Linux-Development-Sys Digest #678, Volume #6 Tue, 4 May 99 19:14:08 EDT
Contents:
Re: what is safe optimisation level for kernel? (Mark Brown)
Re: exhausted memory (Kelly Burkhart)
Re: glibc-2.1 and incompatible apps ("Stefan Monnier "
<[EMAIL PROTECTED]>)
Re: what is safe optimisation level for kernel? (Johan Kullstam)
Re: suggestion to scsi-drivers (Andreas Dilger)
Re: glibc-2.1 and incompatible apps (David T. Blake)
Re: [ANN] CodeWarrior for Red Hat Linux, GNU ed. Shipping (Robert Wuest)
Re: glibc-2.1 and incompatible apps (Nix)
----------------------------------------------------------------------------
From: Mark Brown <[EMAIL PROTECTED]>
Subject: Re: what is safe optimisation level for kernel?
Date: 04 May 1999 19:05:12 +0100
Johan Kullstam <[EMAIL PROTECTED]> writes:
> i think you can compile linux 2.2.x with either gcc-2.7.2.x or any of
> the egcses. i'd use the latest egcs which is 1.1.2 since it's 2
> bugfix versions deep.
It might fail - some people still have problems, but in general it's
safe. Keep your old kernel around just in case.
> > What about pgcc? I read that XFree-3.3.3 hangs when
> > compiled with pgcc, so pgcc is bad and egcs is safe? Is it bug in
> > pgcc and will it be corrected?
> i do not know. call me lazy, but i don't bother compiling xfree86.
It's a bug in PGCC. However, XFree86 is not the kernel and it may be
that the kernel doesn't excercise that bug. Again, try it and revert
to what you were using before if things don't work out.
--
Mark Brown mailto:[EMAIL PROTECTED] (Trying to avoid grumpiness)
http://www.tardis.ed.ac.uk/~broonie/
EUFS http://www.eusa.ed.ac.uk/societies/filmsoc/
------------------------------
Subject: Re: exhausted memory
From: Kelly Burkhart <[EMAIL PROTECTED]>
Date: 03 May 1999 21:32:01 -0500
[EMAIL PROTECTED] (Jerzy Tarasiuk) writes:
> >>>>> Neil Schemenauer <[EMAIL PROTECTED]> writes:
> > [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
> >> Is there a way to free up already used memory? I really hate to
> >> reboot after compiling many programs because there is no more
> >> free memory, I'm always left with about 1k free. Can anyone help
> >> me find a solution. Thanks a lot.
> > You are misinterpeting the word "free" as Linux uses it. Try
> > substituting "wasted" for "free". Here is what the "free"
> > program gives me right now on my machine:
> > total used free shared buffers cached
> > Mem: 31124 30392 732 19204 5580 12264
> > -/+ buffers/cache: 12548 18576
> > Swap: 17132 1276 15856
> > What this says is that I have 732k of wasted memory. I have
> > 18576k of free memory available to programs (more if stuff was
>
> Ok, Neil. I don't know if the posting you responded to is about
> same problem as this I have seen on my computer - maybe it is.
>
> I noticed sometimes amount of used memory - this really used,
> NOT including buffers/cache - is growing, and amount of really
> free (i.e. free + buffers + cache) becomes very small. Finally
> I can reboot the computer by reset button only - shutdown or
> init cannot be loaded because of lack of memory.
>
> The problem occurs on RH 5.1 with kernel 2.0.34.
>
> I need some way to diagnose reason - any ideas?
>
> Is it possible to ask system about all chunks (pages?) of
> memory, and get info which is used by which process?
>
> I hope to either find unallocated memory which kernel cannot
> find to be free, or memory allocated to some process which
> doesn't know it possesses so much, or to non-existent process.
> But I need get access to tables or control blocks describing
> memory, and to scan it (hope r/o access for root is allowed).
>
> I suppose the reason is a bug either in kernel or in module.
>
> If anyone knows how to solve the problem, please reply by
> e-mail, too, because I read this newsgroup very rarely.
