Re: lynx

[Jim Pick <jim@jimpick.com>]

Scott Bambrough <scottb@corelcomputer.com> writes:

> Jim Pick wrote:
> > 
> > bdale@gag.com (Bdale Garbee) writes:
> > 
> > > I have 3.3.2.3a-11.arm.1 built, with fbdev and svga servers.  I'm tweaking
> > > the package building control scripts, and haven't tried any of it yet.
> > >
> > > Along the way, I realized 'lynx' failed to build in my dbuild run
> > > the other day, and the package build script for the xterm package
> > > wants it to process a .html of the faq to make a .txt copy.  I'm
> > > going out of town until Tuesday in a few hours, If anyone has time
> > > to work on lynx while I'm gone, that would be nice.
> > 
> > I'll take a look at it.  I tried building it, and it fails like this:
> > 
> > gcc -DHAVE_CONFIG_H -DNSL_FORK  -I../../.. -I../../../src -I../../.. -I../../../src -I../../../WWW/Library/Implementation -O2 -DLINUX    -I../../../WWW/Library/Implementation/ -DXMOSAIC_HACK -DACCESS_AUTH  -c  ../../../WWW/Library/Implementation/HTTP.c
> > /tmp/cc7yumcg.s: Assembler messages:
> > /tmp/cc7yumcg.s:2388: Error: Can not represent OFFSET_IMM relocation in this object file format (1)
> > make[2]: *** [../../Library/unix/HTTP.o] Error 1
> > make[2]: Leaving directory `/home/jim/src/lynx/lynx-2.8.1/WWW/Library/unix'
> > make[1]: *** [all] Error 2
> > 
> > My guess is that it is an egcs bug.  The offending assembly looks like:
> > 
> >   2385  .L859:
> >   2386          cmp     r7, #0
> >   2387          addne   r6, r7, #1
> >   2388          ldreq   r6, .LC138
> >   2389          beq     .L862
> >   2390          ldr     ip, [fp, #-2320]
> >   2391          rsb     r3, r9, r6
> >   2392          rsb     r3, r3, ip
> >   2393          str     r3, [fp, #-2272]
> >   2394          b       .L863
> > 
> > But there is no .LC138 symbol defined anywhere else.
> > 
> > Lynx did build successfully back when we were using Corel's gcc.
> > 
> > Cheers,
> > 
> >  - Jim
> 
> This is a known bug in EGCS.  Turn optimization off on the file.

Ugh.  That's the wrong way to fix it.

The proper thing to do is to fix the bug in egcs.  We shouldn't be
putting kludgy workarounds into the main Debian source tree (unless
it's really important).

I'll take a look at the egcs bug when I get some more time.  Is there
anything else I should know about it?

Cheers,

 - Jim