From: vischne AT ibm DOT net
Subject: Re: ncurses 4.1
3 Mar 1998 10:28:05 -0800
Message-ID: <199803030221.CAA150008.cygnus.gnu-win32@out5.ibm.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
To: gnu-win32 AT cygnus DOT com

>   Geoffrey Noer (noer AT cygnus DOT com)
>   Sun, 1 Mar 1998 14:59:09 -0800 (PST)
>Morph wrote:
>[...]
>> I then ran 'sh make', and this appeared to be going through OK (although it
>> takes an awfully long time to run through the script), and on the 3rd or 4th
>> ranlib I got the following output, at which point it stopped.
>>
>> a - ../obj_g/m_req_name.o
>> a - ../obj_g/m_spacing.o
>> a - ../obj_g/m_userptr.o
>> a - ../obj_g/m_win.o
>> a - ../obj_g/m_adabind.o
>> ranlib ../lib/libmenu_g.a
>> make: *** wait: Invalid argument.  Stop.
>
>This is only a problem under Win 95.  I do not yet understand the
>cause of it.  With B19, I was actually able to rebuild the comp tools
>sources under Win 95 (for the first time), but the make would stop as
>shown above at a seemingly random point.  Restarting the build, it
>would make it farther and then die in the same way.  Doing this three
>or four times, the make finally completed successfully.
>
>So the workaround is, run make repeatedly until the compile succeeds.
>Or, better yet, run Windows NT.  :-)
>
>If anyone finds a cause or solution, please let me know!
>
This is interesting.  gcc2.6 used to have this problem under Linux.  You
had make crashing for no obvious or repeatable reason, yet make succeeded
if you re-started after the crash.  I believe the problem at that time
was that gcc was running out of memory.

BTW, it's obvious that someone did some work on improving the cygwin
terminal, since people are reporting success with ncurses on this release,
and I can vouch for improvements in display for Midnight Commander using
its built-in Unix slang library.  There is one problem remaining:  Keyboard
repeat rate.  If you hold a key down, like left arrow, you get a sequence
of [D[D on the screen.  I don't know why cygwin terminal insists on
echoing the keyboard to the screen using ansi sequences, but you might
do well to check out the keyboard-handling code in the latest slang1.03
release, especially in the slw32tty.c file.  There is a `waitforobject'
call in slw32tty.c that is my candidate for inclusion in the great cygwin
callback function.
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request AT cygnus DOT com" with one line of text: "help".