Skip to contentSkip to conversation

Replacing calls of general functions by more specific ones with SmPL

Cocci SF Markus Elfring 2 participants 23 messages Oct 14, 2016 – Oct 16, 2016

lore
Share & tools
lore
#0 lore raw

Hello,

I came during my software development activities along a few Linux source files where I got an advice by the script "checkpatch.pl" like the following.

WARNING: Prefer seq_puts to seq_printf

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/scripts/checkpatch.pl?id=29fbff8698fc0ac1a1d74584b258e0bf18b469f9#n5610

Joe Perches pointed out that the function "seq_putc" would be more appropriate at a specific source code place.
https://patchwork.kernel.org/patch/9364455/
https://lkml.kernel.org/r/<[email protected]>

Now I am curious on how good the semantic patch language is applicable for advanced source code transformations in similar use cases.
With which metavariables and corresponding programming interfaces can strings that are passed to such functions be checked in more detail by the Coccinelle software?

Regards,
Markus

#1 lore raw
Quoted 16 lines · SF Markus Elfring

Hello,

I came during my software development activities along a few Linux source files where I got an advice by the script "checkpatch.pl" like the following.

WARNING: Prefer seq_puts to seq_printf

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/scripts/checkpatch.pl?id=29fbff8698fc0ac1a1d74584b258e0bf18b469f9#n5610

Joe Perches pointed out that the function "seq_putc" would be more appropriate at a specific source code place.
https://patchwork.kernel.org/patch/9364455/
https://lkml.kernel.org/r/<[email protected]>

Quoted 4 lines

Now I am curious on how good the semantic patch language is applicable for advanced source code transformations in similar use cases.
With which metavariables and corresponding programming interfaces can strings that are passed to such functions be checked in more detail by the Coccinelle software?

The simplest would be to use python or ocaml code to make the checks that are needed. You can use cocci.include_match(False) in python or Coccilib.include_match false in ocaml when your test detects that something is not suitable. coccinelle/demos/first.cocci may be helpful.

julia

#2 lore raw
Quoted 7 lines
Quoted 2 lines

With which metavariables and corresponding programming interfaces can strings that are passed to such functions be checked in more detail by the Coccinelle software?

The simplest would be to use python or ocaml code to make the checks that are needed. You can use cocci.include_match(False) in python or Coccilib.include_match false in ocaml when your test detects that something is not suitable.

Thanks for your suggestion.

Quoted 1 line

coccinelle/demos/first.cocci may be helpful.

I have got the impression that this small SmPL script demonstrates the usage of the mentioned function from your software library only for a simple position variable. I do not see the way so far how desired data could be extracted from a function (or expression?) parameter. Would you like to make the introspection options a bit clearer for the SmPL interface?

Regards,
Markus

#3 lore raw
Quoted 7 lines
Quoted 1 line

coccinelle/demos/first.cocci may be helpful.

I have got the impression that this small SmPL script demonstrates the usage of the mentioned function from your software library only for a simple position variable. I do not see the way so far how desired data could be extracted from a function (or expression?) parameter. Would you like to make the introspection options a bit clearer for the SmPL interface?

You can inherit and test properties of anything you like in python or ocaml code. If you inherit some kind of metavariable other than a position variable, the term will be represented as a string. But that should be fine for detecting eg if a string has only one character. If it has multiple characters, you can do the include match false call, so that in the last rule when yu inherit again the metavariable containing th string, it won't be seen.

julia

#4 lore raw
Quoted 4 lines

You can inherit and test properties of anything you like in python or ocaml code. If you inherit some kind of metavariable other than a position variable, the term will be represented as a string. But that should be fine for detecting eg if a string has only one character.

Your explanation sounds promising.

Quoted 3 lines

If it has multiple characters, you can do the include match false call, so that in the last rule when yu inherit again the metavariable containing th string, it won't be seen.

I see another software development challenge then.

* I imagine that this "false call" will work for the exclusion of unwanted source code fragments.

* But I will also need a way to construct a source code change ("plus code" in the semantic patch) with OCaml or Python scripting, won't I?

Regards,
Markus

#5 lore raw
Quoted 2 lines

* But I will also need a way to construct a source code change ("plus code" in the semantic patch) with OCaml or Python scripting, won't I?

If you need to do this, you can follow the examples in demos/camltococci.cocci and demos/pythontococci.cocci

julia

#6 lore raw
In reply to Julia Lawall
Quoted 1 line

But that should be fine for detecting eg if a string has only one character.

