[Caml-list] Unix.getlogin () fails when stdin is redirected

[Caml-list] Unix.getlogin () fails when stdin is redirected

[Ricardo Catalinas Jiménez]

I found out the next issue in this simple code:

let () =
    print_endline "Hello";
    print_endline (Unix.getlogin ())

Running in the normal case, with `./a.out' gives:

Hello
ricardo

But running like `./a.out </dev/null' makes Unix.getlogin fail:

Hello
Fatal error: exception Unix.Unix_error(20, "getlogin", "")

A simple strace reveals the problem:

   open("/etc/passwd", O_RDONLY|O_CLOEXEC) = 3
   fstat(3, {st_mode=S_IFREG|0644, st_size=509, ...}) = 0
   mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fb125554000
   read(3, "root:x:0:0:root:/root:/bin/bash\n"..., 4096) = 509
   read(3, "", 4096)                       = 0
   close(3)                                = 0
   munmap(0x7fb125554000, 4096)            = 0
-> ioctl(0, SNDCTL_TMR_TIMEBASE or SNDRV_TIMER_IOCTL_NEXT_DEVICE or TCGETS, 0x7fff12682c98) = -1 ENOTTY (Inappropriate ioctl for device)
   write(2, "Fatal error: exception Unix.Unix"..., 59) = 59
   exit_group(2)                           = ?

Any idea how to fix this?


Best regards
-- 
Ricardo (http://r.untroubled.be/)

[Caml-list] Re: Unix.getlogin () fails when stdin is redirected

[Ricardo Catalinas Jiménez]

On Tue, Mar 20, 2012 at 06:51:13PM +0100, Ricardo Catalinas Jiménez wrote:
> I found out the next issue in this simple code:
> 
> let () =
>     print_endline "Hello";
>     print_endline (Unix.getlogin ())
> 
> Running in the normal case, with `./a.out' gives:
> 
> Hello
> ricardo
> 
> But running like `./a.out </dev/null' makes Unix.getlogin fail:
> 
> Hello
> Fatal error: exception Unix.Unix_error(20, "getlogin", "")
> 
> A simple strace reveals the problem:
> 
>    open("/etc/passwd", O_RDONLY|O_CLOEXEC) = 3
>    fstat(3, {st_mode=S_IFREG|0644, st_size=509, ...}) = 0
>    mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fb125554000
>    read(3, "root:x:0:0:root:/root:/bin/bash\n"..., 4096) = 509
>    read(3, "", 4096)                       = 0
>    close(3)                                = 0
>    munmap(0x7fb125554000, 4096)            = 0
> -> ioctl(0, SNDCTL_TMR_TIMEBASE or SNDRV_TIMER_IOCTL_NEXT_DEVICE or TCGETS, 0x7fff12682c98) = -1 ENOTTY (Inappropriate ioctl for device)
>    write(2, "Fatal error: exception Unix.Unix"..., 59) = 59
>    exit_group(2)                           = ?


Someone knew the answer, man 3 getlogin reads:

       Note that glibc does not follow the POSIX specification and uses
       stdin instead of /dev/tty.  A bug.  (Other recent systems, like
       SunOS 5.8 and HP-UX 11.11 and FreeBSD 4.8 all return the login
       name also when stdin is redirected.)


Regards
-- 
Ricardo (http://r.untroubled.be/)

Re: [Caml-list] Re: Unix.getlogin () fails when stdin is redirected

[David House]

Note that Jane Street's core library [1] does not use getlogin(3) in 
its replacement Unix module, for exactly this reason:

(* The standard getlogin function goes through utmp which is unreliable,
   see the BUGS section of getlogin(3) *)
let getlogin_orig = Unix.getlogin
let getlogin () = (Unix.getpwuid (getuid ())).Unix.pw_name

[1]: https://bitbucket.org/yminsky/ocaml-core/wiki/Home

I just tested your specific example, and it worked fine.

