for this kind of problem,
> which resulted in discarding the actual error message.


Does the commit “don't throw away the message for a lexical parse error”
belong to corresponding software adjustments?
https://gitlab.inria.fr/coccinelle/coccinelle/-/commit/43f8e0e3c195870cb43548e5bd350aaa720f2561

How do you think about to add a test case for this issue?

Would you like to achieve any improvements for the software documentation?

* https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/20fdb67f4b20a242f222337e13091115884cf6bb/docs/manual/cocci_syntax.tex#L1207
  https://github.com/coccinelle/coccinelle/blob/ae337fce1512ff15aabc3ad5b6d2e537f97ab62a/docs/manual/cocci_syntax.tex#L1207

* https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/20fdb67f4b20a242f222337e13091115884cf6bb/docs/manual/cocci_syntax.tex#L1095
  https://github.com/coccinelle/coccinelle/blob/ae337fce1512ff15aabc3ad5b6d2e537f97ab62a/docs/manual/cocci_syntax.tex#L1095

Regards,
Markus

Re: [cocci] Improved error handling for mixed SmPL data processing indicators

[Julia Lawall]

[-- Attachment #1: Type: text/plain, Size: 1415 bytes --]



On Sun, 20 Mar 2022, Markus Elfring wrote:

> >> A friendly error like "You can't mix * and + in a single semantic patch.
> >> The patch you get would be strange" would be useful ;)
> > This has been fixed.  It resulted from a well intentioned but incorrect
> > attempt to provide file names and line numbers for this kind of problem,
> > which resulted in discarding the actual error message.
>
>
> Does the commit “don't throw away the message for a lexical parse error”
> belong to corresponding software adjustments?
> https://gitlab.inria.fr/coccinelle/coccinelle/-/commit/43f8e0e3c195870cb43548e5bd350aaa720f2561
>
> How do you think about to add a test case for this issue?
>
> Would you like to achieve any improvements for the software documentation?

Th documentation already says that * cannot be mixed with - and +.  What
more do you want?

julia

>
> * https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/20fdb67f4b20a242f222337e13091115884cf6bb/docs/manual/cocci_syntax.tex#L1207
>   https://github.com/coccinelle/coccinelle/blob/ae337fce1512ff15aabc3ad5b6d2e537f97ab62a/docs/manual/cocci_syntax.tex#L1207
>
> * https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/20fdb67f4b20a242f222337e13091115884cf6bb/docs/manual/cocci_syntax.tex#L1095
>   https://github.com/coccinelle/coccinelle/blob/ae337fce1512ff15aabc3ad5b6d2e537f97ab62a/docs/manual/cocci_syntax.tex#L1095
>
> Regards,
> Markus
>
>

Re: [cocci] Improved error handling for mixed SmPL data processing indicators

[Markus Elfring]

>> Would you like to achieve any improvements for the software documentation?
> Th documentation already says that * cannot be mixed with - and +.


This information is available.



> What more do you want?


I imagine that tweaks for text formatting and layout adjustments can help
to read corresponding data better.

Regards,
Markus

Re: [cocci] Improving descriptions for SmPL data processing indicators?

[Markus Elfring]

>> What more do you want?


I propose to explain SmPL data processing indicators/annotations better by
a corresponding taxonomy.
Will such a documentation approach trigger a different visual representation?

Regards,
Markus