Linux-Advocacy Digest #261, Volume #30 Thu, 16 Nov 00 01:13:03 EST
Contents:
Re: OT: Could someone explain C++ phobia in Linux? (Donovan Rebbechi)
Re: Of course, there is a down side... ("Les Mikesell")
Re: NT/2000 true multiuser? ("Les Mikesell")
Re: OT: Could someone explain C++ phobia in Linux? (Craig Kelley)
Re: OT: Could someone explain C++ phobia in Linux? (Craig Kelley)
Re: Need some advice on Linux ("Les Mikesell")
Re: Need some advice on Linux (Craig Kelley)
Re: I WANT WIN2k drivers! ("Les Mikesell")
Re: Of course, there is a down side... (Craig Kelley)
Re: Help!! ("Les Mikesell")
Re: OT: Could someone explain C++ phobia in Linux? ("tony roth")
Re: OT: Could someone explain C++ phobia in Linux? ("Les Mikesell")
Re: OT: Could someone explain C++ phobia in Linux? ("Les Mikesell")
----------------------------------------------------------------------------
From: [EMAIL PROTECTED] (Donovan Rebbechi)
Subject: Re: OT: Could someone explain C++ phobia in Linux?
Date: 16 Nov 2000 04:27:55 GMT
On Thu, 16 Nov 2000 03:24:37 GMT, Les Mikesell wrote:
>
>"Donovan Rebbechi" <[EMAIL PROTECTED]> wrote in message
>news:[EMAIL PROTECTED]...
>> On Wed, 15 Nov 2000 08:19:41 GMT, Les Mikesell wrote:
>> >
>>
>> >Part of the problem is debugging this complexity. If you have an obscure
>> >problem in your own complex C code, you have a reasonable chance
>> >of debugging it. I
>>
>> But you're considerably more likely to have the problem.
>
>Only if the compiler is perfect. Has that been your experience?
I'm not sure what you mean. I haven't had big problems with compilers
choking on good code. There are some obscure things with templates that
some compilers don't support, but things tend to either work or not work.
And yes, I believe that C code is more likely to have bugs than C++ code.
I've just decided to throw out large amounts of code in a project and
redo it because the guy who coded it did so with a C mentality, and
there are strdups al over the place (few of them freed) as well as a
lot of strings with hard coded lengths.
>The first c++ programs I saw come around in the unix
>open source area, like Sam Leffler's Hylafax had to be
>matched precisely with a version of g++. Every time either
>the source or the compiler version changed it took a new
>round of patches to make it work, and then people
>who couldn't match the compiler version on the OS they
>used would have trouble.
This kind of thing usually happens because the code has some glitch
that the earlier compiler didn't catch (for example, not declaring
the return type of a function)
>> > f the bug is in the complex things the compiler
>> >has done you
>>
>> Not likely in practice. What kind of "complex things" are you referring to
>?
>
>Who knows? One set of developers here is working on a test wrapper
>for some other things, and depending on which service pack they
>have applied to the compiler, a whole loop may disappear.
I'm not clear on what you're talking about. Can you be more specific ?
C++ doesn't just write half of your code for you.
>If it is completely reliable. If it isn't what do you do? You can fix
>the part you did wrong in your own code.
If it isn't you can write your own implementation and write it once, or
if it's open source you can fix it.
Either way, I don't see how you can keep a straight face and try to argue
that it's easier to manage memory in 1000 blocks of code than it is
to manage it in one block of code.
Fixing your own code is going to be a hell of a lot harder if your bugs
are scattered between thousands of lines of code, as opposed to localised
in the one file.
>I'm not. But with a decent debugger I can usually track down
>mistakes in C.
That's a very good thing -- because there are going to be a lot more
mistakes in C. While you chase down 1000 bugs caused by some idiot using
strdup() or strcpy(), I'll simply use the string class (or if it's broken,
which is highly unlikely, I'll write my own, and at least localise
the bugs)
>> > g++ has not been all that reliable at
>> >least until recently (does it have a complete STL yet?). The
>>
>> Nope. Close, but not completely-complete.
>
>How do you deal with portability issues then? And how do
You avoid the more obscure features of STL.
>you learn the language if you can't follow a standard reference?
You can. You just need to avoid the more obscure features. To be honest,
I've only seen one feature that doesn't work in the STL on g++.