On Tue 20 Mar 2012 06:07:59 PM GMT, Ricardo Catalinas Jiménez wrote:
> On Tue, Mar 20, 2012 at 06:51:13PM +0100, Ricardo Catalinas Jiménez wrote:
>> I found out the next issue in this simple code:
>>
>> let () =
>>      print_endline "Hello";
>>      print_endline (Unix.getlogin ())
>>
>> Running in the normal case, with `./a.out' gives:
>>
>> Hello
>> ricardo
>>
>> But running like `./a.out</dev/null' makes Unix.getlogin fail:
>>
>> Hello
>> Fatal error: exception Unix.Unix_error(20, "getlogin", "")
>>
>> A simple strace reveals the problem:
>>
>>     open("/etc/passwd", O_RDONLY|O_CLOEXEC) = 3
>>     fstat(3, {st_mode=S_IFREG|0644, st_size=509, ...}) = 0
>>     mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fb125554000
>>     read(3, "root:x:0:0:root:/root:/bin/bash\n"..., 4096) = 509
>>     read(3, "", 4096)                       = 0
>>     close(3)                                = 0
>>     munmap(0x7fb125554000, 4096)            = 0
>> ->  ioctl(0, SNDCTL_TMR_TIMEBASE or SNDRV_TIMER_IOCTL_NEXT_DEVICE or TCGETS, 0x7fff12682c98) = -1 ENOTTY (Inappropriate ioctl for device)
>>     write(2, "Fatal error: exception Unix.Unix"..., 59) = 59
>>     exit_group(2)                           = ?
>
>
> Someone knew the answer, man 3 getlogin reads:
>
>         Note that glibc does not follow the POSIX specification and uses
>         stdin instead of /dev/tty.  A bug.  (Other recent systems, like
>         SunOS 5.8 and HP-UX 11.11 and FreeBSD 4.8 all return the login
>         name also when stdin is redirected.)
>
>
> Regards

Re: [Caml-list] Re: Unix.getlogin () fails when stdin is redirected

[Till Varoquaux]

getpwuid(getuid()) is not a synonym for get_login (refer to the
discussion in the POSIX standard[^1]). You should not shadow posix
functions by functions with different semantics in the Unix modules;
providing your own abstraction over the OS is a commendable goal but
you should do so without silently bypassing core functions.

Till
[1]:http://pubs.opengroup.org/onlinepubs/007904975/functions/getlogin.html


On Tue, Mar 20, 2012 at 2:28 PM, David House <dhouse@janestreet.com> wrote:
> Note that Jane Street's core library [1] does not use getlogin(3) in its
> replacement Unix module, for exactly this reason:
>
> (* The standard getlogin function goes through utmp which is unreliable,
>  see the BUGS section of getlogin(3) *)
> let getlogin_orig = Unix.getlogin
> let getlogin () = (Unix.getpwuid (getuid ())).Unix.pw_name
>
> [1]: https://bitbucket.org/yminsky/ocaml-core/wiki/Home
>
> I just tested your specific example, and it worked fine.
>
>
> On Tue 20 Mar 2012 06:07:59 PM GMT, Ricardo Catalinas Jiménez wrote:
>>
>> On Tue, Mar 20, 2012 at 06:51:13PM +0100, Ricardo Catalinas Jiménez wrote:
>>>
>>> I found out the next issue in this simple code:
>>>
>>> let () =
>>>     print_endline "Hello";
>>>     print_endline (Unix.getlogin ())
>>>
>>> Running in the normal case, with `./a.out' gives:
>>>
>>> Hello
>>> ricardo
>>>
>>> But running like `./a.out</dev/null' makes Unix.getlogin fail:
>>>
>>> Hello
>>> Fatal error: exception Unix.Unix_error(20, "getlogin", "")
>>>
>>> A simple strace reveals the problem:
>>>
>>>    open("/etc/passwd", O_RDONLY|O_CLOEXEC) = 3
>>>    fstat(3, {st_mode=S_IFREG|0644, st_size=509, ...}) = 0
>>>    mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1,
>>> 0) = 0x7fb125554000
>>>    read(3, "root:x:0:0:root:/root:/bin/bash\n"..., 4096) = 509
>>>    read(3, "", 4096)                       = 0
>>>    close(3)                                = 0
>>>    munmap(0x7fb125554000, 4096)            = 0
>>> ->  ioctl(0, SNDCTL_TMR_TIMEBASE or SNDRV_TIMER_IOCTL_NEXT_DEVICE or
>>> TCGETS, 0x7fff12682c98) = -1 ENOTTY (Inappropriate ioctl for device)
>>>    write(2, "Fatal error: exception Unix.Unix"..., 59) = 59
>>>    exit_group(2)                           = ?
>>
>>
>>
>> Someone knew the answer, man 3 getlogin reads:
>>
>>        Note that glibc does not follow the POSIX specification and uses
>>        stdin instead of /dev/tty.  A bug.  (Other recent systems, like
>>        SunOS 5.8 and HP-UX 11.11 and FreeBSD 4.8 all return the login
>>        name also when stdin is redirected.)
>>
>>
>> Regards
>
>
>
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa-roc.inria.fr/wws/info/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>

Re: [Caml-list] Re: Unix.getlogin () fails when stdin is redirected

[Yaron Minsky]

