Linux-Development-Sys Digest #38, Volume #8 Sat, 22 Jul 00 15:13:09 EDT
Contents:
Problem with alsa on 2.4.0-test4 (kris)
Re: Mount of Zip doesn't work since linux-2.4.0-test2 (kris)
Re: measuring time on LINUX ("Norman Black")
Re: A good IDE (Bruce Stephens)
Re: Problem with alsa on 2.4.0-test4 (David A Knight)
Re: floating point issue (Grant Edwards)
How to wait for TCP data to be ack'ed? (Grant Edwards)
Re: How to wait for TCP data to be ack'ed? (Kaz Kylheku)
Re: How to wait for TCP data to be ack'ed? (Dave Platt)
Re: How to wait for TCP data to be ack'ed? (benno)
Re: how to install Gentus in one disk partition only? (John Tsai)
approach to linux kernel internals??? ([EMAIL PROTECTED])
Re: how to install Gentus in one disk partition only? ("Alvin Beach")
Re: How to wait for TCP data to be ack'ed? (Grant Edwards)
SCO emulation mode??? (Pizzamaxsw)
Re: floating point issue (Larry Ebbitt)
[BSD Sockets Programming]: How to determine if a (TCP) socket has closed without
reading (Waleed)
HTML mail? ("Marc A. Lepage")
----------------------------------------------------------------------------
Date: Fri, 21 Jul 2000 22:25:19 +0200
From: kris <[EMAIL PROTECTED]>
Subject: Problem with alsa on 2.4.0-test4
I have following problem
unresolved symbol when loading freshly compiled alsa-driver ...
:: lookup_dentry
this behaviour started with 2.4.0-test2
I checked for differences ... saw there is an __emul_lookup_dentry but
no lookup_dentry in the kernel source ( most probable reference was in
fs/namei.c)
in the old kernel source ( 2.2x ) there is a lot more .. and the struct
lookup_dentry is there ... not in the new kernel ... how to fix this ?
------------------------------
Date: Fri, 21 Jul 2000 22:31:53 +0200
From: kris <[EMAIL PROTECTED]>
Subject: Re: Mount of Zip doesn't work since linux-2.4.0-test2
bill davidsen wrote:
> In article <[EMAIL PROTECTED]>,
> David A Knight <[EMAIL PROTECTED]> wrote:
>
> | Some of the SCSI modules have changed names and are therefore probably
> | not loaded.
> |
> | I encountered the same problem which was that sd_mod is now just sd
>
> Let's make a meaningless change without thinking, which buys us
> nothing but a little namespace clarity and breaks every SCSI system
> using modules. Hacker mentality at its finest. The folks in Redmond must
> be smiling on yet another tarnish on Linux.
>
> I'd like to let all the security bugs in IE be in the news, with no
> complaints about Linux. "I upgraded and my system doesn't work" sounds
> like another o/s I won't name.
>
mmm kinda different ...
You are aware to be using software that has problem spots ...
And as always with Linux development kernels things change from time to
time .. they tend to get resolved before they get in the "stable tree".
So stop whinin..
In redmond they sell release software of this Quality ...
And call it proffesional.
At least U could have stayed in the stable kernel tree... no problem there
...
U have to know what U're playing with ... highly developmental stuff ...
always has some good features.. but tend to break once in a while what is
supposed to happen.
------------------------------
From: "Norman Black" <[EMAIL PROTECTED]>
Subject: Re: measuring time on LINUX
Date: Fri, 21 Jul 2000 13:54:53 -0700
Reply-To: "Norman Black" <[EMAIL PROTECTED]>
gettimeofday has microsecond resolution on 2.? (2.2 for sure) kernels. It
uses the RDTSC instruction if >= Pentium processor. If you want single clock
resolution look into using the RDTSC instruction directly. gettimeofday is a
kernel call and as such is slow compared to pure user mode code.
--
Norman Black
Stony Brook Software
the reply, fubar => ix.netcom
<[EMAIL PROTECTED]> wrote in message news:8l9vja$t9f$[EMAIL PROTECTED]...