Did you come along similar change possibilities at other places already?

http://lxr.free-electrons.com/source/include/linux/seq_file.h?v=4.8#L114

?????????????????????????????????????????????????????? ? update candidate ? 1. replacement ? 2. replacement ? ??????????????????????????????????????????????????????
? seq_printf ? seq_puts ? seq_putc ?
??????????????????????????????????????????????????????

Would it make sense to consider any more source code search patterns here?

Regards,
Markus

#7 lore raw
Quoted 6 lines · SF Markus Elfring
Quoted 1 line

But that should be fine for detecting eg if a string has only one character.

Did you come along similar change possibilities at other places already?

http://lxr.free-electrons.com/source/include/linux/seq_file.h?v=4.8#L114

Quoted 8 lines

?????????????????????????????????????????????????????? ? update candidate ? 1. replacement ? 2. replacement ? ??????????????????????????????????????????????????????
? seq_printf ? seq_puts ? seq_putc ?
??????????????????????????????????????????????????????

Would it make sense to consider any more source code search patterns here?

I'm not familiar with the details of this issue.

julia

#8 lore raw
Quoted 3 lines
Quoted 1 line

Would it make sense to consider any more source code search patterns here?

I'm not familiar with the details of this issue.

Have you heard about that any other Linux developers dare to think about further collateral software evolution around a function like "seq_printf"?

How interesting can it become to take also the POSIX variants into account?

Regards,
Markus

#9 lore raw
Quoted 8 lines · SF Markus Elfring
Quoted 3 lines
Quoted 1 line

Would it make sense to consider any more source code search patterns here?

I'm not familiar with the details of this issue.

Have you heard about that any other Linux developers dare to think about further collateral software evolution around a function like "seq_printf"?

How interesting can it become to take also the POSIX variants into account?

Same answer as above :) I have no idea.

You can look at patches people have previously submitted and see to what extent there is interest in this issue.

julia

#10 lore raw
Quoted 2 lines

You can look at patches people have previously submitted and see to what extent there is interest in this issue.

I see another software development challenge here.

The discussed source code search pattern builds on a "general function" like "seq_printf" as an update candidate from which each instance (probably with a different prefix) is connected with a few replacement possibilities. Did any other software developer fiddle with corresponding combinations of function names?

Regards,
Markus

#11 lore raw
Quoted 10 lines · SF Markus Elfring
Quoted 2 lines

You can look at patches people have previously submitted and see to what extent there is interest in this issue.

I see another software development challenge here.

The discussed source code search pattern builds on a "general function" like "seq_printf" as an update candidate from which each instance (probably with a different prefix) is connected with a few replacement possibilities. Did any other software developer fiddle with corresponding combinations of function names?

No idea. Look at the code and see what is available.

julia

#12 lore raw
Quoted 4 lines
Quoted 2 lines

Did any other software developer fiddle with corresponding combinations of function names?

No idea. Look at the code and see what is available.

Is it interesting to handle more than one combination efficiently?

Is a loop needed for corresponding record sets?

Regards,
Markus

#13 lore raw
In reply to Julia Lawall
Quoted 6 lines
Quoted 4 lines

Have you heard about that any other Linux developers dare to think about further collateral software evolution around a function like "seq_printf"?

How interesting can it become to take also the POSIX variants into account?

Same answer as above :) I have no idea.

Do you eventually know any source code search databases which could be queried for function names with the suffix "_printf" (or similar wordings)?

Regards,
Markus

#14 lore raw
In reply to Julia Lawall
Quoted 5 lines
Quoted 3 lines

I do not see the way so far how desired data could be extracted from a function (or expression?) parameter. Would you like to make the introspection options a bit clearer for the SmPL interface?

You can inherit and test properties of anything you like in python or ocaml code.

I have got an SmPL approach working.

Quoted 2 lines

If you inherit some kind of metavariable other than a position variable, the term will be represented as a string.

This software design aspect might look sufficient for a basic variant of the discussed source code search pattern.

Quoted 1 line

But that should be fine for detecting eg if a string has only one character.

I have got a few concerns for further software development considerations. How do you think about to support also an advanced interface by which it could be determined if a preprocessor symbol would correspond to such a desired character (instead of fiddling with regular expressions within a rule for the embedded programming language script)?