[-- Attachment #1: Type: text/plain, Size: 3471 bytes --]

Is there a concrete difference in behavior you're concerned about?

y

On Tue, Mar 20, 2012 at 2:48 PM, Till Varoquaux <till@pps.jussieu.fr> wrote:

> getpwuid(getuid()) is not a synonym for get_login (refer to the
> discussion in the POSIX standard[^1]). You should not shadow posix
> functions by functions with different semantics in the Unix modules;
> providing your own abstraction over the OS is a commendable goal but
> you should do so without silently bypassing core functions.
>
> Till
> [1]:http://pubs.opengroup.org/onlinepubs/007904975/functions/getlogin.html
>
>
> On Tue, Mar 20, 2012 at 2:28 PM, David House <dhouse@janestreet.com>
> wrote:
> > Note that Jane Street's core library [1] does not use getlogin(3) in its
> > replacement Unix module, for exactly this reason:
> >
> > (* The standard getlogin function goes through utmp which is unreliable,
> >  see the BUGS section of getlogin(3) *)
> > let getlogin_orig = Unix.getlogin
> > let getlogin () = (Unix.getpwuid (getuid ())).Unix.pw_name
> >
> > [1]: https://bitbucket.org/yminsky/ocaml-core/wiki/Home
> >
> > I just tested your specific example, and it worked fine.
> >
> >
> > On Tue 20 Mar 2012 06:07:59 PM GMT, Ricardo Catalinas Jiménez wrote:
> >>
> >> On Tue, Mar 20, 2012 at 06:51:13PM +0100, Ricardo Catalinas Jiménez
> wrote:
> >>>
> >>> I found out the next issue in this simple code:
> >>>
> >>> let () =
> >>>     print_endline "Hello";
> >>>     print_endline (Unix.getlogin ())
> >>>
> >>> Running in the normal case, with `./a.out' gives:
> >>>
> >>> Hello
> >>> ricardo
> >>>
> >>> But running like `./a.out</dev/null' makes Unix.getlogin fail:
> >>>
> >>> Hello
> >>> Fatal error: exception Unix.Unix_error(20, "getlogin", "")
> >>>
> >>> A simple strace reveals the problem:
> >>>
> >>>    open("/etc/passwd", O_RDONLY|O_CLOEXEC) = 3
> >>>    fstat(3, {st_mode=S_IFREG|0644, st_size=509, ...}) = 0
> >>>    mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS,
> -1,
> >>> 0) = 0x7fb125554000
> >>>    read(3, "root:x:0:0:root:/root:/bin/bash\n"..., 4096) = 509
> >>>    read(3, "", 4096)                       = 0
> >>>    close(3)                                = 0
> >>>    munmap(0x7fb125554000, 4096)            = 0
> >>> ->  ioctl(0, SNDCTL_TMR_TIMEBASE or SNDRV_TIMER_IOCTL_NEXT_DEVICE or
> >>> TCGETS, 0x7fff12682c98) = -1 ENOTTY (Inappropriate ioctl for device)
> >>>    write(2, "Fatal error: exception Unix.Unix"..., 59) = 59
> >>>    exit_group(2)                           = ?
> >>
> >>
> >>
> >> Someone knew the answer, man 3 getlogin reads:
> >>
> >>        Note that glibc does not follow the POSIX specification and uses
> >>        stdin instead of /dev/tty.  A bug.  (Other recent systems, like
> >>        SunOS 5.8 and HP-UX 11.11 and FreeBSD 4.8 all return the login
> >>        name also when stdin is redirected.)
> >>
> >>
> >> Regards
> >
> >
> >
> >
> > --
> > Caml-list mailing list.  Subscription management and archives:
> > https://sympa-roc.inria.fr/wws/info/caml-list
> > Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> > Bug reports: http://caml.inria.fr/bin/caml-bugs
> >
>
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa-roc.inria.fr/wws/info/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>
>

[-- Attachment #2: Type: text/html, Size: 5139 bytes --]

Re: [Caml-list] Re: Unix.getlogin () fails when stdin is redirected

[Till Varoquaux]

As far as I can tell getpwuid(geteuid()) is what the (deprecated)
cuserid function does. The linux man page explains the difference
between the two functions like this:

  _ These functions let your program identify positively the user who
is running (cuserid()) or the user who logged in this session
(getlogin()). (These can differ when set-user-ID programs are
involved.)

I expect there to be more points were the behaviours of those two
functions diverge (e.g.: sudo, deamons etc...).

In general I am very skeptic any time I see a one liner that is sold
as a working alternative to a glibc function. The glibc has issues and
pitfalls but they tend to be very well documented. "Better the devil
you know"...

