Linux-Development-Sys Digest #973, Volume #6 Sun, 18 Jul 99 16:13:56 EDT
Contents:
Re: One thing Windows has over Linux. (Jan Andres)
32 Char usernames (David Nillesen)
Re: Problem with fopen and _FILE_OFFSET_BITS on glibc-2.1.1 (Collin Rogowski)
Re: 32 Char usernames (Miquel van Smoorenburg)
Re: when will Linux support > 2GB file size??? (Byron A Jeff)
Re: Bug of GCC (David T. Blake)
Re: Problem with fopen and _FILE_OFFSET_BITS on glibc-2.1.1 (Andreas Jaeger)
Re: when will Linux support > 2GB file size??? (Byron A Jeff)
Re: when will Linux support > 2GB file size??? (Byron A Jeff)
Re: 32 Char usernames ("R.K.Aa")
Call for Authors (Re: NT to Linux port questions) (Gary Lawrence Murphy)
Re: Drivers for Linux... (Christopher B. Browne)
Re: Pro/ENGINEER FlexLM and LINUX (Michael Serban)
Re: Asynchronous IO networking Stack? (XuYifeng)
General Call for Authors (Gary Lawrence Murphy)
----------------------------------------------------------------------------
From: [EMAIL PROTECTED] (Jan Andres)
Crossposted-To: comp.os.linux.x,redhat.config
Subject: Re: One thing Windows has over Linux.
Date: 18 Jul 1999 06:30:07 GMT
In article <[EMAIL PROTECTED]>, Thomas Steffen wrote:
>Zach Brown <[EMAIL PROTECTED]> writes:
>
>> The current solution is to alias loadkeys and xmodmap to commands that use
>> letters that are the same in the different layouts (i.e. "a" and "m" and
>> the number keys are the same in Qwerty and in Dvorak). This becomes
>> tedious because a user can log out and another sits down to use the system
>> only to find that their username "phil" comes out on the screen
>> "ldcn".
>
>Then your system is misconfigured, X should restore the default
>keyboard after a logout. imho the unix way is to integrate language
>selection into xdm, so you can do it by mouse. i think both kde and
>gnome have some program to configure the keyboard on a per user basis,
X11 itself lets a user specify his/her own keytable. This is done via
the .Xmodmap file. Don't ask me about the syntax of this file, I only
know that it can be used for this purpose.
--
Jan Andres
[EMAIL PROTECTED]
Ham radio: DH2JAN
------------------------------
From: David Nillesen <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.setup
Subject: 32 Char usernames
Date: Sun, 18 Jul 1999 19:51:07 +1000
We run an ISP and are interested in using longer then 8 char usernames.
We tried making a username with a dot in it ie "Kaptain.Krash" and
everything seemed to work OK. We tried all the services we could think
of and nothing unexpected happened.
Has anyone else been doing this or are there any problem with longer
usernames?
The info i found on glibc 2 seemed to indicate that upto 32 charachters
would be fine.
If you could let me know of any potential pitfalls it would be much
appreciated.
We are running RedHat so that means we have glibc2.1
Thanks,
David Nillesen
Northnet Internet Services
+61 2 67749300
Type Bits/KeyID Date User ID
pub 2048/BBCA4E3D 1998/01/03 David A Nillesen <[EMAIL PROTECTED]>
=====BEGIN PGP PUBLIC KEY BLOCK=====
Version: 2.6.3ia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------------------------------
From: [EMAIL PROTECTED] (Collin Rogowski)
Subject: Re: Problem with fopen and _FILE_OFFSET_BITS on glibc-2.1.1
Date: Sun, 18 Jul 1999 10:32:23 GMT
>Are you really linking against the freshly installed glibc 2.1.2?
>Check with gcc -Wl,-verbose -o teste test.c
I did that and found out that I used an old glibc. :-(
But I wonder why the gcc uses the old one.
I installed the glibc-2.1.1 in /usr/local/lib (because I didn't want
to break my old programs). I then changed /etc/ld.so.conf so that
/usr/local/lib is the first entry. I even set my LD_LIBRARY_PATH to
just include /usr/local/lib. But the gcc (or the linker) tries to link
with /usr/lib/glibc.
How can I change the search-order the linker uses to find the lib to
link to?
thanks,
Collin Rogowski
------------------------------
From: [EMAIL PROTECTED] (Miquel van Smoorenburg)
Crossposted-To: comp.os.linux.setup
Subject: Re: 32 Char usernames
Date: 18 Jul 1999 13:34:06 +0200
In article <[EMAIL PROTECTED]>,
David Nillesen <[EMAIL PROTECTED]> wrote:
> We run an ISP and are interested in using longer then 8 char usernames.