In practice, you need to be prepared to target a few STL implementations
(and in practice, there are only a few that you need to target. )
>Using STL seems to me to be the whole point of c++ and I
No, it is not the "whole point". Support for generic programming is
nice, but the main reason to use C++ is the fact that it supports
run time polymorphism within an OO framework.
>don't think I'd want to learn how to work around not having it.
You don't have to "work around not having it".
>> If we wanted everything to be "reliable" in the above sense, we'd all
>> program in ancient languages and ancient APIs and resist all technology.
>
>I don't understand. What is the alternative that you consider better?
Well, perhaps the GNOME developers should all just program in Xlib, and
the Windows developers should program in raw Win32, for fear of learning
anything new !
>> Again, this would be true to the spirit of the "luddite" movement.
>However,
>> those who join the luddite movement should not use GNOME or GTK, because
>> these are object oriented, make heavy use of CORBA, and are certainly not
>> stable.
>
>Is there some plan to make them so?
I don't know. Probably. Certainly, there's a plan to make C++ stable, it's
called the 1998 ansi standards document. Why is this relevant ?
--
Donovan
------------------------------
From: "Les Mikesell" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.ms-windows.advocacy,comp.os.ms-windows.nt.advocacy
Subject: Re: Of course, there is a down side...
Date: Thu, 16 Nov 2000 04:34:30 GMT
"Chad Myers" <[EMAIL PROTECTED]> wrote in message
news:MCIQ5.8699$[EMAIL PROTECTED]...
>
> >
> > rpm -Uvh wine*.rpm
>
> This is brain dead? Why not rpm wine*.rpm?
Rpm does more than one thing, and unlike windows 'setup'
programs it doesn't take a half hour of mousing around
followed by a reboot to tell it what you want.
> Why not have one wine file, why are their multiple?
Why won't 'setup' process as many packages as you
want in one command?
> What's -v and -h for? Yes, I'm sure that it's all in the docs (if
> there are any), but seriously, simply extract an rpm file I must
> really have THREE command line arguments?
Those are for the warm-fuzzy feedback. If you prefer the down
to business no-news-is-good-news approach, leave them off.
> How many are required to get a listing of the contents of the rpm?
> 8? 9?
Do you want the listing from the file or the installed package?
How do you get a listing of a *.cab file both ways?
> > man wine.conf
>
> Ah yes, two things here:
>
> - man the always unintuitive, vague, and rarely helpful Jargon-o
> Machine that seems to only really assist the people who actually
> developed the application you're attempting to get assistance for.
You mean the concise reference page devoid of sales fluff.
> And "man"? I want "help" or "assistance". The term "man" is
> completely back-asswards. Like everything, I guess, in Linux and
> Unix.
Help is for when you are confused. The manual page is for
when you know what you want.
> - The ever confusing, never consistent, and rarely in order .conf
> files.
After all, they are trying to emulate the win32 interface. How
else could it be?
Les Mikesell
[EMAIL PROTECTED]
------------------------------
From: "Les Mikesell" <[EMAIL PROTECTED]>
Subject: Re: NT/2000 true multiuser?
Date: Thu, 16 Nov 2000 04:37:48 GMT
"The Great Suprendo" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> A certain Les Mikesell, of comp.os.linux.advocacy "fame", writes :
> >> What freely redistributable versions are available for those platforms
?
> >> The Citrix client is free on all of the platforms I listed.
> >
> >The X source is available for anyone who wants to port it.
>
> As yet no-one has come up with a free replacement for the (buggy and
> slow) XFree86 implementation.
You must have this confused with something else. XFree86 is
not buggy or slow. You might try WeirdX (the java implementation).
I'd expect it to be slower, but it is free.
> >How much does the Citrix server cost?
>
> I imagine it would be competitive with a commercial X implementation.
There is nothing wrong with the free X versions.
Les Mikesell
[EMAIL PROTECTED]
------------------------------
Subject: Re: OT: Could someone explain C++ phobia in Linux?
From: Craig Kelley <[EMAIL PROTECTED]>
Date: 15 Nov 2000 21:55:10 -0700
"Les Mikesell" <[EMAIL PROTECTED]> writes:
> "Craig Kelley" <[EMAIL PROTECTED]> wrote in message
> news:[EMAIL PROTECTED]...