Till

On Tue, Mar 20, 2012 at 3:41 PM, Yaron Minsky <yminsky@gmail.com> wrote:
> Is there a concrete difference in behavior you're concerned about?
>
> y
>
>
> On Tue, Mar 20, 2012 at 2:48 PM, Till Varoquaux <till@pps.jussieu.fr> wrote:
>>
>> getpwuid(getuid()) is not a synonym for get_login (refer to the
>> discussion in the POSIX standard[^1]). You should not shadow posix
>> functions by functions with different semantics in the Unix modules;
>> providing your own abstraction over the OS is a commendable goal but
>> you should do so without silently bypassing core functions.
>>
>> Till
>> [1]:http://pubs.opengroup.org/onlinepubs/007904975/functions/getlogin.html
>>
>>
>> On Tue, Mar 20, 2012 at 2:28 PM, David House <dhouse@janestreet.com>
>> wrote:
>> > Note that Jane Street's core library [1] does not use getlogin(3) in its
>> > replacement Unix module, for exactly this reason:
>> >
>> > (* The standard getlogin function goes through utmp which is unreliable,
>> >  see the BUGS section of getlogin(3) *)
>> > let getlogin_orig = Unix.getlogin
>> > let getlogin () = (Unix.getpwuid (getuid ())).Unix.pw_name
>> >
>> > [1]: https://bitbucket.org/yminsky/ocaml-core/wiki/Home
>> >
>> > I just tested your specific example, and it worked fine.
>> >
>> >
>> > On Tue 20 Mar 2012 06:07:59 PM GMT, Ricardo Catalinas Jiménez wrote:
>> >>
>> >> On Tue, Mar 20, 2012 at 06:51:13PM +0100, Ricardo Catalinas Jiménez
>> >> wrote:
>> >>>
>> >>> I found out the next issue in this simple code:
>> >>>
>> >>> let () =
>> >>>     print_endline "Hello";
>> >>>     print_endline (Unix.getlogin ())
>> >>>
>> >>> Running in the normal case, with `./a.out' gives:
>> >>>
>> >>> Hello
>> >>> ricardo
>> >>>
>> >>> But running like `./a.out</dev/null' makes Unix.getlogin fail:
>> >>>
>> >>> Hello
>> >>> Fatal error: exception Unix.Unix_error(20, "getlogin", "")
>> >>>
>> >>> A simple strace reveals the problem:
>> >>>
>> >>>    open("/etc/passwd", O_RDONLY|O_CLOEXEC) = 3
>> >>>    fstat(3, {st_mode=S_IFREG|0644, st_size=509, ...}) = 0
>> >>>    mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS,
>> >>> -1,
>> >>> 0) = 0x7fb125554000
>> >>>    read(3, "root:x:0:0:root:/root:/bin/bash\n"..., 4096) = 509
>> >>>    read(3, "", 4096)                       = 0
>> >>>    close(3)                                = 0
>> >>>    munmap(0x7fb125554000, 4096)            = 0
>> >>> ->  ioctl(0, SNDCTL_TMR_TIMEBASE or SNDRV_TIMER_IOCTL_NEXT_DEVICE or
>> >>> TCGETS, 0x7fff12682c98) = -1 ENOTTY (Inappropriate ioctl for device)
>> >>>    write(2, "Fatal error: exception Unix.Unix"..., 59) = 59
>> >>>    exit_group(2)                           = ?
>> >>
>> >>
>> >>
>> >> Someone knew the answer, man 3 getlogin reads:
>> >>
>> >>        Note that glibc does not follow the POSIX specification and uses
>> >>        stdin instead of /dev/tty.  A bug.  (Other recent systems, like
>> >>        SunOS 5.8 and HP-UX 11.11 and FreeBSD 4.8 all return the login
>> >>        name also when stdin is redirected.)
>> >>
>> >>
>> >> Regards
>> >
>> >
>> >
>> >
>> > --
>> > Caml-list mailing list.  Subscription management and archives:
>> > https://sympa-roc.inria.fr/wws/info/caml-list
>> > Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
>> > Bug reports: http://caml.inria.fr/bin/caml-bugs
>> >
>>
>>
>> --
>> Caml-list mailing list.  Subscription management and archives:
>> https://sympa-roc.inria.fr/wws/info/caml-list
>> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
>> Bug reports: http://caml.inria.fr/bin/caml-bugs
>>
>

Re: [Caml-list] Re: Unix.getlogin () fails when stdin is redirected

[oliver]

Are there so many bugs in glibc that fixing this one means waiting
in a long, long queue... and bugfixes will need months or years?

