Hi
Thanks a lof for infos and so on.
I used the static libm.a.
It dosn't work.
But you are truth, I extract all pow-Object and there is no link - really !
So I seem that the problem ist really in the libm.a itself and it better to
use
the rt_math-lib.
But I am a luck men, because I could prevent pow in my Code, alternative I
used left-shift and so on.
So is the problem solved !!!!!!!
And I lerned very much about libs, rtl and so on - thank
Michael
> -----Original Message-----
> From: Dresner, Norman A. [SMTP:[EMAIL PROTECTED]]
> Sent: Tuesday, November 27, 2001 4:58 PM
> To: Back Michael (extern)
> Subject: RE: [rtl] fputs, stderr
>
> Michael.
>
> Your use of the alpha and my use of the Pentium is indeed causing
> the confusion -- and probably your problem too. I just looked at the
> contents of libm.a and found that there are 9 object modules with names
> like
> s_pow.o, ... s_cpowl.o. presumably for different data types. I
> disassembled
> one of them at random (s_pow.o) and there are absolutely no
> call-instructions in it at all.
>
> So... the first question I have for you is whether you're linking
> with the shared library (libm.so) or the static archive (libm.a). If it's
> the latter then AFAIK the linker extracts from the archive only those
> routines needed by your module/program and includes them in the output.
> When you link with the shared version (libm.so), you essentially get all
> of
> the library.
> If you're linking with the shared library (libm.so), try linking
> with the static (libm.a) and see what remains unresolved. That could
> solve
> your problem.
>
> If you're linking with the static library, it's still possible that
> the "pow" routine in the alpha library uses fputs, etc, but the best way
> to
> "prove" that is to extract it from the library
> ar -x /usr/lib/libm.a {module name}
> (unfortunately you need to know the exact module name but you should be
> able
> to get that from the "ld" linker if you don't already know it or you can
> list the contents of the library and search for appropriate modules:
> ar -t /usr/lib/libm.a | grep -i pow
> and select one or all for "extraction" which is a copy-out operation, not
> a
> removal).
> Then you can disassemble the pow-function(s) with
> objdump -S {pow.o-modules}
> and see what they're really calling.
>
> If you're really interested, you can see what each module in the static
> library is calling with the command
> objdump --reloc /usr/lib/libm.a
>
>
> If you want to have any further communications with me after 12 Noon EST
> (it's almost 11 AM now), please direct the e-mail to my alternate address:
>
> sendto:[EMAIL PROTECTED]
> where I work in the afternoon.
>
>
> Norm
>
> Norman Dresner
> Fellow Avionics Support Systems Engineer & (SGI) Advanced Signal
> Processing
> Laboratory Administrator
> Radar Systems Engineering Department
> Northrop Grumman Corporation
> Electronic Systems
> Baltimore-Washington International Airport
> 7323 Aviation Boulevard
> Baltimore Maryland 21240
>
> Voice: (410) 993 - 2096 Mornings; all-day voice-mail
> (410) 969 - 8068 Afternoons with answering machine
> FAX: (410) 993 - 8084 On-site
> (410) 969 - 8068 Afternoons; call first to arrange
> E-Mail: Mornings: mailto:[EMAIL PROTECTED]
> Afternoons: mailto:[EMAIL PROTECTED]
> > -----Original Message-----
> > From: Back Michael (extern) [SMTP:[EMAIL PROTECTED]]
> > Sent: Tuesday, November 27, 2001 10:19 AM
> > To: Dresner, Norman A.
> > Cc: '[EMAIL PROTECTED]'
> > Subject: RE: [rtl] fputs, stderr
> >
> > Hi,
> > I now that and I don't use "fputs" and "stderr" in my code.
> > But I use "pow" and so I link like you adviced the Math-lib.
> > Additional it is show - like you adviced - in the fp-examples.
> > So I try now to like with math-lib and without:
> > With:
> > example.o: unresolved symbol fputs
> > example.o: unresolved symbol stderr
> > example.o: unresolved symbol __errno_location
> > Without:
> > example.o: unresolved symbol pow
> > So I am sure that the math-Lib is the problem.
> > If I look with nm, the reault is that fputs - for example - is from
> > /usr/src/bs/BUILD/glibc-2.2.3/math/../sysdeps/ieee754/k_standard.c
> > If I grep in lib-path so I find this path only in some math-libs.
