On 11 Mar 1999 08:41:17 +0900, [EMAIL PROTECTED] wrote:
>Jean-Marc Lasgouttes <[EMAIL PROTECTED]> wrote:
>
>> Concerning the problems particular to the OS/2 port, I prefer to
>> accept the changes after a (virtual) nod from SMiyata, since I have no
>> way to check whether things work as advertized.
I will clean up most of the things, though. At least this is my
intention now.... ;) In the meantime just some remarks/proposals, in
case someone has time to make the changes:
>The fix to lib/configure.cmd is a correct one.
Proposal: rename files and references in the sources to
lib/LyX2LocConf.cmd for OS/2 and /lib/LyXLocConf for UNIX in future
releases?
I got confused, as there are four configure scripts, and I accidentally
just run the wrong one.
>I'm not sure about the
>one to (toplevel) configure.cmd
Rename to OS2conf.cmd?
Changes (Already cleaned some small bugs.):
1) --pentium-optimize option
....
if opt='yes' then
do
os2config.16='@CXXFLAGS@ -m486 -O2 -malign-loops=2 -malign-jumps=2
-malign-functions=2 -fno-strength-reduce -Zmtd -Zsysv-signals
-D__ST_MT_ERRNO__ -Wall -Wno-unused -ffast-math'
os2config.17='@CFLAGS@ 'subword(os2config.16,2)
end
else
...
Maybe
-m486 -O2 -malign-loops=2 -malign-jumps=2 -malign-functions=2
is of interest for other gcc 2.7.x users as well. Seems to be a bit
faster on pentium and 30% smaller.
2) explicit (dynamic) linking, changed order a bit, xpm 3.4k seems to
like that more
....
os2config.21='@LYX_LIBS@ -lforms -lXpm'
os2config.22='@LIBS@ -lSM -lICE -lX11 -lsocket'
....
>There are so few who build LyX on
>OS/2 and these people can figure out what s/he likes, I suppose.
Nevertheless it builds very easy. And even install works, provided you
built it with configure.cmd and make all, then edit config/makefile a
bit and make the Makefiles. Only small and easy changes in Makefile
required afterwards. May seem a bit complicated, but as I was not so
sure where which stuff should be installed as what, it was of real
value.
>I personally do not mind running LaTeX while configuring. I always
>build LyX background while I am doing some other works.
If you want to release, you will have to change things by hand
afterwards.
>The documents in overall seem better than mine (especially those
>detailed indications for the ghostscript usages), although I am
>not sure if they belong to the LyX package.
As it seems, those indications are not easily found elsewhere and I
personally heard of people giving up on LyX/2 because of that (OS/2
users often get confused by all this UNIX-like stuff; suspect the
number of frustrated users is rather high, here). So, if there is a
better alternative, just let me know... My favourite here would be a
kind of official distribution with official documentation (already
heard of CD-ROM's sold with comprehensive gimp/2 distributions).
>there are a few inaccuracy
Yes, there are, already found some. I'm always grateful, when others
clean up my mistakes... ;)
> For example,
> - xpm.dll is not required as long as you are using my build,
Unfortunately I had no luck downloading your port lyx-1_0_1-os2-bin.zip
from
ftp or http://tsikora.tiac.net/pub/os2/unix/xfree86/ports/
due to my slow/unstable internet connection.
Maybe more luck for me if provided on faster server, like lyx.org....?
Will you forward it to Hobbes? My uni has a local mirror, the uni's
server being the only ftp server really working for me.
Nevertheless. the explicit dynamic linking (-lxpm) to 4.11 (3.4k) did
work well.
> - LaTeX is required, not an option unless you manually edit
> textclass.lst
Do you mean, what you put in lib/configure.cmd is not enough?
"...
call SysFileTree 'textclass.lst', 'file', 'FO'
if file.0=0 then
do
rc=lineout('textclass.lst','# This file declares layouts and their
associated definition files')
rc=lineout('textclass.lst','# (include dir. relative to the place
where this file is).')
rc=lineout('textclass.lst','# It contains only default values, since
chkconfig.ltx could not be run')
rc=lineout('textclass.lst','# for some reason. Run ./configure if
you need to update it after a')
rc=lineout('textclass.lst','# configuration change.')
rc=lineout('textclass.lst','article article article')
rc=lineout('textclass.lst','report report report')
rc=lineout('textclass.lst','book book book')
rc=lineout('textclass.lst','linuxdoc linuxdoc linuxdoc')
rc=lineout('textclass.lst','letter letter letter')
rc=lineout('textclass.lst')
rc=lineout('chkconfig.sed','s/@.*@/???/g')
rc=lineout('chkconfig.sed')
rc_graphics='none'
end
..."