> I need to measure the amount of time some code
> path takes and need a way to get the finest time
> stamp before and after the code. This is for
> LINUX on an intel 386 architecture.
>
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
------------------------------
From: Bruce Stephens <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development.apps,comp.os.linux.development
Subject: Re: A good IDE
Date: 21 Jul 2000 23:07:38 +0100
[EMAIL PROTECTED] writes:
> what is the best IDE for Linux ?
This question can't really be answered without more information.
What do you think an IDE ought to contain? What language(s) do you
want to develop in? If your idea of an IDE contains a GUI designer,
what widget set(s) are you interested in?
I suspect most people use a variety of independent tools, however,
there are certainly a few capable IDEs around, too.
Most recently, Source Navigator was released under the GNU GPL.
Source Navigator is the last bit of Code Fusion (Cygnus/RedHat's IDE
thingy) to be made freely available, so all of Code Fusion (I think,
anyway---are there any bits still not available?) is now out there,
free for anybody to use and improve.
Specifically, I'd recommend that anyone working with C/C++ ought to
get Source Navigator and Insight. Insight's not perfect (it's a gdb
made into a GUI), but it has some nice features, and matches Source
Navigator nicely.
------------------------------
From: David A Knight <[EMAIL PROTECTED]>
Subject: Re: Problem with alsa on 2.4.0-test4
Date: 22 Jul 2000 00:34:06 +0100
kris <[EMAIL PROTECTED]> writes:
> I have following problem
>
> unresolved symbol when loading freshly compiled alsa-driver ...
> :: lookup_dentry
>
> this behaviour started with 2.4.0-test2
<snip>
> lookup_dentry is there ... not in the new kernel ... how to fix this ?
try upgrading alsa-driver to 0.5.8b
David
--
Make your site SCREEM - Site Creating & Editing EnvironMent
URL: http://www.screem.org/
Mail: [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Grant Edwards)
Subject: Re: floating point issue
Date: Sat, 22 Jul 2000 00:30:53 GMT
On 21 Jul 2000 20:06:32 GMT, Paul Hughett <[EMAIL PROTECTED]> wrote:
>One practical approach is to write your own function approximations
>using fixed-point arithmetic; this used to be a standard practice
>back in the days when floating point hardware was an extra cost
>option, if it was available at all. The basic idea is that you
>put a conceptual binary point at some convenient place in the
>machine word, do your arithmetic using integers, and shift as
>necessary to keep the binary point where you want it. I think some
>of the lower-end DSP chips provide only fixed-point arithmetic, so
>you might be able to find more details in a DSP text. Programming
>for embedded processors is another possible source of information.
>Alternately, a numeric programming textbook dating back to the
>60s or 70s might be relevant.
This is possible but it's also very tricky. I know of more
than one case where home-made floating point was found to work
incorrectly after a few hundred thousand units are in the field.
Fixing a bug in a device driver is a bit less of a logistical
problem. ;)
--
Grant Edwards grante Yow! Loni Anderson's hair
at should be LEGALIZED!!
visi.com
------------------------------
From: [EMAIL PROTECTED] (Grant Edwards)
Subject: How to wait for TCP data to be ack'ed?
Date: Sat, 22 Jul 2000 00:35:34 GMT
I'm working on a device driver that writes chunks of data via a
TCP connection. I need to be able to wait at certain points
until all of the data written so far has been ACK'ed.
How do I find out if all data in a TCP stream has been ACK'ed
before proceeding?
--
Grant Edwards grante Yow! I don't know WHY I
at said that... I think it
visi.com came from the FILLINGS inmy
read molars...
------------------------------
From: [EMAIL PROTECTED] (Kaz Kylheku)
Subject: Re: How to wait for TCP data to be ack'ed?
