[CalendarServer-users] calendarserver 3.2 on Debian unstable throwing error "socket.error: [Errno 22] Invalid argument" out of the box?

[Fredrik Unger]

Hi,

While looking into sendmsg implementation I found a python implementation :

http://hg.python.org/cpython/file/default/Modules/socketmodule.c#l3478

I found it via http://bugs.python.org/issue6560 and it seems to be for python 3 only?

"> Also is there any remote chance for a backport to 2.7?

I'd be happy to do one, but I'm pretty sure python.org's 2.x line
is closed to new features.  Perhaps some fork of CPython might be
willing to accept it, though - I don't know."

Maybe one could use this instead ? 
In anyway I will use try to compare the two and try to find the problem.
But as I said, I might not get to it the next two weeks.

/Fred


On Tue, 17 Jul 2012 10:04:37 +0200
Fredrik Unger <fred at tree.se> wrote:

> 
> Hi,
> 
> Here is a first shot. Maybe in the dark..
> 
> > Thanks for this investigation. So far it sounds like everything is
> > right, but I will follow your further analysis with interest.
> 
> One problem could be socklen_t. Linux uses size_t [1].
> (yes I read the comment in the code "To forestall portability problems")
> (I do not know Mac:s structs, also 32 bit probably size_t and socklen_t 
> are 4 )
> I have not enough Python/C interface knowledge to know what dragons
> are hidden there..
> 
> For example the cast :
> message_header.msg_controllen = (socklen_t) all_data_len;
> 
> puts an int in the msg.controllen but an long long is expected?
> That cast might not be as bad as setting some length using &.
> 
> Under linux[2] here, the sizeof is :
> sizeof socklen_t 4
> sizeof size_t 8
> 
> There are several casts (socklen_t) and removing them did not
> yet solve the problem. There must be some other error as well.
> The transformer t# means char* + int. I added a tmp varible for iovec
> as it is also size_t[3].
> 
> There are several points and I have tried to fix things, but was
> not yet successful.
> 
> I need to read up on and create a working under linux sendmsg I think to 
> have something to compare to.
> Rebuilding the full calendar server for each test does not make sense.
> 
> > Please feel free to bother me until I respond when you do :).
> 
> Hehe, leaving for vacation today, so I might not be able to follow up 
> for a while.
> 
> > Given that the problem thus far has been in some C wrapping for the
> > standard APIs, I'd rather figure out what the existing problem is than
> > introduce a new dependency and then write a bunch of _new_ C code
> > wrapping those APIs.
> 
> Sure, I understand. The above library also does not seem to be actively 
> developed.
> 
> [1]
> http://www.kernel.org/doc/man-pages/online/pages/man2/send.2.html
> 
> struct msghdr {
>         void         *msg_name;       /* optional address */
>         socklen_t     msg_namelen;    /* size of address */
>         struct iovec *msg_iov;        /* scatter/gather array */
>         size_t        msg_iovlen;     /* # elements in msg_iov */
>         void         *msg_control;    /* ancillary data, see below */
>         size_t        msg_controllen; /* ancillary data buffer len */
>         int           msg_flags;      /* flags on received message */
> };
> 
> http://pubs.opengroup.org/onlinepubs/007908775/xns/syssocket.h.html
> void         *msg_name        optional address
> socklen_t     msg_namelen     size of address
> struct iovec *msg_iov         scatter/gather array
> int           msg_iovlen      members in msg_iov
> void         *msg_control     ancillary data, see below
> socklen_t     msg_controllen  ancillary data buffer len
> int           msg_flags       flags on received message
> 
> [2]
> #include <sys/socket.h>
> #include <stdio.h>
> 
> main() {
> socklen_t t;
> size_t s;
> printf("sizeof socklen_t %d\n", sizeof(t));
> printf("sizeof size_t %d\n", sizeof(s));
> }
> 
> [3]
> struct iovec
>    {
>      void *iov_base;     /* Pointer to data.  */
>      size_t iov_len;     /* Length of data.  */
>    };
> _______________________________________________
> calendarserver-users mailing list
> calendarserver-users at lists.macosforge.org
> http://lists.macosforge.org/mailman/listinfo/calendarserver-users