Slashdot Mirror
glibc 2.1 pulled due to license problems
el_nino writes "It seems glibc 2.1 has been withdrawn because of a license problem. A mail from Adam J. Richter to the bugs-glibc mailing list: "The source files in the db2 section of glibc-2.1 point to a nonexistant
file named "LICENSE." The files are apparently written by Sleepycat
software, and the copying conditions found on www.sleepycat.com
for this package are the BSD copying conditions with the advertising
restrictions that FSF claims are incompatible with the GPL." The glibc-2.1-README on ftp.gnu.org says: "glibc-2.1 has been (temporarily) removed, until some political issues are worked out."
" Zack Weinberg wrote with an update: I'd like to correct some factual errors in the story you posted on
Slashdot regarding glibc 2.1 being pulled from ftp.gnu.org.
The pull has nothing to do with the license conditions for db2. There
is some disagreement over whether the BSD license is compatible with
the GPL, but FSF's official stance is that they are. The license for
db2 is clearly stated in an appendix to the libc manual (see
manual/contrib.texi).
glibc 2.1 has been pulled since it cannot be compiled with gcc 2.8 and this conflicts with FSF policy. We are working with RMS to resolve the issue. In the meantime, glibc 2.1 remains available from sourceware.cygnus.com and its mirrors.
zw
Sure, you can switch to Solaris and help us finish making Glibc 2.1 work on it. =) There's about 6 functions that need to be fixed. Unfortunetly, I don't know enough about Solaris internals to do much more than I have.
Seriously though, if anyone wants to help email libc-alpha, I'm on that list. I need people who know details on the dirent64 stuctures, and who understand the utmp/wtmp stuff on Solaris.
Tks,
Jeff Bailey
This is a serious bug.... Any decent Haskell compiler would detect this kind of infinite recursion.... even an indecent one.
(I am David Feuer: dfeuer@his.com
Yeah, right. Release the source code. Give it away, but ONLY if the FSF control your lives. Get a grip. Get a life. And STOP sounding like furking Microsoft.
FSF: YOU ARE IDIOTS.
You failed to correctly parse the statement from the sleepycat folks.
The statement says that glibc can dynamically *load* the dbm code, not that it can dynamically *link* the dbm code.
This means that glibc has code to call dlopen, and that the dbm code is *totally seperate*, and thus there is no license conflict.
If people receiving copies of glibc can't rip that code out and use it under the terms of the LGPL, then that exception is useless. OTOH, if that exception allows the code to be fully distributed undel LGPL with glibc, then it effectively becomes LGPL to anyone receiving it that way, and there's no problem.
- RF (dfelker@cnu.edu)
RMS will be just thrilled that you referred to a GNU product as a "product"!
Its like the Ministry of Truth on Babylon 5. Just when you think that its safe to use glibc, Adam springs forward to KILL IT!!!!
:-)
Hey Adam, how about an update to the Fall 96 Plug'n'Play distro??? Its only been -3- years
I've found glibc 2.1 to be mostly compatible with glibc 2.0.7. Few apps broke. I've since recompiled many packages on my system -- and while a number of them required tweaking to get them to compile, most didnt. Most of the time, adding a single #define fixed the problem.
At this point, I havent got the utmp/utmpx, wtmp/wtmpx business sorted out -- even though I followed their instructions.
The only thing that absoutely wont run is realplayer (not due to the 2.1 O_NDELAY problem)...it segfaults. That's probably a C++
2.7.2 problem.
I've compiled: fileutils/shellutils/textutils, mpg123, named, sendmail, fetchmail, vim, shadow password suite, egcs bootstrap, jmr, minicom, bitchx, XFMail, IceWM, rxvt, libjpeg, ImageMagick.
All these seem to deal with Glibc 2.1 perfectly.
Sorry man, couldn't help myself ;-)
In any case, "portable" is relative. Portability is defined by popularity. Various important UNIX systems supply a useful return value, so it is somewhat portable. Linux is the only major system with a bare-minimum fputs(). Nice fputs() behavior is portable as soon as glibc is updated. (In case you hadn't noticed, Linux is starting to define the standards. Nearly every vendor plans to support our ABI.)