Reply-To: [EMAIL PROTECTED]
Date: Sat, 22 Jul 2000 03:45:33 GMT
On Sat, 22 Jul 2000 00:35:34 GMT, Grant Edwards <[EMAIL PROTECTED]> wrote:
>I'm working on a device driver that writes chunks of data via a
>TCP connection. I need to be able to wait at certain points
>until all of the data written so far has been ACK'ed.
>
>How do I find out if all data in a TCP stream has been ACK'ed
>before proceeding?
IMHO, the best way to do this is to implement an application-level
end to end acknowledgement.
--
Any hyperlinks appearing in this article were inserted by the unscrupulous
operators of a Usenet-to-web gateway, without obtaining the proper permission
of the author, who does not endorse any of the linked-to products or services.
------------------------------
From: [EMAIL PROTECTED] (Dave Platt)
Subject: Re: How to wait for TCP data to be ack'ed?
Date: Sat, 22 Jul 2000 05:19:41 GMT
>>I'm working on a device driver that writes chunks of data via a
>>TCP connection. I need to be able to wait at certain points
>>until all of the data written so far has been ACK'ed.
>>
>>How do I find out if all data in a TCP stream has been ACK'ed
>>before proceeding?
>
>IMHO, the best way to do this is to implement an application-level
>end to end acknowledgement.
I agree.
As far as I know, there's nothing at all in TCP which requires that
ACKs be sent only for data which has actually been read by the
application. All an ACK really tells you is that the TCP stack at the
far end has acknowledged receiving the data and that the sending
system's TCP stack need no longer buffer it against loss.
--
Dave Platt [EMAIL PROTECTED]
Visit the Jade Warrior home page: http://www.radagast.org/jade-warrior/
I do _not_ wish to receive unsolicited commercial email, and I will
boycott any company which has the gall to send me such ads!
------------------------------
From: benno <[EMAIL PROTECTED]>
Subject: Re: How to wait for TCP data to be ack'ed?
Date: Sat, 22 Jul 2000 09:39:29 +0200
On Sat, 22 Jul 2000 03:45:33 GMT, [EMAIL PROTECTED] (Kaz
Kylheku) wrote:
>On Sat, 22 Jul 2000 00:35:34 GMT, Grant Edwards <[EMAIL PROTECTED]> wrote:
>>I'm working on a device driver that writes chunks of data via a
>>TCP connection. I need to be able to wait at certain points
>>until all of the data written so far has been ACK'ed.
>>
>>How do I find out if all data in a TCP stream has been ACK'ed
>>before proceeding?
>
>IMHO, the best way to do this is to implement an application-level
>end to end acknowledgement.
You can also try using UDP and let the other end answer when data is
written with an UDP packet.
Benno
------------------------------
From: John Tsai <[EMAIL PROTECTED]>
Crossposted-To: alt.comp.periphs.mainboard.abit
Subject: Re: how to install Gentus in one disk partition only?
Reply-To: [EMAIL PROTECTED]
Date: Sat, 22 Jul 2000 16:06:30 +0800
Hi Vero,
You may manually add two partitions, swap and root.
<swap> if there are 128M memory, 128M~256M all okay for swap.
</> at lease 1.7GB
Best Regards.
John Tsai
FAE department, ABIT Computer Co.
[EMAIL PROTECTED]
On Fri, 21 Jul 2000 09:29:05 -0500, Vero <[EMAIL PROTECTED]>
wrote:
>can Gentus [Red Hat6.1] be installed in just one partition?
>> at least initially just for evaluation.
> recommends three disk partition
>doc recommends /boot ./swap and /root at 16, 8, 500 MB min size
------------------------------
From: [EMAIL PROTECTED]
Subject: approach to linux kernel internals???
Date: Sat, 22 Jul 2000 12:23:17 GMT
hello friends,
can anyone suggest me the correct and effective approach
to read the book 'Linux Kernel Internals' (by 5 authors).pleease
thanks in advance
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: "Alvin Beach" <[EMAIL PROTECTED]>
Crossposted-To: alt.comp.periphs.mainboard.abit
Subject: Re: how to install Gentus in one disk partition only?
