On Wed, May 11, 2011 at 12:28:29PM +0200, Michel Hermier 
<[email protected]> wrote:
> I just saw something *big*
>    case PM_OP_PS:      ret = pspkg(pm_targets);      break;
> and then the declaration is:
>    int pspkg();
> You should add the void or the pm_targets type to the call.

Fixed.

> (maybe all our
> () prototype should also be changed to (void) ?)

That's possible (kernel coding style requeres that, for example), but is
not a must.

> Smaller issues:
> * ps_parse lacks a static in the declaration.

Fixed.

> * lsof is executed using execvp which might be problematic since it
> exports all the environment, leading to potentioal locale issues at least.

That's not a problem, lsof is not localized and we parse a machine
output, which won't even contain English strings.

> * on execvp failure child process should die not return

Fixed.

> * The parent should also check the fdopen return, just in case the child
> was faster, and execvp failed (leading to an appempt to read a pipe
> without writer, which is an error)

Fxied.

> * end_lsof should be input values guarded (checking NULL file and 0 pid)

That's not necessary, start_lsof() guarantees these constrains as we
have checks for these there.

> * struct __ps_t don't have to be library public, and pid should be of
> pid_t type.

__ps_t is in the frontend at the moment, so it makes no sense to make it
private. We can do that if we move it to libpacman.

I fixed the int vs pid_t issue.

> I wait review for now, maybe some more on the next round ;)

Thanks a lot, pushed out. :)

Attachment: pgprIauIsjaPp.pgp
Description: PGP signature

_______________________________________________
Frugalware-devel mailing list
[email protected]
http://frugalware.org/mailman/listinfo/frugalware-devel

Reply via email to