Are there any reasons this bug is not already fixed?
And since when is it known?

Ciao,
  Oliver


On Tue, Mar 20, 2012 at 04:12:36PM -0400, Till Varoquaux wrote:
> As far as I can tell getpwuid(geteuid()) is what the (deprecated)
> cuserid function does. The linux man page explains the difference
> between the two functions like this:
> 
>   _ These functions let your program identify positively the user who
> is running (cuserid()) or the user who logged in this session
> (getlogin()). (These can differ when set-user-ID programs are
> involved.)
> 
> I expect there to be more points were the behaviours of those two
> functions diverge (e.g.: sudo, deamons etc...).
> 
> In general I am very skeptic any time I see a one liner that is sold
> as a working alternative to a glibc function. The glibc has issues and
> pitfalls but they tend to be very well documented. "Better the devil
> you know"...
> 
> Till
> 
> On Tue, Mar 20, 2012 at 3:41 PM, Yaron Minsky <yminsky@gmail.com> wrote:
> > Is there a concrete difference in behavior you're concerned about?
> >
> > y
> >
> >
> > On Tue, Mar 20, 2012 at 2:48 PM, Till Varoquaux <till@pps.jussieu.fr> wrote:
> >>
> >> getpwuid(getuid()) is not a synonym for get_login (refer to the
> >> discussion in the POSIX standard[^1]). You should not shadow posix
> >> functions by functions with different semantics in the Unix modules;
> >> providing your own abstraction over the OS is a commendable goal but
> >> you should do so without silently bypassing core functions.
> >>
> >> Till
> >> [1]:http://pubs.opengroup.org/onlinepubs/007904975/functions/getlogin.html
> >>
> >>
> >> On Tue, Mar 20, 2012 at 2:28 PM, David House <dhouse@janestreet.com>
> >> wrote:
> >> > Note that Jane Street's core library [1] does not use getlogin(3) in its
> >> > replacement Unix module, for exactly this reason:
> >> >
> >> > (* The standard getlogin function goes through utmp which is unreliable,
> >> >  see the BUGS section of getlogin(3) *)
> >> > let getlogin_orig = Unix.getlogin
> >> > let getlogin () = (Unix.getpwuid (getuid ())).Unix.pw_name
> >> >
> >> > [1]: https://bitbucket.org/yminsky/ocaml-core/wiki/Home
> >> >
> >> > I just tested your specific example, and it worked fine.
> >> >
> >> >
> >> > On Tue 20 Mar 2012 06:07:59 PM GMT, Ricardo Catalinas Jiménez wrote:
> >> >>
> >> >> On Tue, Mar 20, 2012 at 06:51:13PM +0100, Ricardo Catalinas Jiménez
> >> >> wrote:
> >> >>>
> >> >>> I found out the next issue in this simple code:
> >> >>>
> >> >>> let () =
> >> >>>     print_endline "Hello";
> >> >>>     print_endline (Unix.getlogin ())
> >> >>>
> >> >>> Running in the normal case, with `./a.out' gives:
> >> >>>
> >> >>> Hello
> >> >>> ricardo
> >> >>>
> >> >>> But running like `./a.out</dev/null' makes Unix.getlogin fail:
> >> >>>
> >> >>> Hello
> >> >>> Fatal error: exception Unix.Unix_error(20, "getlogin", "")
> >> >>>
> >> >>> A simple strace reveals the problem:
> >> >>>
> >> >>>    open("/etc/passwd", O_RDONLY|O_CLOEXEC) = 3
> >> >>>    fstat(3, {st_mode=S_IFREG|0644, st_size=509, ...}) = 0
> >> >>>    mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS,
> >> >>> -1,
> >> >>> 0) = 0x7fb125554000
> >> >>>    read(3, "root:x:0:0:root:/root:/bin/bash\n"..., 4096) = 509
> >> >>>    read(3, "", 4096)                       = 0
> >> >>>    close(3)                                = 0
> >> >>>    munmap(0x7fb125554000, 4096)            = 0
> >> >>> ->  ioctl(0, SNDCTL_TMR_TIMEBASE or SNDRV_TIMER_IOCTL_NEXT_DEVICE or
> >> >>> TCGETS, 0x7fff12682c98) = -1 ENOTTY (Inappropriate ioctl for device)
> >> >>>    write(2, "Fatal error: exception Unix.Unix"..., 59) = 59
> >> >>>    exit_group(2)                           = ?
> >> >>
> >> >>
> >> >>
> >> >> Someone knew the answer, man 3 getlogin reads:
> >> >>
> >> >>        Note that glibc does not follow the POSIX specification and uses
> >> >>        stdin instead of /dev/tty.  A bug.  (Other recent systems, like
> >> >>        SunOS 5.8 and HP-UX 11.11 and FreeBSD 4.8 all return the login
> >> >>        name also when stdin is redirected.)
> >> >>
> >> >>
> >> >> Regards
> >> >
> >> >
> >> >
> >> >
> >> > --
> >> > Caml-list mailing list.  Subscription management and archives:
> >> > https://sympa-roc.inria.fr/wws/info/caml-list
> >> > Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> >> > Bug reports: http://caml.inria.fr/bin/caml-bugs
> >> >
> >>
> >>
> >> --
> >> Caml-list mailing list.  Subscription management and archives:
> >> https://sympa-roc.inria.fr/wws/info/caml-list
> >> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> >> Bug reports: http://caml.inria.fr/bin/caml-bugs
> >>
> >
> 
> 
> -- 
> Caml-list mailing list.  Subscription management and archives:
> https://sympa-roc.inria.fr/wws/info/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
> 