Regards,
Markus

#15 lore raw
Quoted 16 lines · SF Markus Elfring
Quoted 5 lines
Quoted 3 lines

I do not see the way so far how desired data could be extracted from a function (or expression?) parameter. Would you like to make the introspection options a bit clearer for the SmPL interface?

You can inherit and test properties of anything you like in python or ocaml code.

I have got an SmPL approach working.

Quoted 2 lines

If you inherit some kind of metavariable other than a position variable, the term will be represented as a string.

This software design aspect might look sufficient for a basic variant of the discussed source code search pattern.

Quoted 7 lines
Quoted 1 line

But that should be fine for detecting eg if a string has only one character.

I have got a few concerns for further software development considerations. How do you think about to support also an advanced interface by which it could be determined if a preprocessor symbol would correspond to such a desired character (instead of fiddling with regular expressions within a rule for the embedded programming language script)?

What C code are you concerned about exactly?

julia

#16 lore raw
Quoted 7 lines
Quoted 5 lines

I have got a few concerns for further software development considerations. How do you think about to support also an advanced interface by which it could be determined if a preprocessor symbol would correspond to such a desired character (instead of fiddling with regular expressions within a rule for the embedded programming language script)?

What C code are you concerned about exactly?

I guess that this "story" affects some general aspects. After I got a better understanding of related software dependencies, I begin to think also about further possibilities to make such a source code search pattern safer and more efficient.
How much would you like to increase the support for data processing around macros with your software?

Regards,
Markus

#17 lore raw
Quoted 6 lines · SF Markus Elfring
Quoted 6 lines
Quoted 5 lines

I have got a few concerns for further software development considerations. How do you think about to support also an advanced interface by which it could be determined if a preprocessor symbol would correspond to such a desired character (instead of fiddling with regular expressions within a rule for the embedded programming language script)?

Quoted 7 lines
Quoted 1 line

What C code are you concerned about exactly?

I guess that this "story" affects some general aspects. After I got a better understanding of related software dependencies, I begin to think also about further possibilities to make such a source code search pattern safer and more efficient.
How much would you like to increase the support for data processing around macros with your software?

Ther are no particular plans in this direction.

julia

#18 lore raw
Quoted 4 lines
Quoted 2 lines

How much would you like to increase the support for data processing around macros with your software?

Ther are no particular plans in this direction.

How does this kind of feedback fit to an information like ?Add identifier list metavariables for #define parameter lists? from the recent software release?

Regards,
Markus

#19 lore raw
Quoted 7 lines · SF Markus Elfring
Quoted 4 lines
Quoted 2 lines

How much would you like to increase the support for data processing around macros with your software?

Ther are no particular plans in this direction.

How does this kind of feedback fit to an information like ?Add identifier list metavariables for #define parameter lists? from the recent software release?

That was done already, so it's not a plan. Furthermore, it was not planned for, the need arose. Need, not hypothetical need.

julia

#20 lore raw
In reply to Julia Lawall
Quoted 4 lines

The simplest would be to use python or ocaml code to make the checks that are needed. You can use cocci.include_match(False) in python or Coccilib.include_match false in ocaml when your test detects that something is not suitable.

Will it occasionally matter then to distinguish better if a specific source code fragment or a complete source file should be excluded from an analysis run on demand?

Regards,
Markus

#21 lore raw
Quoted 8 lines · SF Markus Elfring
Quoted 4 lines

The simplest would be to use python or ocaml code to make the checks that are needed. You can use cocci.include_match(False) in python or Coccilib.include_match false in ocaml when your test detects that something is not suitable.

Will it occasionally matter then to distinguish better if a specific source code fragment or a complete source file should be excluded from an analysis run on demand?

I have no idea what you mean by this. But since I know very little about this printing problem in general, I doubt that I can give an answer. Try it and see what requirements turn up.

julia

#22 lore raw
Quoted 5 lines
Quoted 3 lines

Will it occasionally matter then to distinguish better if a specific source code fragment or a complete source file should be excluded from an analysis run on demand?

I have no idea what you mean by this.

I assume that there are more chances.

I am just unsure around the application of a function call like "cocci.include_match(False)" so far.

Quoted 1 line

But since I know very little about this printing problem in general,

It is not a "problem". - Is it just the beginning for another software development challenge (or adventure)?

Quoted 1 line

I doubt that I can give an answer.

I imagine that we can exchange some ideas to improve the affected software situation further, can't we?

Quoted 1 line

Try it and see what requirements turn up.

Which items will be picked up for further clarification?

Regards,
Markus