home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
TeX 1995 July
/
TeX CD-ROM July 1995 (Disc 1)(Walnut Creek)(1995).ISO
/
tex-k
/
tex-k-archive.past
/
tex-k-archive.gz
/
tex-k-archive
/
000287_kb@cs.umb.edu_Tue Feb 8 07:11:57 1994.msg
< prev
next >
Wrap
Internet Message Format
|
1994-10-11
|
3KB
Received: from terminus.cs.umb.edu by cs.umb.edu with SMTP id AA16223
(5.65c/IDA-1.4.4 for <tex-k-exp@cs.umb.edu>); Tue, 8 Feb 1994 12:11:58 -0500
Received: by terminus.cs.umb.edu id AA22338
(5.65c/IDA-1.4.4); Tue, 8 Feb 1994 12:11:57 -0500
Date: Tue, 8 Feb 1994 12:11:57 -0500
From: "K. Berry" <kb@cs.umb.edu>
Message-Id: <199402081711.AA22338@terminus.cs.umb.edu>
To: karl@cs.umb.edu, tex-k@cs.umb.edu
Subject: -g and -O
The principle of least surprise says web2c should act like all the other
GNU/Autoconf-ized packages. That's why I made -g the default. If all the
other packages change, I will too.
In 5.851d (or was it earlier?) web2c did use -O by default. (It also
used -g, if compiling with gcc.) This caused endless trouble because
the optimizers broke.
Anyway, it seems Linux folks make up binary packages that others can
then use. This solves the problem in a much better way (imo) than
changing the way web2c works: naive installers would (should) much
rather install a binary package than compile stuff from sources anyway.
If they prefer to compile stuff from sources, then they should accept
the responsibility of configuring them. If they don't want to do that,
then they are free to make up their own package configured however they
like -- all the TeX (and Kpathsea) software is freely available.
So, I'm not going to change this. (Nor am I going to strip installed
binaries by default, for all the same reasons). Aside from my personal
preferences and needs, I think it's better in general for web2c to be
like other configure packages than for it to strike out on its own.
My point is that you have been so successful
that many of the people now installing UnixTeX --
Victim of my own success, huh? That's certainly how I've been feeling ...
know little or nothing of subtle points like -g vs -O.
Sorry, I still don't consider the differences between -g/-O/-g -O a
subtle point!
I feel you should aim for an installation regime
which allows users hopefully just to say
"configure", "make", "make install".
Right. That is of course what I try to do. As far as I know, this works
on Linux. The results may not be *ideal*, but that's how it goes when
you take the defaults for everything. The defaults have to work well on
all systems, not just yours.
I haven't checked with TeX,
but some programs are 10 times as large under Linux
if compiled with -g and not stripped.
Sounds like a bug in Linux to me. There's no inherent reason that this
should happen -- it doesn't under SunOS (much as I hate SunOS). Maybe
work has to be done on gdb and ld so that you can debug with shared
executables on Linux.