Re: lynx-dev Just an FYI

[Michael Sokolov]

   Bela Lubkin <address@hidden> wrote:
> If you spend a couple of minutes reading through that output, you should
> be able to figure out the problem.
   
   Figuring out the workings of a 200 KB infernal script is not something a
normal (or even paranormal) human can do in "a couple of minutes". Sorry, I
have better uses for my time.
   
> configure assumes that ultrix cc
> will support a "-Olimit 600" directive (which controls some aspect of
> the optimizer).  Yours doesn't.
   
   Actually I know this problem very well. As it happens, VAX Ultrix and
MIPS Ultrix are completely different beasts in this regard. The C compiler
in VAX Ultrix (as well as in Berkeley UNIX(R) and other faithful True
UNIX(R) systems) is the Holy Original UNIX(R) C Compiler written by K&R
themselves. (Just in case someone still doesn't understand, it's not "a"
K&R C compiler, it's _THE_ K&R C compiler.) MIPS Ultrix, on the other hand,
uses some DEC-made ANSI C crap. The "-Olimit 600" option is obviously
targeted for the latter, since the Holy Original UNIX(R) C Compiler doesn't
have options that even look like this one. The test in Mr. Dickey's
infernal script obviously makes the extremely insulting and offensive
assumption that all Ultrix systems are MIPSen.
   
   Speaking of this, v2.7.2's Makefile was no different. On yesterday's
round I have made one mod that I forgot to document in the CHANGES file:
   
* Renamed the ultrix and ultrix-slang targets into mipsultrix and
  mipsultrix-slang, respectively, since they apply only to MIPS Ultrix.
  VAX Ultrix users should use my new unified BSD target. Also removed the
  decstation and decstation-slang aliases for these targets. Not all DEC MIPS
  machines are called DECstations, nor do they have to run Ultrix. This also
  reduces the Makefile target clutter. - MS
   
   (I will include this in CHANGES on the next round.) As one would expect,
the mipsultrix and mipsultrix-slang targets do specify "-Olimit 600".
   
> Find the one occurrence of "600" in
> configure, and fix it appropriately.  The appropriate fix for you is to
> remove the offending option(*).
   
   I can bet that even with this change Mr. Dickey's version of Lynx will
still (always!) crash miserably on Real UNIX(R). Quoting from my very first
mod:
   
> [...] I have also modified this Makefile not to use the -o option for cc.
> See the comments inside for more info. - MS
   
   If you look in WWW/Library/unix/Makefile in lynx2-7-2ms.zip, you'll find
a description of a MAJOR misguidance in all versions of Lynx before 2-7-2MS
that prevents them from ever compiling on Real UNIX(R).
   
> You won't be successful as a patch integrator if simple problems like
> this stop you...
   
   Wrong: since I am a faithful worshiper of True UNIX(R) and stay away
from autoconf and other heresy, I am spared of such problems.
   
> There's a long history of Lynx failing to compile on particular systems
> because the Makefile stanza for that system, or now the configure
> assumptions, have obsolete compiler options embedded in them.
   
   In this case it is not an "obsolete" compiler option (it doesn't matter
here whether it's Ultrix v4.00 or v1.00), but the difference between Real
Computers (VAXen) and toys (MIPSen).
   
   Sincerely,
   Michael Sokolov
   Phone: 440-449-0299
   ARPA Internet SMTP mail: address@hidden