Date: Sat, 22 Jul 2000 08:44:03 -0400
Yes. Create one partition labeled, /, (<--forward slash) and a swap
partition. A Linux Native partition has an ID of 82 and a Swap partition has
a ID of 83. You need at least these two partitions. It's not necessary to
have the /boot partition separate, once you have a / (forward slash called
root from now on) partition boot will be placed underneath hence the /boot.
However, I recommend you make a /boot partition. If you some reason you mess
up the linux OS you'll still be able to boot into linux since the kernel and
such will not be on the effected partition. Also 8MB for swap is too small I
would think. The swap partition should be around the amount of MB's of RAM.
The theory is that the OS should be able to dump everything in memory into
the swap partition. This hardly happens, but it depends on what you plan to
do. I'm assuming you're a newbie to linux or at most a little familiar. I
suggest you do some reading on the BOOTING, INSTALLING, GENERAL USAGE and
once you're comfortable with those topics and you've used Linux for awhile,
read about the KERNEL and CUSTOMIZING THE KERNEL. Additionally, if you
decide to make a /boot partition, don't make it 16MB but rather 24MB or so.
You'll be happy once you've advanced a bit with linux and get more
experimental.
Good Luck,
Alvin Beach
LINUX RULEZ!
"Vero" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> can Gentus [Red Hat6.1] be installed in just one partition?
>
> > at least initially just for evaluation.
>
> recommends three disk partition
>
> doc recommends /boot ./swap and /root at 16, 8, 500 MB min size
>
>
>
> --
> Please be complete in any email submissions
> else suffer Suspensional Consequenced Reality
>
> serious email to ---- [EMAIL PROTECTED](Vero)
------------------------------
From: [EMAIL PROTECTED] (Grant Edwards)
Subject: Re: How to wait for TCP data to be ack'ed?
Date: Sat, 22 Jul 2000 15:02:05 GMT
In article <[EMAIL PROTECTED]>, Dave Platt wrote:
>>>I'm working on a device driver that writes chunks of data via a TCP
>>>connection. I need to be able to wait at certain points until all of the
>>>data written so far has been ACK'ed.
>>>
>>>How do I find out if all data in a TCP stream has been ACK'ed before
>>>proceeding?
>>
>>IMHO, the best way to do this is to implement an application-level end to
>>end acknowledgement.
>
>I agree.
That would be nice, but unfortunately that's not an option. The device to
which my driver is to talk is already in the field.
>As far as I know, there's nothing at all in TCP which requires that ACKs be
>sent only for data which has actually been read by the application.
In this case, an ACK is only sent when the data has been read by the
application. Therefore, knowing how much data has been ACKed is useful, and
since I have no other option, that's what I need to do.
I believe I've found some the code in /usr/src/linux/net/ipv2/proc.c that
calculates the number of bytes in the tx queue:
/* Format a single socket into tmpbuf. */
static inline void get__sock(struct sock *sp, char *tmpbuf, int i, int format)
{
[...]
sprintf(tmpbuf, "%4d: %08lX:%04X %08lX:%04X"
" %02X %08X:%08X %02X:%08lX %08X %5d %8d %ld",
i, src, srcp, dest, destp, sp->state,
(tw_bucket ?
0 :
(format == 0) ?
tp->write_seq-tp->snd_una : atomic_read(&sp->wmem_alloc)),
[...]
}
It appears that the 4-line expression following sp->state is the number of
unacknowledged bytes. Correct?
--
Grant Edwards grante Yow! I joined scientology
at at a garage sale!!
visi.com
------------------------------
From: [EMAIL PROTECTED] (Pizzamaxsw)
Subject: SCO emulation mode???