What should be added?
>Please revise them carefully, I think they are really valuable to
>XFree86/OS2 users.
One question here:
Should we generally advise users (using sh.exe of pdksh.zip, as others
are a bit buggy) to
set EMXSHELL=DRIVE:\bin\sh.exe
With your previous build of version 1.0.0 (not having tested the
current one) I occasionally had the problem that otherwise cmd.exe was
not recognized for some obscure reason, despite setting comspec and
emxshell to d:\os2\cmd.exe. And I have still no idea what triggered
this behaviour...
Another minor bug:
Since lyx 0.12 the painting of inline images with gs does not work for
me; the behaviour being somewhat obscure (or should I say interesting).
1) Your build of 1.0.0 had the forked (did you use fork() here, I have
no idea) child processes of LyX just hanging in an endless loop, while
gs.exe (v 4.03 for x11) received signal hangup.
2) Now, with my build of 1.0.1 the childs of lyx.exe (only the lyx/2
child processes; parent remaining completely stable and intact) just
explode to a sigsegv.
The parent lyxus.exe (unstripped -g compiled bin):
GDB 4.16 (emx), Copyright 1996 Free Software Foundation, Inc...
(gdb) run
Starting program: d:/tcpip/netscape/plugins/lyx-1.0.1/src/lyxus.exe
[New thread 1]
[New thread 2]
[Thread terminated: 1]
[Thread terminated: 2]
Program exited normally.
(gdb) quit
I have no idea how to run the immediately exploding childs in gdb.
The coredump:
gdb -c core
GDB 4.16 (emx), Copyright 1996 Free Software Foundation, Inc.
#0 0x1f3fbbcb in ?? ()
(gdb) bt
#0 0x1f3fbbcb in ?? ()
#1 0x1f3f645a in ?? ()
#2 0x15c50 in ?? ()
#3 <signal handler called>
#4 0xa81a7 in ?? ()
#5 0xa8c4d in ?? ()
#6 0xa8e70 in ?? ()
#7 0xac83b in ?? ()
#8 0xaad14 in ?? ()
#9 0xc86a5 in ?? ()
#10 0xcab04 in ?? ()
#11 0x1052bc in ?? ()
#12 0x1060d4 in ?? ()
#13 0x9647f in ?? ()
#14 0x95e08 in ?? ()
#15 0x13bd96f8 in ?? ()
#16 0x13be3b55 in ?? ()
#17 0x35c6a in ?? ()
#18 0x10849 in ?? ()
#19 0x10175 in ?? ()
(gdb) info locals
No symbol table info available.
(gdb) up
#1 0x1f3f645a in ?? ()
(gdb) bt
#0 0x1f3fbbcb in ?? ()
#1 0x1f3f645a in ?? ()
#2 0x15c50 in ?? ()
#3 <signal handler called>
#4 0xa81a7 in ?? ()
#5 0xa8c4d in ?? ()
#6 0xa8e70 in ?? ()
#7 0xac83b in ?? ()
#8 0xaad14 in ?? ()
#9 0xc86a5 in ?? ()
#10 0xcab04 in ?? ()
#11 0x1052bc in ?? ()
#12 0x1060d4 in ?? ()
#13 0x9647f in ?? ()
#14 0x95e08 in ?? ()
#15 0x13bd96f8 in ?? ()
#16 0x13be3b55 in ?? ()
#17 0x35c6a in ?? ()
#18 0x10849 in ?? ()
#19 0x10175 in ?? ()
(gdb) quit
So I think the problem is hidden in some library (stripped).
>If it is to be distributed with LyX, then it
>must be in development/OS2 directory.
Or as README.OS2 in top dir. I don't think, it will confuse there, but
it almost certainly won't be found in some deep subdir.