>We tried making a username with a dot in it ie "Kaptain.Krash" and
>everything seemed to work OK. We tried all the services we could think
>of and nothing unexpected happened.
> Has anyone else been doing this or are there any problem with longer
>usernames?
We use up to 10 characters, but that's only because the DOS-based
accounting program limits us to that.
> The info i found on glibc 2 seemed to indicate that upto 32 charachters
>would be fine.
> If you could let me know of any potential pitfalls it would be much
>appreciated.
2 things:
- most mailers change the name to lowercase before delivering mail,
so don't use uppercase characters or the account cannot receive mail
- The old chown syntax is "chown user.group file [file..]". The newer
versions also support a ":" as seperation character, but there might
be utilities or scripts still using the old version. If you have
dots in usernames you might run into some unexpected trouble.
Mike.
--
... somehow I have a feeling the hurting hasn't even begun yet
-- Bill, "The Terrible Thunderlizards"
------------------------------
From: [EMAIL PROTECTED] (Byron A Jeff)
Crossposted-To: comp.os.linux.advocacy
Subject: Re: when will Linux support > 2GB file size???
Date: 18 Jul 1999 10:57:08 -0400
In article <[EMAIL PROTECTED]>,
Stefan Monnier <[EMAIL PROTECTED]> wrote:
->>>>> "Christopher" == Christopher Browne <[EMAIL PROTECTED]> writes:
-> No, there are *two* missing things:
-> a) FS that naturally supports large files,
-
-ext2 has been supporting files up to 1TB for quite a while now.
That is incorrect for the standand 32 bit Intel kernel. The filesystem in
toto can be upwards of 2TB. However the largest size of any single file in
that filesystem is currently 2G by default.
-
-> b) Applications that support large files.
-
-You can have that right now:
-- apply the Large-File-Summit (LFS) patch to your kernel
-- use a recent glibc
-- recompile your apps with I_CANT_REMEMBER_THE_FLAG_NAME
-
-and that should be it. Some bad apps might break, but it's rather unlikely
-since they would probably break in similar ways on all other Unixes.
What effect does the LFS patch have upon the rest of the system, specifically
ext2 and VFS? Are the bad apps the recompiled ones or are they existing apps
that work fine without LFS?
BAJ
------------------------------
From: [EMAIL PROTECTED] (David T. Blake)
Subject: Re: Bug of GCC
Date: 18 Jul 1999 13:55:48 GMT
Reply-To: [EMAIL PROTECTED]
Villy Kruse <[EMAIL PROTECTED]> wrote:
>How about:
>
>
>#if 0
> char abc[] = "this is a test
>#endif
> char abc[] = "this is another test"
The program:
********************
#if 0
char abc[] = "this is a test
#endif
char abc[] = "this is another test";
main(){
printf("%s\n",abc);
}
********************
DEC OSF cc runs the program without error.
OSF gcc barfs at unterminated string.
egcs on linux barfs similarly.
cc on the SGI compiles and runs it without error.
gcc on linux barfs on unterminated string constant.
Looks like they divide - free vs proprietary compilers.
--
Dave Blake
[EMAIL PROTECTED]
------------------------------
From: Andreas Jaeger <[EMAIL PROTECTED]>
Subject: Re: Problem with fopen and _FILE_OFFSET_BITS on glibc-2.1.1
Date: 18 Jul 1999 16:19:25 +0200
>>>>> Collin Rogowski writes:
>> Are you really linking against the freshly installed glibc 2.1.2?
>> Check with gcc -Wl,-verbose -o teste test.c
> I did that and found out that I used an old glibc. :-(
> But I wonder why the gcc uses the old one.
> I installed the glibc-2.1.1 in /usr/local/lib (because I didn't want
> to break my old programs). I then changed /etc/ld.so.conf so that
> /usr/local/lib is the first entry. I even set my LD_LIBRARY_PATH to
> just include /usr/local/lib. But the gcc (or the linker) tries to link
> with /usr/lib/glibc.
> How can I change the search-order the linker uses to find the lib to
> link to?
Read the glibc FAQ, question 3.18.
Andreas
--
Andreas Jaeger [EMAIL PROTECTED] [EMAIL PROTECTED]
for pgp-key finger [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Byron A Jeff)
Crossposted-To: comp.os.linux.advocacy
Subject: Re: when will Linux support > 2GB file size???
