[9fans] mkstemp()

[9fans] mkstemp()

[cej]

Thank you all for patience with my questions, and for help!
 
++pac.
 

[9fans] mkstemp()

[cej]

need mkstemp() [BSD unistd.h]
any better idea than to copy the code from BSD | GNU ?
(I understand that returns a fd of a temp file)
 
thanks,
++pac.
 

Re: [9fans] mkstemp()

[Gorka Guardiola Múzquiz]

Creating a name with a random number in /tmp like /tmp/acme.paurea.987987 
wouldn't suffice?. Given we have private namespaces and random numbers
normally don't collide /tmp/acme.234234 or even /tmp/23234234 would probably 
suffice too.
----- Original Message ----- 
From: <cej@gli.cas.cz>
To: <9fans@cse.psu.edu>
Sent: Friday, January 28, 2005 2:39 PM
Subject: [9fans] mkstemp()


> need mkstemp() [BSD unistd.h]
> any better idea than to copy the code from BSD | GNU ?
> (I understand that returns a fd of a temp file)
>
> thanks,
> ++pac.
>
> 

Re: [9fans] mkstemp()

[Scott Schwartz]

| Creating a name with a random number in /tmp like /tmp/acme.paurea.987987 
| wouldn't suffice?. Given we have private namespaces and random numbers
| normally don't collide /tmp/acme.234234 or even /tmp/23234234 would probably 
| suffice too.

But still, it's nice to have a general solution.  Sometimes you do need
to create a file in a shared space.  One could even imagine a fileserver
that filled in the filename for you, rather like the special purpose
ones under /net.  In general you probably don't want to reserve the name
"clone", but "..." or "" might be ok.

Re: [9fans] mkstemp()

[boyd, rounin]

other deal is to remove the file, freeing up the name after its creation (if i have my 9P right).
--
MGRS 31U DQ 52572 12604

Re: [9fans] mkstemp()

[Russ Cox]

> other deal is to remove the file, freeing up the name after its creation (if i have my 9P right).

if you remove the file, it's up to the file server
whether you can still access it with the open fd.
the plan 9 disk file servers (kfs, ken fs, fossil) don't let you.

russ

Re: [9fans] mkstemp()

[arisawa]

Hello,

I hope temporal filename is made clear so that I can safely remove
files that are no longer required .
In my /tmp, there are many temporal files that is created by acme
and others.
I hesitate whether I can safely remove these annoying files.

Kenji Arisawa

Re: [9fans] mkstemp()

[Russ Cox]

> I hope temporal filename is made clear so that I can safely remove
> files that are no longer required .
> In my /tmp, there are many temporal files that is created by acme
> and others.
> I hesitate whether I can safely remove these annoying files.

If you log in and don't start any programs, then it's safe
to remove everything in /tmp (unless you have precious data
stored there!).

To find out what is being used in /tmp, grep /tmp /proc/*/fd.

Russ

Re: [9fans] mkstemp()

[arisawa]

Hello Russ

> To find out what is being used in /tmp, grep /tmp /proc/*/fd.
>

Oh thanks. That's very nice!

By the way,  I don't know what to do: assume I find a "listen" is 
listening a port
that I don't intended to listen.
I will fix some files in /rc/bin/service/ and then I will want to kill 
the "listen".
But how to find the pid ?

Kenji Arisawa

Re: [9fans] mkstemp()

[Russ Cox]

> By the way,  I don't know what to do: assume I find a "listen" is
> listening a port
> that I don't intended to listen.
> I will fix some files in /rc/bin/service/ and then I will want to kill
> the "listen".
> But how to find the pid ?

There's not a good way.  You could kill all the listeners
and then restart aux/listen though.

Russ

Re: [9fans] mkstemp()

[geoff]

No, removing the open file is a Unix idiom.

On Plan 9, create your temporaries with the ORCLOSE bit and they'll be
removed when closed, including if the application crashes or the
client machine crashes (but not, alas, if the file server is halted
with the file still open, at least in ken fs).  Or run ramfs first.

Re: [9fans] mkstemp()

[Russ Cox]

Enjoy.

int
mkstemp(char *template)
{
    char *s;
    int i, fd;

    s = strdup(template);
    if(s == nil)
        return -1;
    for(i=0; i<20; i++){
        strcpy(s, template);
        mktemp(s);
        if((fd = create(s, ORDWR|OEXCL, 0666)) >= 0){
            strcpy(template, s);
            free(s);
            return fd;
        }
    }
    free(s);
    return -1;
}

On Fri, 28 Jan 2005 14:39:41 +0100, cej@gli.cas.cz <cej@gli.cas.cz> wrote:
> need mkstemp() [BSD unistd.h]
> any better idea than to copy the code from BSD | GNU ?
> (I understand that returns a fd of a temp file)
> 
> thanks,
> ++pac.
> 
>

Re: [9fans] mkstemp()

[Ronald G. Minnich]

On Fri, 28 Jan 2005 cej@gli.cas.cz wrote:

> need mkstemp() [BSD unistd.h]

why? Your tmp is not my tmp. I see no need for the actual function, just 
make something with the same name that doesn't worry about 'unique' file 
names. 

ron

Re: [9fans] mkstemp()

[Gorka Guardiola Múzquiz]

What about two instances of the same program?. What I meant in my other mail 
is that a big random number would probably suffice, as the probability of 
collision is really low
(as you say we already have user separation).
Even more if it has an identifier per applications (ex. /tmp/acme.23423) 
which also makes the name readable. In any case, if the create fails, you 
can always generate the
random number again.

----- Original Message ----- 
From: "Ronald G. Minnich" <rminnich@lanl.gov>
To: "Fans of the OS Plan 9 from Bell Labs" <9fans@cse.psu.edu>
Sent: Friday, January 28, 2005 4:38 PM
Subject: Re: [9fans] mkstemp()


>
>
> On Fri, 28 Jan 2005 cej@gli.cas.cz wrote:
>
>> need mkstemp() [BSD unistd.h]
>
> why? Your tmp is not my tmp. I see no need for the actual function, just
> make something with the same name that doesn't worry about 'unique' file
> names.
>
> ron
> 

Re: [9fans] mkstemp()

[Russ Cox]

> What about two instances of the same program?. What I meant in my other mail
> is that a big random number would probably suffice, as the probability of
> collision is really low
> (as you say we already have user separation).
> Even more if it has an identifier per applications (ex. /tmp/acme.23423)
> which also makes the name readable. In any case, if the create fails, you
> can always generate the
> random number again.

which is exactly what mkstemp does.