Skip to contentSkip to conversation

Finding usage of the statement "goto" with SmPL

Cocci SF Markus Elfring 2 participants 3 messages Dec 9, 2015

lore
Share & tools
lore
#0 lore raw

Hello,

I have tried another tiny SmPL script out.

@jumping@
identifier label;
@@
*goto label;

This test example works as expected.

elfring at Sonne:~/Projekte/Coccinelle/janitor> spatch.opt show_goto_usage1.cocci ~/Projekte/Linux/next-patched/drivers/block/zram/zram_drv.c ?
@@ -1424,7 +1400,6 @@ static int __init zram_init(void)

                ret = zram_add();
                mutex_unlock(&zram_index_mutex);
                if (ret < 0)
-                       goto out_error;
                num_devices--;
        }

Now I am trying also to generate other display formats for the corresponding data with a SmPL approach like the following.

?
@contains_goto@
identifier label, work;
position g_pos, pos;
type return_type;
@@
return_type work at pos(...)
{
... when any
goto label at g_pos;
... when any
}

@script:python collection depends on contains_goto@ fun << contains_goto.work;
places << contains_goto.pos;
go_to << contains_goto.label;
jump_at << contains_goto.g_pos;
@@
print("collection: " + "fun=" + fun + " | go_to=" + go_to) store_positions(fun, places, go_to, jump_at)
?

elfring@Sonne:~/Projekte/Coccinelle/janitor/jumping1> spatch.opt list_functions_with_goto1.cocci ~/Projekte/Linux/next-patched/drivers/block/zram/zram_drv.c ?
No result for this analysis!

Unfortunately, I get the impression that the SmPL rule does not get executed because the base rule "contains_goto" might not produce the desired data. How should this surprise be resolved?

Regards,
Markus

#1 lore raw
Quoted 50 lines · SF Markus Elfring

Hello,

I have tried another tiny SmPL script out.

@jumping@
identifier label;
@@
*goto label;

This test example works as expected.

elfring at Sonne:~/Projekte/Coccinelle/janitor> spatch.opt show_goto_usage1.cocci ~/Projekte/Linux/next-patched/drivers/block/zram/zram_drv.c ?

(quoted diff) +0 -1
@@ -1424,7 +1400,6 @@ static int __init zram_init(void)
                ret = zram_add();
                mutex_unlock(&zram_index_mutex);
                if (ret < 0)
-                       goto out_error;
                num_devices--;
        }
 

Now I am trying also to generate other display formats for the corresponding data with a SmPL approach like the following.

?
@contains_goto@
identifier label, work;
position g_pos, pos;
type return_type;
@@
return_type work at pos(...)
{
... when any
goto label at g_pos;
... when any
}

@script:python collection depends on contains_goto@ fun << contains_goto.work;
places << contains_goto.pos;
go_to << contains_goto.label;
jump_at << contains_goto.g_pos;
@@
print("collection: " + "fun=" + fun + " | go_to=" + go_to) store_positions(fun, places, go_to, jump_at)
?

Quoted 3 lines

elfring at Sonne:~/Projekte/Coccinelle/janitor/jumping1> spatch.opt list_functions_with_goto1.cocci ~/Projekte/Linux/next-patched/drivers/block/zram/zram_drv.c ?
No result for this analysis!

It doesn't match anything because when you write a rule that has no *, then it is considered that you want the pattern to match on every execution path. It is unlikely that the goto is on every execution path, so you get nothing. To solve the problem, just put exists in the initial @@

If you don't care about the position of the function header of the containing function, but could live with just the function name, you can access it much more efficiently from a position variable on the goto, in the field current_element.

julia

#2 lore raw
Quoted 4 lines

It doesn't match anything because when you write a rule that has no *, then it is considered that you want the pattern to match on every execution path. It is unlikely that the goto is on every execution path, so you get nothing.

This is true.

Quoted 1 line

To solve the problem, just put exists in the initial @@

Thanks for your quick clarification.

I'm sorry that I overlooked this SmPL script setting somehow. Now I get also the expected results for this SmPL approach.

function|"source file"|line|column|go_to|label_line|label_column __zram_make_request|"/home/elfring/Projekte/Linux/next-patched/drivers/block/zram/zram_drv.c"|841|13|out|874|10 ? zram_rw_page|"/home/elfring/Projekte/Linux/next-patched/drivers/block/zram/zram_drv.c"|937|12|put_zram|952|8

Quoted 2 lines

If you don't care about the position of the function header of the containing function, but could live with just the function name,

I reuse the information "line" and "column" from this metavariable intentionally.

Quoted 2 lines

you can access it much more efficiently from a position variable on the goto, in the field current_element.

I will consider such fine-tuning for other source code analysis.

Regards,
Markus