> Thanks in advance, Jerzy
Try running top. It will show the same info as from free plus a list
of running processes sorted by %cpu (i think). If you hit 'M'
(capital M) top will sort processes by memory (%MEM or SIZE, I can't
remember). That may give you an idea what is sucking up all of your
memory.
--
Kelly R. Burkhart
[EMAIL PROTECTED]
------------------------------
From: "Stefan Monnier <[EMAIL PROTECTED]>"
<[EMAIL PROTECTED]>
Subject: Re: glibc-2.1 and incompatible apps
Date: 04 May 1999 14:34:07 -0400
>>>>> "Juergen" == Juergen Heinzl <[EMAIL PROTECTED]> writes:
> In article <[EMAIL PROTECTED]>, "Stefan Monnier wrote:
>> How do people work around the problem of applications compiled
>> for glibc-2.0 that don't work with glibc-2.1 any more ?
>> I guess I can keep a glibc-2.0 around and play around with LD_LIBRARY_PATH,
> No.
Well, my current solution goes like this:
- setup an directory with old glibc-2.0 libraries in /foo/lib
- ln -s /foo/lib/ld-linux.so.2 /lib/ld-linux.2.0
- edit the executable and change the "ld-linux.so.2" into "ld-linux.2.0\0".
the trailing zero char is important to maintain the same number of chars.
- play around with wrappers and LD_LIBRARY_PATH
It's butt-ugly but it seems to work.
Stefan
------------------------------
From: Johan Kullstam <[EMAIL PROTECTED]>
Subject: Re: what is safe optimisation level for kernel?
Date: 04 May 1999 08:59:40 -0400
[EMAIL PROTECTED] (Jacek Pop�awski) writes:
> I read that people compile 2.2.x with egcs, and I heard that 100%
> safe is only gcc. What is the true?
i think you can compile linux 2.2.x with either gcc-2.7.2.x or any of
the egcses. i'd use the latest egcs which is 1.1.2 since it's 2
bugfix versions deep.
> Can I safe compile kernel with
> -march=pentium?
i use -march=pentiumpro all the time - including kernels[1]. just
edit /usr/src/linux/arch/i386/Makefile and adjust the arch depdendent
CFLAGS.
> What about pgcc? I read that XFree-3.3.3 hangs when
> compiled with pgcc, so pgcc is bad and egcs is safe? Is it bug in
> pgcc and will it be corrected?
i do not know. call me lazy, but i don't bother compiling xfree86.
[1] the various performance tests i've run on my ppro system show no
substantial improvement in using -march=pentiumpro over -m486.
-mpentium on a ppro hurts by about 10%. however, if you have a
classic pentium, then -mpentium could be a big win. ymmv.
--
johan kullstam
------------------------------
From: [EMAIL PROTECTED] (Andreas Dilger)
Subject: Re: suggestion to scsi-drivers
Date: 4 May 1999 15:47:50 GMT
In article <[EMAIL PROTECTED]>,
John Donald <[EMAIL PROTECTED]> wrote:
>it's still a bit stupid though. depending on my zipdrive being on or off
>my cdrom drive is either /dev/sdc or /dev/sdd. a 'symbolic link' to the
>scsi-vendor information would be a way to solve this, but who's going to
>implement that?
With my IDE CD-ROM, the boot messages are of the form "hdc: <blah> CDROM",
so in my startup, I simply do a
ln -sf `dmesg | grep CDROM | cut -d: -f1` /dev/cdrom
and then /dev/cdrom is always correct. I admit I had the same problem with
my JAZ drive, so all I did was make sure that it had an address higher than
my other drives, so it didn't confuse the drive names.
It's still not as good a solution as having a unique ID on the disks, and then
using that to determine the mount point. This would be especially good with
removable media, so that one disk will be mounted on /usr/local/mp3, and
another would be mounted on /usr/local/lib/fonts or whatever.