>
> > > Without good memory management, you can't do a non-trivial program very
> > > well. I don't care what anyone says, garbage collectors don't work in
> > > complex systems. You either end up running too slowly because the memory
> > > manager is taking up too much time, or your process bloats because the
> > > memory manager isn't doing enough.
> >
> > Care to explain that to my Java servlets then?
>
> Someone here wrote some java servlets using xml and an xslt
> transformation and they appear to grow by the size of every response,
> never releasing any memory. I kill the whole jvm when it hits 40 megs
> (every hour or so). Fortunately apache knows how to restart it. How do you
> make a servlet free it's memory?
I haven't noticed it with jakarta at all; but you can tell Java to
garbage collect when you think you're smarter than the garbage
collector is (which happens from time to time). The garbage
collector in 1.3 is more intelligent than the one in earlier
versions.
It is still possible to create persistent references, which cannot be
collected, but that is bad code even in C++.
--
The wheel is turning but the hamster is dead.
Craig Kelley -- [EMAIL PROTECTED]
http://www.isu.edu/~kellcrai finger [EMAIL PROTECTED] for PGP block
------------------------------
Subject: Re: OT: Could someone explain C++ phobia in Linux?
From: Craig Kelley <[EMAIL PROTECTED]>
Date: 15 Nov 2000 21:58:34 -0700
mlw <[EMAIL PROTECTED]> writes:
> Craig Kelley wrote:
> >
> > mlw <[EMAIL PROTECTED]> writes:
> >
> > > Michael Livshin wrote:
> > > >
> > > > [EMAIL PROTECTED] (Donovan Rebbechi) writes:
> > > >
> > > > > I'm not saying you should use C++ without polymorphism. I'm pointing
> > > > > out that you don't need to use much manual memory management in C++.
> > > >
> > > > any non-zero amount is enough for me, thankyouverymuch.
> > >
> > > Without good memory management, you can't do a non-trivial program very
> > > well. I don't care what anyone says, garbage collectors don't work in
> > > complex systems. You either end up running too slowly because the memory
> > > manager is taking up too much time, or your process bloats because the
> > > memory manager isn't doing enough.
> >
> > Care to explain that to my Java servlets then?
>
> I remember seeing some alerts last year that the garbage collection in a
> version of Java cased a problem because it was not freeing the memory
> correctly. I wish I saved the URL.
Yes, it's happened in more than one JVM. Bugs are everywhere
(unfortunately). C++ may be getting garbage colleciton in the next
version, which would be a very good thing in my book.
> > Not that Java is a perfect language (it isn't), but it sure beats C
> > and C++ in the usability department.
>
> I disagree, but then again I do things that are very process intensive.
> Java is too bloated for what I do.
True, the right tool for the right job and all. Apple is pushing Java
as a first class language for OS X. Some of their control panels even
use it.
> > [snip]
> >
> > > > possibility of recovery (or at least very graceful reporting) from
> > > > *any* error is quite important for big long-running applications. you
> > > > can't recover from a segfault terribly well, I'm afraid.
> > >
> > > Sure you can.
> >
> > Considering that the defn of a segfault is that your program is in an
> > inconsistent state....
> >
> > Besides, how exactly to you "catch" a segfault without being in a
> > debugger?
>
> signal(SEGSIGV, sigHandler)
>
> It isn't a catch(..), but it works. If you structure your code well, you
> can recover.
... unless a stray pointer destroyed some of your code or data; which
is usually what has happened by the time a segmentation fault happens.
> > [snip]
> >
> > > The question is why does it suck? and more to the point, how can someone
> > > claim it sucks more than C?
> >
> > It probably doesn't suck more than C, but it definately sucks more
> > than most intellegent languages; for everything apart from devices
> > drivers and anywhere else you need a macro assembler.
>
> I write a lot of low level stuff, and things that process information.
> If I use Java or perl, it takes 5 to 10 times longer to process than if
> I write it in C++.
As do I, but I place my intense code in know black boxes and do most
development in modern languages. n-tier systems are cool for that.
--
The wheel is turning but the hamster is dead.
Craig Kelley -- [EMAIL PROTECTED]
http://www.isu.edu/~kellcrai finger [EMAIL PROTECTED] for PGP block
------------------------------
From: "Les Mikesell" <[EMAIL PROTECTED]>
Subject: Re: Need some advice on Linux
Date: Thu, 16 Nov 2000 04:59:54 GMT
<[EMAIL PROTECTED]> wrote in message
news:8uvg7n$b6n$[EMAIL PROTECTED]...
