On Fri, Sep 11, 2015 at 3:14 PM, Ruediger Pluem <[email protected]> wrote:
>
> On 09/11/2015 03:00 PM, [email protected] wrote:
>>
>> #include "httpd.h"
>> #include "http_main.h"
>> -#ifdef AP_NEED_SET_MUTEX_PERMS
>> -#include "unixd.h"
>> -#endif
>> -
>> -#if APR_HAVE_UNISTD_H
>> -#include <unistd.h> /* for getpid() */
>> -#endif
>> -
>> -#if HAVE_SYS_SEM_H
>> -#include <sys/shm.h>
>> -#if !defined(SHM_R)
>> -#define SHM_R 0400
>> -#endif
>> -#if !defined(SHM_W)
>> -#define SHM_W 0200
>> -#endif
>> -#endif
>>
>
> Why can the above be removed and how is this related to what you try to fix?
There is no Unix specific code (anymore?) in the module, these macros
(and getpid) are used nowhere.
This is moderatly related to the commit I agree (though Unixes vs
Windows somehow), that's why I also mentionned the change in the
commit message ("This commit also axes unused Unix specifics about
mutexes/semaphores.")...
>
>> #define AP_SLOTMEM_IS_PREGRAB(t) (t->desc.type & AP_SLOTMEM_TYPE_PREGRAB)
>> #define AP_SLOTMEM_IS_PERSIST(t) (t->desc.type & AP_SLOTMEM_TYPE_PERSIST)
>> @@ -59,6 +42,7 @@ typedef struct {
>>
>> struct ap_slotmem_instance_t {
>> char *name; /* per segment name */
>
> name now stores the filename. Maybe that should reflected in the comment?
It was already storing the filename before (the leading '/' was used
to shortcut ap_runtime_dir_relative() only).
But yes, the comment needs to be changed, will do it.
Thanks for the review.