Re: [Caml-list] Unix.getlogin () fails when stdin is redirected

[malc]

On Tue, 20 Mar 2012, Ricardo Catalinas Jim?nez wrote:

> I found out the next issue in this simple code:
> 
> let () =
>     print_endline "Hello";
>     print_endline (Unix.getlogin ())
> 
> Running in the normal case, with `./a.out' gives:

FWIW on my machine (linux) calling getlogin from any PTY always fails
(be it screen or some sort of X terminal)

[..snip..]

-- 
mailto:av1474@comtv.ru

Re: [Caml-list] Unix.getlogin () fails when stdin is redirected

[Gerd Stolpmann]

> On Tue, 20 Mar 2012, Ricardo Catalinas Jim?nez wrote:
>
>> I found out the next issue in this simple code:
>>
>> let () =
>>     print_endline "Hello";
>>     print_endline (Unix.getlogin ())
>>
>> Running in the normal case, with `./a.out' gives:
>
> FWIW on my machine (linux) calling getlogin from any PTY always fails
> (be it screen or some sort of X terminal)

I guess /var/run/utmp is not world-readable - which is a perfectly
reasonable configuration, and means that you disabled getlogin
administratively.

Does "last -f /var/run/utmp" give an output?

Gerd

> [..snip..]
>
> --
> mailto:av1474@comtv.ru
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa-roc.inria.fr/wws/info/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>
>
>


-- 
Gerd Stolpmann, Darmstadt, Germany    gerd@gerd-stolpmann.de
Creator of GODI and camlcity.org.
Contact details:        http://www.camlcity.org/contact.html
Company homepage:       http://www.gerd-stolpmann.de
*** Searching for new projects! Need consulting for system
*** programming in Ocaml? Gerd Stolpmann can help you.

Re: [Caml-list] Re: Unix.getlogin () fails when stdin is redirected

[Török Edwin]

On 03/21/2012 12:45 AM, oliver wrote:
> Are there so many bugs in glibc that fixing this one means waiting
> in a long, long queue... and bugfixes will need months or years?
> 
> Are there any reasons this bug is not already fixed?
> And since when is it known?

Using stdin might be on purpose, this is what a glibc comment says:
  /* Get name of tty connected to fd 0.  Return NULL if not a tty or
     if fd 0 isn't open.  Note that a lot of documentation says that
     getlogin() is based on the controlling terminal---what they
     really mean is "the terminal connected to standard input".  The
     getlogin() implementation of DEC Unix, SunOS, Solaris, HP-UX all
     return NULL if fd 0 has been closed, so this is the compatible
     thing to do.  Note that ttyname(open("/dev/tty")) on those
     systems returns /dev/tty, so that is not a possible solution for
     getlogin().  */

--Edwin

Re: [Caml-list] Unix.getlogin () fails when stdin is redirected

[malc]

On Wed, 21 Mar 2012, Gerd Stolpmann wrote:

> 
> > On Tue, 20 Mar 2012, Ricardo Catalinas Jim?nez wrote:
> >
> >> I found out the next issue in this simple code:
> >>
> >> let () =
> >>     print_endline "Hello";
> >>     print_endline (Unix.getlogin ())
> >>
> >> Running in the normal case, with `./a.out' gives:
> >
> > FWIW on my machine (linux) calling getlogin from any PTY always fails
> > (be it screen or some sort of X terminal)
> 
> I guess /var/run/utmp is not world-readable - which is a perfectly
> reasonable configuration, and means that you disabled getlogin
> administratively.
> 
> Does "last -f /var/run/utmp" give an output?