>
> > Defragment disk and use Partition Magic or fips >=2.0 to resize
> WinME
> > partition size and then install Linux on one or several partitions in
> the
> > remaining
> > space. I suggest 4GB for Windows and 22GB for GNU/Linux (one or two
> > distributions) ... the other 4GB are ok for Solaris on Intel :)
>
> Is there any inexpensive way of doing this? I mean can I get some kind
> of free utilities to defragment and resize the partition?
> Thanks for the advice.
Windows should have it's own defrag utility. Fips will do it should be
included in the install on Mandrake 7.2. However, if your defrag
doesn't move all the files up (and there are some that won't move),
fips can't adjust the partitions. Gnu partd (also free) seems to
work in this case.
Les Mikesell
[EMAIL PROTECTED]
------------------------------
Subject: Re: Need some advice on Linux
From: Craig Kelley <[EMAIL PROTECTED]>
Date: 15 Nov 2000 22:09:45 -0700
[EMAIL PROTECTED] writes:
> > Defragment disk and use Partition Magic or fips >=2.0 to resize
> WinME
> > partition size and then install Linux on one or several partitions in
> the
> > remaining
> > space. I suggest 4GB for Windows and 22GB for GNU/Linux (one or two
> > distributions) ... the other 4GB are ok for Solaris on Intel :)
>
> Is there any inexpensive way of doing this? I mean can I get some kind
> of free utilities to defragment and resize the partition?
> Thanks for the advice.
Most distributions ship with FIPS.EXE that can do this for vfat
partitions (unless Microsoft changed it *again* with ME).
Also be careful: There are reports that the Widnows ME installer
erases any other operating system on your machine; so backup before
you reinstall it.
--
The wheel is turning but the hamster is dead.
Craig Kelley -- [EMAIL PROTECTED]
http://www.isu.edu/~kellcrai finger [EMAIL PROTECTED] for PGP block
------------------------------
From: "Les Mikesell" <[EMAIL PROTECTED]>
Crossposted-To: alt.destroy.microsoft,comp.os.ms-windows.advocacy
Subject: Re: I WANT WIN2k drivers!
Date: Thu, 16 Nov 2000 05:13:30 GMT
"Curtis" <alliem@kas*spam*net.com> wrote in message
news:[EMAIL PROTECTED]...
> If Linux had great application support, I might have stayed and slugged
> it out since I like a challenge that's worth my while, but unfortunately
> I couldn't see why learning Linux would be worth my while so I ditched
> it. I have no problems with it personally. I just see it for what it is
> and appreciate which the OS is appropriate for and it doesn't include the
> average user ... not by a long shot. It will be a few years before I'll
> even remotely reconsider this point.
Don't judge Linux's speed of change by Window's standards... Take
a look at Mandrake 7.2 plus StarOffice 5.2.
Les Mikesell
[EMAIL PROTECTED]
------------------------------
Crossposted-To: comp.os.ms-windows.advocacy,comp.os.ms-windows.nt.advocacy
Subject: Re: Of course, there is a down side...
From: Craig Kelley <[EMAIL PROTECTED]>
Date: 15 Nov 2000 22:13:35 -0700
"Chad Myers" <[EMAIL PROTECTED]> writes:
> "Gary Hallock" <[EMAIL PROTECTED]> wrote in message
> news:[EMAIL PROTECTED]...
> > [EMAIL PROTECTED] wrote:
> >
> > > My my but we are getting nasty today.
> > >
> > > You're starting to sound like jedi :(
> > >
> > > Pretty soon you might start adding the word "hardly" to every
> > > sentence.
> > >
> > > claire
> > >
> >
> > But it really is very simple to install wine. If you can't do it, then you
> > have no brain:
> >
> > rpm -Uvh wine*.rpm
>
> This is brain dead? Why not rpm wine*.rpm?
>
> Why not have one wine file, why are their multiple?
>
> What's -v and -h for? Yes, I'm sure that it's all in the docs (if
> there are any), but seriously, simply extract an rpm file I must
> really have THREE command line arguments?
>
> How many are required to get a listing of the contents of the rpm?
> 8? 9?
Click on the GNOME menu.
Click on gnorpm.
Click install.
Or, just browse to the file using the mouse and double-click on the
RPM. This is not brain surgery.