> > So I think there is a problem with the math-lib.
> > By the way I use RedHeat 7.1 on a Alpha, mybe this is wrong ;-)
> > So I try to compile the math-lib by may self, mybe this works !
> > Thanks
> > Michael
> >
> >
> >
> >
> > > Absolutely. The functions "fputs" and "stderr" are from the standard
> > > library. And you simply cannot use those functions in the kernel.
> > Apart
> > > from the RT-FIFO and anything you write, AFAIK the only "character"
> I/O
> > > you
> > > can use is rtl_printf (or printk in normal portions of the modules).
> > >
> > > Norm
> > >
> > > > -----Original Message-----
> > > > From: Back Michael (extern) [SMTP:[EMAIL PROTECTED]]
> > > > Sent: Tuesday, November 27, 2001 9:59 AM
> > > > To: Dresner, Norman A.; '[EMAIL PROTECTED]'
> > > > Subject: RE: [rtl] fputs, stderr
> > > >
> > > > Thanks for this advice, but I use -lm already ! And it doesn't work.
> > > > I am not sure that the math-lib is the reason for
> > > >
> > > > example.o: unresolved symbol fputs
> > > > example.o: unresolved symbol stderr
> > > > example.o: unresolved symbol __errno_location
> > > >
> > > > but it seems so, if I look with the nm-command.
> > > >
> > > > Michael
> > > >
> > > >
> > > >
> > > > It IS possible to use the math library in modules; I've
> been doing
> > > > it for
> > > > several years. But -- except for certain selected functions
> -- you
> > > > can not
> > > > use the standard library functions.
> > > >
> > > > To link your module with the math library, just add "-lm" to
> the
> > > > partial
> > > > link command. Even if you don't want to run with it, it's
> worth
> > > > doing to
> > > > see if that's where your problems lie.
> > > >
> > > > Norm
> > > >
> > > > > -----Original Message-----
> > > > > From: Back Michael (extern)
> [SMTP:[EMAIL PROTECTED]]
> > > > > Sent: Tuesday, November 27, 2001 2:08 AM
> > > > > To: '[EMAIL PROTECTED]'; rtlinux mailing list
> > > > > Cc: comedi mailing list
> > > > > Subject: RE: [rtl] fputs, stderr
> > > > >
> > > > > Hi,
> > > > > I have the same problem and I am maybe on the way to find
> out what
> > > > is
> > > > > wrong.
> > > > > I think these unresolved symbol are from the math-lib
> > > > > /usr/lib/libm.... and so on.
> > > > > Do you use this lib too ?
> > > > > Michael
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: [EMAIL PROTECTED]
> > > > [SMTP:[EMAIL PROTECTED]]
> > > > > > Sent: Tuesday, November 27, 2001 12:02 AM
> > > > > > To: rtlinux mailing list
> > > > > > Cc: comedi mailing list
> > > > > > Subject: [rtl] fputs, stderr
> > > > > >
> > > > > >
> > > > > > I've upgraded some old code to run on 2.4.1x86-rtl-3.0.
> > > > > > It uses comedi which I've hobbled together to run on
> this rtl
> > > > patch,
> > > > > > comedi 0.7.59 and comedilib 0.7.15.
> > > > > >
> > > > > > The code compiles fine (1 assembler warning that I don't
> > > > understand,
> > > > > > seems insignificant) but when I try inserting the
> resulting
> > > > realtime
> > > > > > module:
> > > > > >
> > > > > > example.o: unresolved symbol fputs
> > > > > > example.o: unresolved symbol stderr
> > > > > >
> > > > > > and subsequently the module doesn't insert. Of course
> I'm not
> > > > linking
> > > > > > stdio, and I'm not trying to use any of the calls
> within. I even
> > > > > commented
> > > > > > out all printk's and rtl_printf's thinking maybe I was
> using the
> > > > wrong
> > > > > > declaration in some bizarre header file, but the same
> message
> > > > appears.
> > > > > >
> > > > > > Any advice before I strip the code to nothing to see
> which
> > > > function
> > > > > > call (I suspect the comedilib functions) is to blame?
> I'd
> > > > greatly
> > > > > > appreciate it...
> > > > > >
> > > > > > -Chuck
-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
--
For more information on Real-Time Linux see:
http://www.rtlinux.org/