Date: 18 Jul 1999 11:08:22 -0400
In article <7mgaej$hkl$[EMAIL PROTECTED]>,
Miquel van Smoorenburg <[EMAIL PROTECTED]> wrote:
-In article <7mfa79$[EMAIL PROTECTED]>,
-Byron A Jeff <[EMAIL PROTECTED]> wrote:
->It impacts the performance of the overall filesystem. One of ext2 greatest
->features is that it is fast. It just doesn't make sense to impact every
->process accessing the filesystem so that a couple of programs can have 2G+
->access at a performance loss.
->
->Simply create a new FS that supports 2G+ files.
-
-ext2 already supports files > 2GB, it's just that the Linux 32 bit kernels
-do not. On an Alpha you can easily create a > 2GB file on a standard ext2 FS.
While true, that point is moot because the majority of systems we are
discussing in Intel 32 bit systems.
-
-Most of ext2 already uses 64 bit values internally, even on 32 bit
-architectures. Use the Source, Luke ;) for example from linux/fs/ext2/file.c:
-
-static long long ext2_file_lseek(
- struct file *file,
- long long offset,
- int origin)
-
-long long == 64 bits
Now that I think about it this really isn't new information. Without 64 bit
ints, ext2 wouldn't be able to have filesystems that were greater than 2G.
Given this then where exactly is the bottleneck for large files?
-
-There are already patches for Large File Support (LFS) available
-for i386 and other 32 bit archs, it's just that they are not integrated
-into the kernel but that will probably happen in 2.3
I simply have a concern about the impact this support will have upon the
existing infrastructure. In other words will the changed kernel/filesystem/
libraries be as fast and as stable as the existing kernel/filesystem/libraries?
BAJ
------------------------------
From: [EMAIL PROTECTED] (Byron A Jeff)
Crossposted-To: comp.os.linux.advocacy
Subject: Re: when will Linux support > 2GB file size???
Date: 18 Jul 1999 11:01:54 -0400
In article <[EMAIL PROTECTED]>, JC <[EMAIL PROTECTED]> wrote:
-On 11 Jul 1999 02:47:06 -0700, [EMAIL PROTECTED] wrote:
-
->In article <7m8qtt$[EMAIL PROTECTED]>, [EMAIL PROTECTED] says...
->
->thanks for the info.
->
->>
->>So BTW why exactly do you need 2GB+ files?
->
-
-VMWare.
The most compelling example I've seen so far. VMWare client OS filesystems are
stored within a single file. BTW VMWare does allow for native partitions yes?
Let's turn this discussion in another direction. ext2 doesn't support 2G+ files
on the existing Intel 32 bit architecture. Does any other filesystem allow for
2G+ files in the same environment?
BAJ
------------------------------
From: "R.K.Aa" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.setup
Subject: Re: 32 Char usernames
Date: Sun, 18 Jul 1999 17:48:22 +0200
David Nillesen wrote:
>
> We run an ISP and are interested in using longer then 8 char usernames.
> We tried making a username with a dot in it ie "Kaptain.Krash" and
> everything seemed to work OK. We tried all the services we could think
> of and nothing unexpected happened.
> Has anyone else been doing this or are there any problem with longer usernames?
> The info i found on glibc 2 seemed to indicate that upto 32 charachters
> would be fine.
> If you could let me know of any potential pitfalls it would be much
> appreciated.
> We are running RedHat so that means we have glibc2.1
long usernames are ok but looks messy and is a chore to TYPE. In
addition the users get awful path statements that way - too long for
comfort. And any change an existing username usually cause only grief
for the users. What IS common however is to have a fairly short real
username and then set up *mail-aliases* for the person, to enable more
"intuitive" adresses: for instance one mail alias consisting of
abbreviated first/middle name and then fulle surname, and another with
the full name. Or whatever fantasy-name the user prefers for that
matter. The dot as a separator is fine then.
K.
------------------------------
From: Gary Lawrence Murphy <[EMAIL PROTECTED]>
Subject: Call for Authors (Re: NT to Linux port questions)
Date: 18 Jul 1999 12:38:20 -0400
Reply-To: [EMAIL PROTECTED]
I think we can consider this topic a hot-item in need of some proper
tutorial resources. We're working on a book project for porting NT
drivers and applications to Linux. If anyone would like to make some
money with this topic (in exchange for their expert advice and
hard-won experience ;), let me know. There is a description of our
projects (and a signup form) at http://www.teledyn.com/authors.shtml
--
Gary Lawrence Murphy <[EMAIL PROTECTED]> TeleDynamics Communications Inc
Business Telecom Services : Internet Consulting : http://www.teledyn.com
Linux/GNU Education Group: http://www.egroups.com/group/linux-education/
"Computers are useless. They can only give you answers."(Pablo Picasso)
------------------------------
From: [EMAIL PROTECTED] (Christopher B. Browne)
Subject: Re: Drivers for Linux...