Yep

~$ last -f /var/run/utmp 
malc     tty1                          Wed Mar 21 21:57   still logged in   
reboot   system boot                   Wed Mar 21 21:57          (00:01)    

utmp begins Wed Mar 21 21:57:23 2012
~$ ls -l /var/run/utmp 
-rw-r--r-- 1 root root 3840 2012-03-21 21:57 /var/run/utmp

[..snip..]

-- 
mailto:av1474@comtv.ru

Re: [Caml-list] Re: Unix.getlogin () fails when stdin is redirected

[Yaron Minsky]

[-- Attachment #1: Type: text/plain, Size: 1759 bytes --]

I'm going to drop caml-list for further discussion of this issue.  People
who are interested should feel free to follow along on ocaml-core:

   https://groups.google.com/forum/?fromgroups#!forum/ocaml-core

y

On Tue, Mar 20, 2012 at 10:10 PM, Yaron Minsky <yminsky@janestreet.com>wrote:

> Looping in ocaml-core list.
>
> On Tue, Mar 20, 2012 at 8:51 PM, Gerd Stolpmann <info@gerd-stolpmann.de>wrote:
>
>>
>> I second this. There is one OS where getlogin does not follow POSIX and is
>> maybe insecure, and the fix cannot be to hide the function for all other
>> OS. IMHO, these differences should be handled on a higher level, and not
>> in the module providing the bindings.
>>
>> Semantically, there is a big difference between getlogin and getuid:
>> getlogin shall also work when the user calls a setuid program which in
>> turn invokes a script. These script commands can then use getlogin to
>> identify the original user (which is defined as the user of the session =
>> the user of the controlling terminal). In contrast, getuid would return
>> the uid to which setuid switched (for the script).
>>
>> So, I'd say, you cannot repair getlogin with getuid. The best fix is
>> probably to just run `/usr/bin/logname </dev/tty` and read the printed
>> name.
>>
>
> I can think of a few solutions here:
>
>    - We can use getuid only on platforms where getlogin is busted
>    - We can name our function something other than "getlogin", to avoid
>    confusion.
>    - We can shell-out, in the way you suggest, to implement getlogin on
>    Linux.  My only worry is that this is also going to be somewhat fragile in
>    its own way.  Does calling out to logname with the suggested redirect
>    always work?
>
> I'm open to other suggestions.
>
> y
>

[-- Attachment #2: Type: text/html, Size: 2585 bytes --]

Re: [Caml-list] Re: Unix.getlogin () fails when stdin is redirected

[Yaron Minsky]

[-- Attachment #1: Type: text/plain, Size: 1420 bytes --]

Looping in ocaml-core list.

On Tue, Mar 20, 2012 at 8:51 PM, Gerd Stolpmann <info@gerd-stolpmann.de>wrote:

>
> I second this. There is one OS where getlogin does not follow POSIX and is
> maybe insecure, and the fix cannot be to hide the function for all other
> OS. IMHO, these differences should be handled on a higher level, and not
> in the module providing the bindings.
>
> Semantically, there is a big difference between getlogin and getuid:
> getlogin shall also work when the user calls a setuid program which in
> turn invokes a script. These script commands can then use getlogin to
> identify the original user (which is defined as the user of the session =
> the user of the controlling terminal). In contrast, getuid would return
> the uid to which setuid switched (for the script).
>
> So, I'd say, you cannot repair getlogin with getuid. The best fix is
> probably to just run `/usr/bin/logname </dev/tty` and read the printed
> name.
>

I can think of a few solutions here:

   - We can use getuid only on platforms where getlogin is busted
   - We can name our function something other than "getlogin", to avoid
   confusion.
   - We can shell-out, in the way you suggest, to implement getlogin on
   Linux.  My only worry is that this is also going to be somewhat fragile in
   its own way.  Does calling out to logname with the suggested redirect
   always work?

I'm open to other suggestions.

y

[-- Attachment #2: Type: text/html, Size: 1806 bytes --]

Re: [Caml-list] Re: Unix.getlogin () fails when stdin is redirected

[Gerd Stolpmann]

I second this. There is one OS where getlogin does not follow POSIX and is
maybe insecure, and the fix cannot be to hide the function for all other
OS. IMHO, these differences should be handled on a higher level, and not
in the module providing the bindings.

Semantically, there is a big difference between getlogin and getuid:
getlogin shall also work when the user calls a setuid program which in
turn invokes a script. These script commands can then use getlogin to
identify the original user (which is defined as the user of the session =
the user of the controlling terminal). In contrast, getuid would return
the uid to which setuid switched (for the script).