> > man wine.conf
>
> Ah yes, two things here:
>
> - man the always unintuitive, vague, and rarely helpful Jargon-o
> Machine that seems to only really assist the people who actually
> developed the application you're attempting to get assistance for.
>
> And "man"? I want "help" or "assistance". The term "man" is
> completely back-asswards. Like everything, I guess, in Linux and
> Unix.
Ahem, just like 'winipcfg' and 'musermgr'?
Give me a break.
--
The wheel is turning but the hamster is dead.
Craig Kelley -- [EMAIL PROTECTED]
http://www.isu.edu/~kellcrai finger [EMAIL PROTECTED] for PGP block
------------------------------
From: "Les Mikesell" <[EMAIL PROTECTED]>
Subject: Re: Help!!
Date: Thu, 16 Nov 2000 05:22:39 GMT
<[EMAIL PROTECTED]> wrote in message
news:8uu6mh$60q$[EMAIL PROTECTED]...
> Hi,
> I am a beginner on learning Linux. In my PC, I've already installed
> Win98. Now I also want to install linux in order to dual-boot between
> 2 os. But every time when i am going on the last stage of installation--
> ----"Lilo setup", whatever i do , Lilo can't be installed.
Which distribution and version? All I've used ask if you want to make
a boot floppy during the install process which will work if there is
a problem with the hard disk. But, the recent versions usually
don't have this problem.
Les Mikesell
[EMAIL PROTECTED]
------------------------------
From: "tony roth" <[EMAIL PROTECTED]>
Subject: Re: OT: Could someone explain C++ phobia in Linux?
Date: Tue, 14 Nov 2000 20:45:09 -0800
ya know MLW I rarily agree with you but in this case I do. I keep on
hearing my friends talk about there hate of c++ but then I realize the size
of there code base is um maybe on average 20K lines mine is on average
100k+. C++ works wonders for me!
tr
"mlw" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> Russ Lyttle wrote:
> >
> > mlw wrote:
------------------------------
From: "Les Mikesell" <[EMAIL PROTECTED]>
Subject: Re: OT: Could someone explain C++ phobia in Linux?
Date: Thu, 16 Nov 2000 05:59:50 GMT
"Donovan Rebbechi" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> >
> >Only if the compiler is perfect. Has that been your experience?
>
> I'm not sure what you mean. I haven't had big problems with compilers
> choking on good code. There are some obscure things with templates that
> some compilers don't support, but things tend to either work or not work.
What if you see different runtime behavior depending on the compiler
version? Do the debuggers typically allow tracing to the extent
that you can with C?
> And yes, I believe that C code is more likely to have bugs than C++ code.
If you pick random lines of code, I suppse so, since the compiler does
more checking. I don't see how that means that C has to be
somehow wrong.
> I've just decided to throw out large amounts of code in a project and
> redo it because the guy who coded it did so with a C mentality, and
> there are strdups al over the place (few of them freed) as well as a
> lot of strings with hard coded lengths.
Are the strings really ones that need to grow or do you just
want the source to look pretty?
> >The first c++ programs I saw come around in the unix
> >open source area, like Sam Leffler's Hylafax had to be
> >matched precisely with a version of g++. Every time either
> >the source or the compiler version changed it took a new
> >round of patches to make it work, and then people
> >who couldn't match the compiler version on the OS they
> >used would have trouble.
>
> This kind of thing usually happens because the code has some glitch
> that the earlier compiler didn't catch (for example, not declaring
> the return type of a function)
No, he was doing it on a standards-conforming SGI compiler, I
assume, then having to break it to match the defects of the
g++ dejour - and there were major differences between the
releases that took massive patches to match. Likewise there
were huge patch sets for OS's that had a different g++ version.
> >Who knows? One set of developers here is working on a test wrapper
> >for some other things, and depending on which service pack they
> >have applied to the compiler, a whole loop may disappear.
>
> I'm not clear on what you're talking about. Can you be more specific ?
Different run-time behavior of the same code.
> C++ doesn't just write half of your code for you.
I thought that was the point - with containers and the STL algorithms to
replace most of the things where you would have to write your
own loops.
> >If it is completely reliable. If it isn't what do you do? You can fix
> >the part you did wrong in your own code.
>
> If it isn't you can write your own implementation and write it once, or
> if it's open source you can fix it.