Reply-To: [EMAIL PROTECTED]
Date: Sun, 18 Jul 1999 16:47:08 GMT
On Sun, 18 Jul 1999 12:30:35 -0400, Rick Maines <[EMAIL PROTECTED]> posted:
>I am in need of a couple of drivers for some distribution of Linux. I need
>a driver for my Crystal PnP sound card, and one for my ATI Rage IIc AGP
>video card. I'd appreciate any help anyone can offer. Also, if I can't
>find these drivers, how can I write my own?
Looking at the sources in /usr/src/linux/drivers/sound, it appears
that there is support for some Crystal sound cards. If you can identify
the chipset in use, it may be possible to come up with straighter answers.
You might look at /usr/src/linux/drivers/sound/cs4232.c...
For the ATI graphics card, it is highly likely that it is supported by
the XFree86 "Mach64" server. This isn't formally a part of Linux; see
<http://www.xfree86.org> for more details.
--
STATED REASON DOES NOT COMPUTE WITH PROGRAMMED FACTS...
[EMAIL PROTECTED] <http://www.hex.net/~cbbrowne/lsf.html>
------------------------------
From: Michael Serban <[EMAIL PROTECTED]>
Crossposted-To: comp.cad.pro-engineer,comp.os.linux.development.apps
Subject: Re: Pro/ENGINEER FlexLM and LINUX
Date: Sun, 18 Jul 1999 02:55:47 -0700
Hello there,
Please let me know how you solved the problem.
Regards, Michael
[EMAIL PROTECTED] wrote:
> Hello all!
>
> Has anyone tried to use the LINUX version of FLEXlm to do software
> license management for Pro/ENGINEER? I'd like to dump my NT
> file/license server in favor of a LINUX box. File serving is no prob
> for LINUX, of course, but with FLEXlm? Does anyone have experience with
> changing FLEXlm servers in general? Also, I'm planning on keeping all
> the same hardware; this is just and OS change (heh, just!).
>
> BTW, sorry if the cross posting is inappropriate, did a search for
> FLEXlm and posted to what may be relevant groups.
>
> some references:
> Parametric Technology Corp. -- http://www.ptc.com
> GLOBEtrotter Software, Inc. -- http://www.globetrotter.com
> LINUX General Info Home Page -- http://www.linux.org
>
> -Adolphe Youssef
>
> (note to PTC: Hurry up with Pro/E for LINUX!!!!!)
>
> Sent via Deja.com http://www.deja.com/
> Share what you know. Learn what you don't.
------------------------------
From: XuYifeng <[EMAIL PROTECTED]>
Subject: Re: Asynchronous IO networking Stack?
Date: Sun, 18 Jul 1999 20:28:07 +0800
Hi,
I just have a question:
how do you handle sockets more than 64 ?
WSAWaitForMultipleEvents only allow wait on 64 objects, if you have 65
objects, how do you wait
on them?
Linux select can wait on 1024 sockets, select() just like
WSAWaitForMultipleEvents(), both are
wait for event happen on objects, I don't think they are different, of course it
doesn't include Win32
object.
---
XuYifeng
Francois Payette wrote:
> Thanks for your help.
>
> Here is my situation: On Windows NT, our server uses Events with
> WSAWaitForMultipleEvents() to pool for IO completion of asynchronous send
> and receives. Under Windows NT, a high number of thread tends to slow down
> the system due to thread context switching. An alternative is to use a fixed
> number of threads to handle all the network sockets and have IO completion
> detection for equal fractions of the total of sockets in those threads. That
> way, thread switching is limited even when there is a lot of sockets, the
> sockets with data pending take longer to get serviced but they don't cause
> extra thread overhead. In a previous implementation of our server, I had
> regular(BSD not Winsock) sockets, and a call to select with the whole set of
> connected sockets in one servicing thread. Every time one socket would be
> marked as ready by select, I would spin off a thread to process the request
> and send back the result. It turned out that spawning one thread per request
> was causing a lot of overhead when lots of request where processed, due to
> the overhead of thread switching on NT.