So, I'd say, you cannot repair getlogin with getuid. The best fix is
probably to just run `/usr/bin/logname </dev/tty` and read the printed
name.

Gerd

> getpwuid(getuid()) is not a synonym for get_login (refer to the
> discussion in the POSIX standard[^1]). You should not shadow posix
functions by functions with different semantics in the Unix modules;
providing your own abstraction over the OS is a commendable goal but you
should do so without silently bypassing core functions.
>
> Till
> [1]:http://pubs.opengroup.org/onlinepubs/007904975/functions/getlogin.html
>
>
> On Tue, Mar 20, 2012 at 2:28 PM, David House <dhouse@janestreet.com> wrote:
>> Note that Jane Street's core library [1] does not use getlogin(3) in
its replacement Unix module, for exactly this reason:
>>
>> (* The standard getlogin function goes through utmp which is
unreliable,  see the BUGS section of getlogin(3) *)
>> let getlogin_orig = Unix.getlogin
>> let getlogin () = (Unix.getpwuid (getuid ())).Unix.pw_name
>>
>> [1]: https://bitbucket.org/yminsky/ocaml-core/wiki/Home
>>
>> I just tested your specific example, and it worked fine.
>>
>>
>> On Tue 20 Mar 2012 06:07:59 PM GMT, Ricardo Catalinas Jiménez wrote:
>>>
>>> On Tue, Mar 20, 2012 at 06:51:13PM +0100, Ricardo Catalinas Jiménez
wrote:
>>>>
>>>> I found out the next issue in this simple code:
>>>>
>>>> let () =
>>>>     print_endline "Hello";
>>>>     print_endline (Unix.getlogin ())
>>>>
>>>> Running in the normal case, with `./a.out' gives:
>>>>
>>>> Hello
>>>> ricardo
>>>>
>>>> But running like `./a.out</dev/null' makes Unix.getlogin fail:
>>>>
>>>> Hello
>>>> Fatal error: exception Unix.Unix_error(20, "getlogin", "")
>>>>
>>>> A simple strace reveals the problem:
>>>>
>>>>    open("/etc/passwd", O_RDONLY|O_CLOEXEC) = 3
>>>>    fstat(3, {st_mode=S_IFREG|0644, st_size=509, ...}) = 0
>>>>    mmap(NULL, 4096, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_ANONYMOUS, -1,
>>>> 0) = 0x7fb125554000
>>>>    read(3, "root:x:0:0:root:/root:/bin/bash\n"..., 4096) = 509  
 read(3, "", 4096)                       = 0
>>>>    close(3)                                = 0  
 munmap(0x7fb125554000, 4096)            = 0
>>>> ->  ioctl(0, SNDCTL_TMR_TIMEBASE or SNDRV_TIMER_IOCTL_NEXT_DEVICE or
TCGETS, 0x7fff12682c98) = -1 ENOTTY (Inappropriate ioctl for device)
   write(2, "Fatal error: exception Unix.Unix"..., 59) = 59  
 exit_group(2)                           = ?
>>>
>>>
>>>
>>> Someone knew the answer, man 3 getlogin reads:
>>>
>>>        Note that glibc does not follow the POSIX specification and
uses
>>>        stdin instead of /dev/tty.  A bug.  (Other recent
systems, like
>>>        SunOS 5.8 and HP-UX 11.11 and FreeBSD 4.8 all return the login
>>>        name also when stdin is redirected.)
>>>
>>>
>>> Regards
>>
>>
>>
>>
>> --
>> Caml-list mailing list.  Subscription management and archives:
https://sympa-roc.inria.fr/wws/info/caml-list
>> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
>> Bug reports: http://caml.inria.fr/bin/caml-bugs
>>
>
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa-roc.inria.fr/wws/info/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>
>
>


-- 
Gerd Stolpmann, Darmstadt, Germany    gerd@gerd-stolpmann.de
Creator of GODI and camlcity.org.
Contact details:        http://www.camlcity.org/contact.html
Company homepage:       http://www.gerd-stolpmann.de
*** Searching for new projects! Need consulting for system
*** programming in Ocaml? Gerd Stolpmann can help you.



-- 
Gerd Stolpmann, Darmstadt, Germany    gerd@gerd-stolpmann.de
Creator of GODI and camlcity.org.
Contact details:        http://www.camlcity.org/contact.html
Company homepage:       http://www.gerd-stolpmann.de
*** Searching for new projects! Need consulting for system
*** programming in Ocaml? Gerd Stolpmann can help you.