Date: 22 Jul 2000 17:28:03 GMT
Is there any way to run programs for SCO Unix under Linux such a emulation to
run my programs in Linux
------------------------------
Date: Sat, 22 Jul 2000 13:33:23 -0400
From: Larry Ebbitt <[EMAIL PROTECTED]>
Subject: Re: floating point issue
Grant Edwards wrote:
>
> On 21 Jul 2000 20:06:32 GMT, Paul Hughett <[EMAIL PROTECTED]> wrote:
>
> >One practical approach is to write your own function approximations
> >using fixed-point arithmetic; this used to be a standard practice
> >back in the days when floating point hardware was an extra cost
> >option, if it was available at all. The basic idea is that you
> >put a conceptual binary point at some convenient place in the
> >machine word, do your arithmetic using integers, and shift as
> >necessary to keep the binary point where you want it. I think some
> >of the lower-end DSP chips provide only fixed-point arithmetic, so
> >you might be able to find more details in a DSP text. Programming
> >for embedded processors is another possible source of information.
> >Alternately, a numeric programming textbook dating back to the
> >60s or 70s might be relevant.
>
> This is possible but it's also very tricky. I know of more
> than one case where home-made floating point was found to work
> incorrectly after a few hundred thousand units are in the field.
> Fixing a bug in a device driver is a bit less of a logistical
> problem. ;)
Many moons ago, Dan McCraken wrote "Programming Digital Computers". It is
dated but still valid and has an excellent treatment of using binary points
for this purpose, including hints on how to keep track of things.
Foolishly, I loaned my copy to someone years ago and I'd love to have
another.
--
Larry Ebbitt - Linux + OS/2 - Atlanta
------------------------------
From: [EMAIL PROTECTED] (Waleed)
Crossposted-To: comp.os.linux.networking,comp.unix.programmer
Subject: [BSD Sockets Programming]: How to determine if a (TCP) socket has closed
without reading
Date: 22 Jul 2000 18:34:17 GMT
Hi,
I need to be able to consistently determine when the remote side of a TCP
socket has closed.
I know that if you read eof from the socket that often means it has closed,
but if you don't get eof, then you have to consume data. I could live with
that, but when the socket is set to nonblocking, receiving eof is ambiguous.
It could just mean that there was no data immediately available.
SIGPIPE might occur when attempting to read from a closed socket (is this
consistent?), but again there is the problem of potentially consuming data.
I wanted to find some way of polling for or of being notified when TCP sockets
close. For example, in System V's TLI interface, you can call t_rcvdis to
receive notification of a closed connection.
My program uses select to handle a list of sockets with a single process and
puts the sockets in nonblocking mode, so as a workaround, I decided to set the
highwater mark less than the lowwater mark so I could get a list of only
closed sockets and sockets with errors from select. I'm only afraid that this
could cause some unexpected behavior in other processes or kernel processes if
my program is interrupted while the buffer size indicators are set that way.
Does anyone know of any specific problems that could be caused?
Also, does anyone know of a cleaner, better (more portable) way to acheive
this?
I'm using Linux kernel v.2.0.36 from Debian 2.1, but I'd like to make my code
as portable as possible.
Thanks in advance!
Waleed
------------------------------
From: "Marc A. Lepage" <[EMAIL PROTECTED]>
Subject: HTML mail?
Date: Sat, 22 Jul 2000 18:49:10 GMT
I'd like to automate the sending of mail, and I've had success with a
cron'd command like this:
cat message.txt | mail -s "TEXT" [EMAIL PROTECTED]
I'd like to send HTML mail, however, with a command like this:
cat message.html | mail -s "HTML" [EMAIL PROTECTED]
The problem is, the email still appears as text (in a reader such as
Netscape Mail) instead of HTML.
I tried adding:
Content-Type: text/html;
to the top of the email, however, because of the immediately preceding
blank line it is not considered a header, but message data. If I
manually edit Netscape's mail file to make it a header, everything's
perfect.
So how do I send HTML email using mail on the command line? Is there a
way to specify that the content of the message is HTML instead of text?
How do I go about this?
Thanks in advance. Feel free to CC responses to my email address.
--
Marc A. Lepage
http://www.antimeta.com/
Minion open source game, RTS game programming, etc.
------------------------------
** 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
******************************
