I've been testing the patch and it seems to work out allright :-)
A few notes;
#include <sys/stat.h>
must be outside #ifdef SunOS - other systems needs it too
in unlock_and_fclose(FILE *fp)
fclose is only called in #else, I suppose that should have been like for
unlock_and_close or we'll run out of filehandles ;-)
Best regards,
Christian
On Wed, Sep 10, 2008 at 10:05 PM, Jason Pollock <
[EMAIL PROTECTED]> wrote:
>
>
> Christian Theil Have wrote:
>
>> Hi
>>
>> I think it looks good :-) I was working on some similar code based on a
>> dict, which I was still testing out, but your patch is cleaner.
>> I look forward to trying it out, but I wont have the opportunity to test
>> it until Monday though.
>>
>> About changing close and fclose everywhere in the codebase, you could get
>> around this without having to change the rest the of codebase by using a few
>> macros , eg.
>>
>> #ifdef SunOS
>> #define close(fd) unlock_and_close(fd)
>> #endif
>>
>> ....
>>
>> int unlock_and_close(fd) {
>> #ifdef close(fd)
>> #undef close(fd)
>> printf("unlock_and_close do the stuff..\n");
>> return close(fd);
>> #define close(fd) unlock_and_close(fd)
>> #endif
>> }
>>
>> and similarly for fclose...
>>
>> Christian.
>>
>
> That would work. I wasn't sure if the Mbuni team would want to go that
> way, or the Kannel way:
>
> #define close(x) you_should_not_call_close_directly(x)
>
> I personally decided against the #define because I wanted to be able to
> check that all instances of close/fclose had been modified in the source I
> was working on. I did look at trying to figure out how to wrap the libc
> call, but I thought that might be a bit too destructive to my schedule. ;)
>
> I'm still new to the gwlib code, I didn't realise it had a dict! Too bad
> it uses octstr's as keys.
>
> Let me know if you run into any problems. If I find any, I'll let the list
> know.
>
> Jason
>
_______________________________________________
Devel mailing list
Devel@mbuni.org
http://lists.mbuni.org/mailman/listinfo/devel