Cheers, Andreas
--
Andreas Dilger University of Calgary \"If a man ate a pound of pasta and
Micronet Research Group \ a pound of antipasto, would they
Dept of Electrical & Computer Engineering \ cancel out, leaving him still
http://www-mddsp.enel.ucalgary.ca/People/adilger/ hungry?" -- Dogbert
------------------------------
From: [EMAIL PROTECTED] (David T. Blake)
Subject: Re: glibc-2.1 and incompatible apps
Date: 04 May 1999 12:38:43 -0700
"Stefan Monnier <[EMAIL PROTECTED]>"
<[EMAIL PROTECTED]> writes:
>Well, my current solution goes like this:
>
>- setup an directory with old glibc-2.0 libraries in /foo/lib
>- ln -s /foo/lib/ld-linux.so.2 /lib/ld-linux.2.0
>- edit the executable and change the "ld-linux.so.2" into "ld-linux.2.0\0".
> the trailing zero char is important to maintain the same number of chars.
>- play around with wrappers and LD_LIBRARY_PATH
>
>It's butt-ugly but it seems to work.
There are several commercial apps that use GNU tools but don't
rely on compatibility across versions. Matlab is one. Running
the program starts a script which established the appropriate
version of libc as the LD_PRELOAD and the other compatible
libraries are in LD_LIBRARY_PATH. That is how Matlab uses
the old libc5 on a current linux glibc2.1 system.
If the new apps are open source just compile the app
again, or grab the new RPMS, or .debs, or whatever.
glibc2.1 is really not so tough. Just grab new xlib, new
ncurses, and work the rest out as you go.
--
Dave Blake
[EMAIL PROTECTED]
------------------------------
From: Robert Wuest <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development.apps,comp.unix.programmer
Subject: Re: [ANN] CodeWarrior for Red Hat Linux, GNU ed. Shipping
Date: Tue, 04 May 1999 15:05:23 +0000
MW Ron wrote:
>
> In article <[EMAIL PROTECTED]>, [EMAIL PROTECTED] wrote:
>
> >GNU Edition? does that mean it's Free Software? wow!
>
> As Sumner replied it means this first edition of CodeWarrior uses the GNU
> tools instead of CodeWarrior's tools. I just wanted to confirm this for
> you "officially"
>
> Ron
Is there any kind of demo or test drive version of this?
Robert
------------------------------
From: Nix <$}xinix{[email protected]>
Subject: Re: glibc-2.1 and incompatible apps
Date: 04 May 1999 17:03:51 +0100
Sid Boyce <[EMAIL PROTECTED]> writes:
> I downloaded some stuff for RedHat 6.0 and upgraded to glibc-2.1.1. I
> noticed that there was a libNoVersion.so.1 showing up when you did a ldd
> on any app compile with glibc-2.0.
> I installed the compat libs, these place the glibc-2.0 libs in
> /usr/i386-glibc20-linux/lib which I added to ld.so.conf after /lib,
> however the 2.0 apps fail as it looks like they attempt to use libc.so.6
> and libm.so.6 for glibc-2.1.1. Even building a wrapper either to preload
> these libs or using LD_LIBRARY_PATH doesn't do the job. If you place the
> 2.0 libs directory ahead of /lib in ld.so.conf, the reverse problem
> occurs, the glibc-2.1.1 binaries fail with undefined symbols.
> libNoVersion has symbols such as .......
> U __curbrk@@GLIBC_2.0
> U __deregister_frame_info@@GLIBC_2.0
[snip]
> I would guess there is a way to get both types of binaries to run, but
> I've not seen anything else specific to achieving that.
It looks like you built or acquired a version of glibc-2.1 without
symbol version information.
If
nm /lib/libc-2.1.so | grep '@@GLIBC_2\.0'
yields no output, you have a glibc without symbol versioning.
Solution: Build glibc-2.1 without the --disable-versioning flag.
IMHO no versions should be built with this flag anyway,
unless they have a set of binutils that do not understand
versioning.
--
/* I hate C so much... */ --- jwz, in driver/xscreensaver.c
------------------------------
** FOR YOUR REFERENCE **
The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:
Internet: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.development.system) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Development-System Digest
******************************