Write your own compiler?
> Either way, I don't see how you can keep a straight face and try to argue
> that it's easier to manage memory in 1000 blocks of code than it is
> to manage it in one block of code.
And how does that relate to c++ vs. c?
> >I'm not. But with a decent debugger I can usually track down
> >mistakes in C.
>
> That's a very good thing -- because there are going to be a lot more
> mistakes in C. While you chase down 1000 bugs caused by some idiot using
> strdup() or strcpy(), I'll simply use the string class (or if it's broken,
> which is highly unlikely, I'll write my own, and at least localise
> the bugs)
Does this mean you can't write bad c++?
> >How do you deal with portability issues then? And how do
>
> You avoid the more obscure features of STL.
>
> >you learn the language if you can't follow a standard reference?
>
> You can. You just need to avoid the more obscure features. To be honest,
> I've only seen one feature that doesn't work in the STL on g++.
As of when? The sstream header has only shown up in the most
recent Linux distributions - has it been hiding somewhere I didn't
know about? And I think it still won't compile a set of trivial
training examples I've seen.
> In practice, you need to be prepared to target a few STL implementations
> (and in practice, there are only a few that you need to target. )
What an odd sentiment, coming from a long history of c's goal
of portability. Has it really come to a point where no one cares
any more?
> >Using STL seems to me to be the whole point of c++ and I
>
> No, it is not the "whole point". Support for generic programming is
> nice, but the main reason to use C++ is the fact that it supports
> run time polymorphism within an OO framework.
Do all problems map to that model? What is an example of something
that you couldn't do by passing structs (or arrays of structs) around?
> >don't think I'd want to learn how to work around not having it.
>
> You don't have to "work around not having it".
Why is it in the standard if you don't need it? Won't portable code
expect all of the standard?
> >> If we wanted everything to be "reliable" in the above sense, we'd all
> >> program in ancient languages and ancient APIs and resist all
technology.
> >
> >I don't understand. What is the alternative that you consider better?
>
> Well, perhaps the GNOME developers should all just program in Xlib, and
> the Windows developers should program in raw Win32, for fear of learning
> anything new !
I don't understand how having to be aware of your own data types keeps
you from doing anything new.
> >Is there some plan to make them so?
>
> I don't know. Probably. Certainly, there's a plan to make C++ stable, it's
> called the 1998 ansi standards document. Why is this relevant ?
Predictable run-time behavior is a nice feature.
Les Mikesell
[EMAIL PROTECTED]
------------------------------
From: "Les Mikesell" <[EMAIL PROTECTED]>
Subject: Re: OT: Could someone explain C++ phobia in Linux?
Date: Thu, 16 Nov 2000 06:06:34 GMT
"Craig Kelley" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> "Les Mikesell" <[EMAIL PROTECTED]> writes:
>
> > "Craig Kelley" <[EMAIL PROTECTED]> wrote in message
> > news:[EMAIL PROTECTED]...
> >
> > > > Without good memory management, you can't do a non-trivial program
very
> > > > well. I don't care what anyone says, garbage collectors don't work
in
> > > > complex systems. You either end up running too slowly because the
memory
> > > > manager is taking up too much time, or your process bloats because
the
> > > > memory manager isn't doing enough.
> > >
> > > Care to explain that to my Java servlets then?
> >
> > Someone here wrote some java servlets using xml and an xslt
> > transformation and they appear to grow by the size of every response,
> > never releasing any memory. I kill the whole jvm when it hits 40 megs
> > (every hour or so). Fortunately apache knows how to restart it. How do
you
> > make a servlet free it's memory?
>
> I haven't noticed it with jakarta at all; but you can tell Java to
> garbage collect when you think you're smarter than the garbage
> collector is (which happens from time to time). The garbage
> collector in 1.3 is more intelligent than the one in earlier
> versions.
>
> It is still possible to create persistent references, which cannot be
> collected, but that is bad code even in C++.
Someone told me that parsing the xml DOM created circular references
which sounds likely, but they also said that at the end of a page the
servlet should completely release everything - but it isn't happening.
I suppose it is time to rebuild with the latest versions of everything.
Unfortunately the guy who wrote the java code is no longer around
so I'll have to hope nothing breaks.
Les Mikesell
[EMAIL PROTECTED]
------------------------------
** 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.advocacy) 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-Advocacy Digest
******************************