>
> I would not mind redoing the select implementation of our server on Linux if
> it's the only solution. I was wondering if someone had a project of
> implementing something similar to events(WSAWaitForMultipleEvents(),
> WSAGetOverlappedResult(), etc..) on Linux. Would it be a desirable feature?
> How is the thread spawning overhead on Linux? For maximum server output, how
> many threads should I force my server to use? (On NT it turns out 4 threads
> per processor is the best for our implementation should it be the same?)
> Should I even limit the number of threads like I did on NT? Is there a
> performance gain/loss to call select multiple times symmetrically?
>
> Last: what is the best development environment with Standard C++
> compatibility on Linux ? Have you tried Metrowerks CodeWarrior? (we used it
> to port our server to NetWare, so I guess our code is compatible on it)
>
> Thanks a lot in advance!!
> Fran�ois
>
> ----- Original Message -----
> From: Dave Platt <[EMAIL PROTECTED]>
> To: <[EMAIL PROTECTED]>
> Sent: Monday, July 12, 1999 10:36 AM
> Subject: Re: Asynchronous IO networking Stack?
>
> > >Is there anything like Winsock Overlapped I/o in the current Linux
> Kernel?
> > >
> > >If not, is there a project on the way?
> >
> > I don't know the definition of "overlapped I/O", but I can offer you a
> > couple of suggestions as to how various forms of non-blocking
> > ("asynchronous") I/O can be implemented in Linux.
> >
> > I believe that the Linux kernel permits individual I/O sockets to be
> > placed into "non-blocking" mode. In this mode, you can issue a read,
> > or a write, without your process being blocked due to lack of data or
> > lack of buffer space. The read/write call does as much work as it can
> > without blocking, returns to you the actual number of bytes
> > transferred (which may be 0) and then returns immediately.
> >
> > Non-blocking socket descriptors are almost always used in conjunction
> > with the select() or poll() calls. These permit your process to be
> > notified when further I/O is possible on the socket (e.g. when data
> > arrives, or when data is successfully transmitted and more buffer
> > space is available). It's also possible for your process to ask that
> > it be sent a signal when I/O is possible... it can do further I/O from
> > the signal handler, or set a flag and let the mainline code do the I/O.
> >
> > Another approach is to use the POSIX-standard aio() asynchronous- I/O
> > library, which is built into the current version of the C library.
> > This library allows you to open I/O and descriptors to almost any sort
> > of device (files, network sockets, raw devices, TTYs, etc.), and
> > submit asynchronous I/O requests to the library. The requests are
> > serviced independently - by a set of threads launched by the library,
> > I believe - and you can poll for completion, or receive signals when
> > the requests are done. This library can be used to perform
> > asynchronous I/O even on devices or files that don't have non-blocking
> > support in the kernel.
>
>
------------------------------
From: Gary Lawrence Murphy <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development.apps
Subject: General Call for Authors
Date: 18 Jul 1999 14:58:01 -0400
Reply-To: [EMAIL PROTECTED]
This is for developers who love to talk about their work and wouldn't
mind being paid to do it: We are looking to staff our Linux/GNU book
projects.
These are not long-term contracts. Contributors are generally
involved for a week or two, and lead authors for 3-5 months depending
on the size of the book. Language/writing skills are not as important
as expertise (we have lots of expert copy editors)
Here's a partial list of the project topics for our 1999 releases:
Linux Network, KDE, QTLib, GTK+, OpenGL and CORBA Programming
An encyclopaedia of Linux programming (and how to program portably for Linux)
Detailed "Linux as a Server" operations/installation guide
Linux/GNU database and web database systems
Advanced guides to using Apache, SAMBA, Python
Complete guides to using TurboLinux, SuSE, SlackWare
Sysadmin guides for all distributions
GNU Software (ie a guide to the entire gnu.org catalog)
Basic CodeWarrior tutorial
Porting Windows/NT drivers and applications
Sysadmin's guides to Linux Security
Kernel Programming (and module programming)
Anyone interested in contributing on any of these topics, as an author
or as a technical editor/reviewer, is invited to sign up at
http://www.teledyn.com/authors.shtml
--
Gary Lawrence Murphy <[EMAIL PROTECTED]> TeleDynamics Communications Inc
Business Telecom Services : Internet Consulting : http://www.teledyn.com
Linux/GNU Education Group: http://www.egroups.com/group/linux-education/
"Computers are useless. They can only give you answers."(Pablo Picasso)
------------------------------
** 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
******************************
