Skip to contentSkip to conversation

checkpatch.pl rejects as error something I think it ought to be allow

BUG ? Corey Ashford 2 participants 4 messages Feb 21, 2011 – Feb 22, 2011

lore
Share & tools
lore
#0 lore raw

Hi,

I have a piece of code where I have two constants defined as follows:

	static const unsigned long polling_interval_sec = 1;
	static const unsigned long polling_interval_ns = 0;

Now, it's clear to me that I want these two values to have the keywords const and static. I could use a #define here, but const static seemed cleaner to me.

When I run checkpatch.pl across this code, I get this error:

ERROR: do not initialise statics to 0 or NULL.

I think the problem here is that another case is needed for "static const" that does allow 0.

What do you think?

Thanks for your consideration,

- Corey

#1 lore raw
Quoted 13 lines · Corey Ashford

Hi,

I have a piece of code where I have two constants defined as follows:

	static const unsigned long polling_interval_sec = 1;
	static const unsigned long polling_interval_ns = 0;

Now, it's clear to me that I want these two values to have the keywords const and static. I could use a #define here, but const static seemed cleaner to me.

When I run checkpatch.pl across this code, I get this error:

Quoted 8 lines

ERROR: do not initialise statics to 0 or NULL.

I think the problem here is that another case is needed for "static const" that does allow 0.

What do you think?

Thanks for your consideration,

The warning is intended to tell you that the = 0 is unnecessary. Any static is 0 by default I believe. At some point the addition of the 0 would move the value from the bss to the data segment bloating the code. This may no longer be true.

-apw

#2 lore raw
Quoted 19 lines · Andy Whitcroft
Quoted 18 lines · Corey Ashford

Hi,

I have a piece of code where I have two constants defined as follows:

	static const unsigned long polling_interval_sec = 1;
	static const unsigned long polling_interval_ns = 0;

Now, it's clear to me that I want these two values to have the keywords const and static. I could use a #define here, but const static seemed cleaner to me.

When I run checkpatch.pl across this code, I get this error:

ERROR: do not initialise statics to 0 or NULL.

I think the problem here is that another case is needed for "static const" that does allow 0.

Quoted 8 lines
Quoted 3 lines

What do you think?

Thanks for your consideration,

The warning is intended to tell you that the = 0 is unnecessary. Any static is 0 by default I believe. At some point the addition of the 0 would move the value from the bss to the data segment bloating the code. This may no longer be true.

OK, but that means I'd have to have a declaration like this, which looks quite odd to me:

static const poll_interval_ns; /* = 0 */

I don't think that is preferable to this:

	static const poll_interval_ns = 0;

- Corey

#3 lore raw
Quoted 29 lines · Corey Ashford
Quoted 19 lines · Andy Whitcroft
Quoted 18 lines · Corey Ashford

Hi,

I have a piece of code where I have two constants defined as follows:

	static const unsigned long polling_interval_sec = 1;
	static const unsigned long polling_interval_ns = 0;

Now, it's clear to me that I want these two values to have the keywords const and static. I could use a #define here, but const static seemed cleaner to me.

When I run checkpatch.pl across this code, I get this error:

ERROR: do not initialise statics to 0 or NULL.

I think the problem here is that another case is needed for "static const" that does allow 0.

Quoted 8 lines
Quoted 3 lines

What do you think?

Thanks for your consideration,

The warning is intended to tell you that the = 0 is unnecessary. Any static is 0 by default I believe. At some point the addition of the 0 would move the value from the bss to the data segment bloating the code. This may no longer be true.

Quoted 8 lines

OK, but that means I'd have to have a declaration like this, which looks quite odd to me:

static const poll_interval_ns; /* = 0 */

I don't think that is preferable to this:

	static const poll_interval_ns = 0;

Well I'd not bother with the comment as that is defined behaviour. But I suspect that gcc has gained the smarts to not break these and perhaps the check is long in the tooth. As for this patch, as checkpatch is intended to be an advisory helper you are free to ignore its recommendations, just expect to be asked about it by any human reviewers.

-apw