Re: LYNX-DEV Tiny patch to fix HP-UX 10.20 + curses + 2.8.pre2

[T.E.Dickey]

> Sorry if this post is too wordy.
no problem - I'm looking for details.  The fix looks plausible (figuring
out the possible values from config.guess isn't easy except by running the
script on the right sort of box).  Comparing with the 9.05 config.cache,
I think you're right.
 
> This bug looks mostly like a simple typo in aclocal.m4 in a workaround
> for a weird HP bug (i.e. there is a typo not in the workaround but in
> the configure code which can activate the workaround!).
> 
> The workaround for the bug in -lcurses is there but it was not 
> functioning right in the current autoconf script.  Not sure what 
> version broke this.  (The curses bug makes the darn thing dump core 
> (bus error) on startup, and the fix was in the old makefile from way 
> back -- pre-autoconf.)
> 
> BTW, I confirmed with a debugger that this is similar to the same 
> select() bug that the PROBLEMS file documents (if not the same).
> I don't know anything about the patch mentioned there (just a student
> user here, don't run these machines).  My arrow keys work too, though.
> 
> I found this in docs/CHANGES2.8:
> 
> 1997-09-17
> * Add configure test for HP-UX -lHcurses library. - TD
> 
> Did the configure test ever work?  Hmm.
maybe not - I've only gotten vague reports about anything but ncurses.
 
> It's all pretty much black magic to me what all of these different 
> curses libraries are for, but the the following fix (you can make it 
> by hand in the configure script just fine too) Works For Me.
My impression is that HP got stuck supporting a variant of SVr3 curses
which has been obsolete for some time (SVr4's been there a while), and
someone decided to rename the old curses to Hcurses.  (But that's a guess).
 
> Note I was using /bin/cc because it's much faster (the compilation
> time, that is) on these machines.  I see no reason why gcc would not
> work equally well, it certainly crashed equally ``well'' with libcurses.
I was unable to use /bin/cc (K&R) because it did not parse the 'assert()'
statements in GridText.c (but someone reported no problem - perhaps it's
a missing patch to cc, or something like that).
 
> Here's the piddly diff for aclocal.m4.  
> 
> *** aclocal.m4.orig   Wed Mar  4 03:01:17 1998
> --- aclocal.m4        Wed Mar  4 03:11:18 1998
> ***************
> *** 265,271 ****
>   freebsd*) #(vi
>       AC_CHECK_LIB(mytinfo,tgoto,[LIBS="-lmytinfo $LIBS"])
>       ;;
> ! *hp-hpux10.*)
>       AC_CHECK_LIB(Hcurses,initscr,[
>               # HP's header uses __HP_CURSES, but user claims _HP_CURSES.
>               LIBS="-lHcurses $LIBS"
> --- 265,271 ----
>   freebsd*) #(vi
>       AC_CHECK_LIB(mytinfo,tgoto,[LIBS="-lmytinfo $LIBS"])
>       ;;
> ! hpux10.*)
>       AC_CHECK_LIB(Hcurses,initscr,[
>               # HP's header uses __HP_CURSES, but user claims _HP_CURSES.
>               LIBS="-lHcurses $LIBS"
> 
> ...
> 
> Can someone else with an HP-UX 10.20 box try this out?  Here is a
> ``hacked configure script diff'' (the configure file this makes is the
> same that the above aclocal.m4 should cause autoconf to generate).  
> Remember to kill your config.cache before running this (-lHcurses has 
> no wborder() and your cache thinks it does if you have built with a 
> not-so-weirdo curses library).
> 
> *** configure.orig    Wed Mar  4 02:59:51 1998
> --- configure Wed Mar  4 03:11:32 1998
> ***************
> *** 3429,3435 ****
>   fi
>   
>       ;;
> ! *hp-hpux10.*)
>       echo $ac_n "checking for initscr in -lHcurses""... $ac_c" 1>&6
>   echo "configure:3435: checking for initscr in -lHcurses" >&5
>   ac_lib_var=`echo Hcurses'_'initscr | sed 'y%./+-%__p_%'`
> --- 3429,3435 ----
>   fi
>   
>       ;;
> ! hpux10.*)
>       echo $ac_n "checking for initscr in -lHcurses""... $ac_c" 1>&6
>   echo "configure:3435: checking for initscr in -lHcurses" >&5
>   ac_lib_var=`echo Hcurses'_'initscr | sed 'y%./+-%__p_%'`
> 
> ...
> 
> Hope this helps, I'm anticipating the 2.8 release.  If you get it out 
> this week, they might install it here on campus over Spring Break 
> (next week).  I'm not holding my breath (that lynx-dev won't find 
> lots more bugs, or that the software people here will have the time).
I'd like to release it this weekend - so far we've found several minor
bugs (and become more aware of the design differences between 2.7.2 and
2.8pre).  If we go back & redesign before release, however, I don't believe
it'll be released.
 
> Also, FYI, I have a confirmed curses build on HP-UX 9.01.  I'll try
> some more platforms as time becomes available.
> 
> 
> -- 
> Jonathan Sergent / address@hidden
> 


-- 
Thomas E. Dickey
address@hidden
http://www.clark.net/pub/dickey