@@GLIBC_2.1

gk4 at us.ibm.com gk4 at us.ibm.com
Thu Nov 18 08:13:01 PST 1999



I have done a little research and have asked around regarding my original
inquiry...

GLIBC 2.1 introduced symbol versioning for backwards compatibility.   So,
in /lib/libc.so.6 the "fopen" API has "fopen at GLIBC_2.0" and
"fopen@@GLIBC_2.1" entries.  The entry with the single "@" is the default
"fopen" that is resolved to.  We can determine which APIs have symbolic
versioning by examining the shared library:

/usr/bin/nm  /lib/libc.so.6

We can determine which symbolic version of an API is used by an application
by examining the binary application.  The binary utilities "objdump" and
"nm -D" will examine stripped binaries and give us the information that we
need to know.

/usr/bin/objdump --dynamic-syms /opt/bin/foo

I have written a simple Perl script that will take one or more binary
programs as arguments then produce the following output:

application0:crc0:rpm0:version0:release0:api0:symversion0:library0:crc0:rpm0:version0:release0
application1:crc1:rpm1:version1:release1:api1:symversion1:library1:crc1:rpm1:version1:release1
application2:crc2:rpm2:version2:release2:api2:symversion2:library2:crc2:rpm2:version2:release2
application3:crc3:rpm3:version3:release3:api3:symversion3:library3:crc3:rpm3:version3:release3

Cheers,

George Kraft IV
gk4 at us.ibm.com
512-838-2688; t/l 678-2688
Linux Technology Center
IBM, Austin Texas


Stuart Anderson <anderson at metrolink.com> on 11/08/99 04:09:05 PM

To:   George Kraft/Austin/IBM at IBMUS
cc:   lsb-discuss at lists.linuxbase.org
Subject:  Re: @@GLIBC_2.1






Appearantly, the fdopen implementation (and several other things) changed
between glibc 2.0 and 2.1. The @@GLIBC_X.X indicates which behavior you
get.

I think there may be some global variable that lives in the application
that
says which version it was linked against, but I'm not sure of the details
(but
need to learn & document it).

The mangle section of the specification is for this type of thing. The
source
signature

FILE * fdopen(int fildes, const char *mode)

is mangled to

FOO * fdopen@@GLIBC_2.0(..????..)

Again, I haven't figured out all of the details, but need to before our
next meeting.

On Mon, 8 Nov 1999 gk4 at us.ibm.com wrote:

>
>
> While trying to inventory the APIs and libraries used by applications
such
> as StarOffice51 and WordPerfect8 I have stumbled across some redundant
APIs
> with @GLIBC_2.0 and @@GLIBC_2.1 suffixes on RedHat 6.0.   So, for
example,
> if I look at Netscape Communicator 4.7 I see that "fdopen" is undefined
and
> needs to be resolved.
>
> [gk4 at act5 gk4]$ nm -D -g /usr/local/netscape/netscape | grep fdopen
>          U fdopen
>
> Looking at what libraries are dynamically linked, I see libc.so.6 is
linked
> to netscape to resolve "fdopen".
>
> [gk4 at act5 gk4]$ ldd /usr/local/netscape/netscape
>         libBrokenLocale.so.1 => /lib/libBrokenLocale.so.1 (0x40018000)
>         libXt.so.6 => /usr/X11R6/lib/libXt.so.6 (0x4001a000)
>         libSM.so.6 => /usr/X11R6/lib/libSM.so.6 (0x40064000)
>         libICE.so.6 => /usr/X11R6/lib/libICE.so.6 (0x4006e000)
>         libXmu.so.6 => /usr/X11R6/lib/libXmu.so.6 (0x40083000)
>         libXpm.so.4 => /usr/X11R6/lib/libXpm.so.4 (0x40095000)
>         libXext.so.6 => /usr/X11R6/lib/libXext.so.6 (0x400a2000)
>         libX11.so.6 => /usr/X11R6/lib/libX11.so.6 (0x400ad000)
>         libdl.so.2 => /lib/libdl.so.2 (0x4014a000)
>         libstdc++-libc6.1-1.so.2 => /usr/lib/libstdc++-libc6.1-1.so.2
> (0x4014d000)
>         libm.so.6 => /lib/libm.so.6 (0x40190000)
>         libc.so.6 => /lib/libc.so.6 (0x401ac000)
>         /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)
>
> However, when I look in libc.so.6 I see two "fdopen" definitions with
> either a @GLIBC_2.0 or @@GLIBC_2.1 suffix!
>
> [gk4 at act5 gk4]$ nm -g /lib/libc.so.6 | grep fdopen
> 00050ae0 T _IO_fdopen@@GLIBC_2.1
> 00053410 T _IO_fdopen at GLIBC_2.0
> 00050ae0 T fdopen@@GLIBC_2.1
> 00053410 T fdopen at GLIBC_2.0
>
> How is "fdopen" being resolved?  What is the purpose of the @GLIBC_2.0 or
> @@GLIBC_2.1 suffixes?
>
> George Kraft IV
> gk4 at us.ibm.com
> 512-838-2688; t/l 678-2688
> Linux Technology Center
> IBM, Austin Texas
>
> --
> To UNSUBSCRIBE, email to lsb-discuss-request at lists.linuxbase.org
> with subject of "unsubscribe". Trouble? Email
listmaster at lists.linuxbase.org
>

                                Stuart

Stuart R. Anderson                               anderson at metrolink.com

Metro Link Incorporated                          South Carolina Office
4711 North Powerline Road                        129 Secret Cove Drive
Fort Lauderdale, Florida 33309                   Lexington, SC 29072
voice: 954.938.0283                              voice: 803.951.3630
fax:   954.938.1982                              SkyTel: 800.405.3401
http://www.metrolink.com/


--
To UNSUBSCRIBE, email to lsb-discuss-request at lists.linuxbase.org
with subject of "unsubscribe". Trouble? Email
listmaster at lists.linuxbase.org






More information about the lsb-discuss mailing list