Re: Plans for fixing libio on i386-linux-gnulibc1?

Re: Plans for fixing libio on i386-linux-gnulibc1?

[Hugh W. Holbrook]

I solved these problems for egcs-970917 on my linux/i386 system by
applying HJ Lu's libio patches and by installing libc-5.4.39 before
building egcs.  The patched version of egcs-970917 built a libstdc++
that seems to work fine and allows you to use the cstdio stuff, as well.

HJ Lu's patch to 970917 is available at 
ftp://ftp.yggdrasil.com/private/hjl/egcs-970917.io.diff.gz 

-Hugh

>Joe Buck  writes:
>> It appears that this can be done just by editing libio/config/linux.mt
>> (or linuxlibc1.mt) and following the directions under
>> 
>> # Comment this out to avoid including the stdio functions in libiostream.a:
>> 
>> It appears that it would be experimental, but hey, we're egcs.
>
>I of course tried that before posting my query ... looks like the new libio 
>code includes a few functions older libio didn't have (libc v5.3.12). I now 
>get something like the following when linking trival program containing
>say fputs():
>
>  libstdc++.so: undefined reference to `_IO_outfloat'
>  libstdc++.so: undefined reference to `_IO_getc'
>  libstdc++.so: undefined reference to `_IO_peekc_locked'
>  libstdc++.so: undefined reference to `_IO_putc'
>
>
>Mumit

Re: Plans for fixing libio on i386-linux-gnulibc1?

[Mumit Khan]

"Hugh W. Holbrook" <holbrook@dsg.stanford.edu> writes:
> 
> I solved these problems for egcs-970917 on my linux/i386 system by
> applying HJ Lu's libio patches and by installing libc-5.4.39 before
> building egcs.  The patched version of egcs-970917 built a libstdc++
> that seems to work fine and allows you to use the cstdio stuff, as well.
> 
> HJ Lu's patch to 970917 is available at 
> ftp://ftp.yggdrasil.com/private/hjl/egcs-970917.io.diff.gz 
> 
> -Hugh
> 

Thanks for the info. I'll try it out against 970924 and see what happens.
I guess I'd like to see these HJL fixes incorporated into egcs, and have 
these work with "standard" distributions like RedHat 4.1/4.2 etc (which 
run libc 5.3.12) without having to upgrade libc if at all possible.

Considering the strides made in egcs compared to gcc2, I'm quite confident
we'll have a viable alternative to other commercial C++ tools I have to
resort to now (mostly EDG based -- eg Kai on Unix). Thanks to all the 
great work by Merrill/Stump/Mitchell et al, it's a reachable goal now.

Regards,
Mumit -- khan@xraylith.wisc.edu
http://www.xraylith.wisc.edu/~khan/

Re: Plans for fixing libio on i386-linux-gnulibc1?

[Jeffrey A Law]

  In message < 199709280732.CAA12094@hp2.xraylith.wisc.edu >you write:
  > I guess I'd like to see these HJL fixes incorporated into egcs, and have 
  > these work with "standard" distributions like RedHat 4.1/4.2 etc (which 
  > run libc 5.3.12) without having to upgrade libc if at all possible.
Well, I just checked in a libio patch from Ulrich, which was based on
some code from HJ.  Hopefully we'll have these problems resolved soon.

Jeff

Plans for fixing libio on i386-linux-gnulibc1?

[Mumit Khan]

What's the current forecast for a working libio under i386-linux? It
seems to work ok for C, but useless for C++ + cstdio (bugs have been 
reported already, so I wouldn't repeat those here).

I notice that libio has the new Ulrich stuff, so is it just a matter of
manually "fixing" _G_CONFIG_H to make things work?

If it's not obvious already, I'm a clueless libio newbie ;-)

Regards,
Mumit -- khan@xraylith.wisc.edu
http://www.xraylith.wisc.edu/~khan/

Re: Plans for fixing libio on i386-linux-gnulibc1?

[Joe Buck]

> What's the current forecast for a working libio under i386-linux? It
> seems to work ok for C, but useless for C++ + cstdio (bugs have been 
> reported already, so I wouldn't repeat those here).

Shouldn't it be possible to set up libstdc++ not to include the iostreams
stuff, like it is on other Unices?  That way it would be easier for users
to try out experimental C++ versions without disrupting their libc.

It appears that this can be done just by editing libio/config/linux.mt
(or linuxlibc1.mt) and following the directions under

# Comment this out to avoid including the stdio functions in libiostream.a:

It appears that it would be experimental, but hey, we're egcs.

My dream is to make it as easy to try out a new g++ / (libg++ , libstdc++)
release on Linux as it is on Solaris: just build and go, no need to
replace your key system libraries.

Re: Plans for fixing libio on i386-linux-gnulibc1?

[Mumit Khan]

Joe Buck <jbuck@synopsys.com> writes:
> It appears that this can be done just by editing libio/config/linux.mt
> (or linuxlibc1.mt) and following the directions under
> 
> # Comment this out to avoid including the stdio functions in libiostream.a:
> 
> It appears that it would be experimental, but hey, we're egcs.

I of course tried that before posting my query ... looks like the new libio 
code includes a few functions older libio didn't have (libc v5.3.12). I now 
get something like the following when linking trival program containing
say fputs():

  libstdc++.so: undefined reference to `_IO_outfloat'
  libstdc++.so: undefined reference to `_IO_getc'
  libstdc++.so: undefined reference to `_IO_peekc_locked'
  libstdc++.so: undefined reference to `_IO_putc'


Mumit

Re: Plans for fixing libio on i386-linux-gnulibc1?

[Jeffrey A Law]

  In message < 9709270001.AA00675@modi.xraylith.wisc.edu >you write:
  > What's the current forecast for a working libio under i386-linux? It
  > seems to work ok for C, but useless for C++ + cstdio (bugs have been 
  > reported already, so I wouldn't repeat those here).
  > 
  > I notice that libio has the new Ulrich stuff, so is it just a matter of
  > manually "fixing" _G_CONFIG_H to make things work?
  > 
  > If it's not obvious already, I'm a clueless libio newbie ;-)
Hopefully some of these problems will be fixed with Ulrich's latest
patch (will be in next snapshot).  But not being a linux user/developer
I don't really know what systems his patch will fix problems for.

jeff