Bullshit. Every UNIX system has the same semantics. I don't know what BSD is, but it sure isn't UNIX. Linux libc 5 has proper signal behavior, just like UNIX. This is not developer-friendly. Default features should be the superset of UnixWare, Solaris, Irix, AIX, and Digital UNIX. (if you actually want BSD crap, you should be running FreeBSD) I would not mind a -D_BSD_SOURCE for compiling abnormal softare.I'd make the morons writing the compilers use an object format that supported symbol attribution.
At the same time, we could then get rid of the idiotic "prototype" crap Microsoft foisted off on ANSI C as a mechanism for hiding near/far pointer autoconversion, since they couldn't resolve caller and callee semantic difference at link time... because they didn't use an object format that supported symbol attribution.
A little SECRET is that ALL YOUR C++ BINARIES WILL BREAK.
I've been going around with some of the players to try to
clarify/resolve this, but I'm very concerned that so little
publicity has been given. I HOPE YOU DON'tT RUN KDE.
I hope you don't refer to the TCP/IP stack in
the Linux kernel. It is 100% pure original code.
Alan Cox asked Berkeley for a GPL version of the
BSD code, and Berkeley refused. Because of that,
it took us about 5 years to get better code than
BSD. Of course, the end result kicks ass.
For me, Linux is quite unstable, unlike FreeBSD!
This sort of political move helps no
user or developer.
The more likely reason they withdrew it is
because its full of bugs.
Yes, this is compiler/library stuff is quickly becoming a mess. On my debian system gcc is gnu-2.7.2 (in order to compile the kernel I suppose), g++ is egcs-1.0.3 ; but I needed gcc-2.8.0 to compile RealNetwork plugins ; on a similar system, I needed egcs-1.1.1 at home to compile specific C++ programs. Even on FreeBSD 2.2.8, I didn't manage to compile wxWindows+a given program with the various compilers ; one of the pre-compiled libstd++ was even broken. Now just throw an uncompatible glibc in, and wait for the next stable release of gtk-1.x, and soon things won't get much better than Windows.
Oh sure...
It starts by giving them a little acknowledgment, and before you know it, they will want to start calling GNU/Linux GNU/BSD/Linux, and that will diminish the recognition that the FSF deserves
Realplayer breaks for me on 2.1 (segfault)...and
I suppose it's possible that Netscape 4.5 might (but it ran okay on 2.0.109 and 112).
In that case I'm using precompiled g++ 2.7.2. I prefer not to use binaries if I can help it.
I have a number of C++ apps based on EGCS's libstdc++ 2.9.0, which work perfectly.
well, speaking as a developer and MCSE who's been using NT for the last couple of years, I'd just like to say that I think NT is the coolest damn thing since sliced bread. I mean, it is SOOOO nice to be able to take a 10-minute coffee break every hour or two while you reboot your system. And the documentation! Oh, its to die for! And i get such a kick out of figuring out which one of M$'s little quirks will pop up next causing my network to go on the fritz and my boss to come looking for my head. I tell you, anyone who thinks linux has a chance against performance like that is insane!!!
keith Gaddis
(not anonymous, lost my password)
Only if you say things like:
o Industry standard BSD TCP/IP!
Or
o Supports feature X!
Where "feature X" is specific to the source you
used.
You can advertise until you're broke, and not have to attach anything to the advertising, so long as you don't specifically mention the software itself and/or features.
No way dude. Stallman can go fuck himself. GNU software is 2% to 20% of what I use, depending on how you count it.
Same goes for BSD stuff.
I'm running Linux 2.2.1 withg glibc-2.1. Now that glibc-2.1 has been banned by GNU, I suppose the term GNU/Linux is totally out of place.
according to a post to gnu.misc.discuss by Joel Klecker , the real political problem is that the readme for glibc recommends egcs and is harsh about gcc. the code for glibc2.1 can be found at ftp://sourceware.cygnus.com/pub/glibc/. The license to Sleepycat's Berkeley-DB isn't a problem, because Sleepycat allows distribution along with glibc. From glibc-2.1/db2/README:
> As a special exception, when Berkeley DB is distributed along with the
> GNU C Library, in any program which uses the GNU C Library in accord
> with that library's distribution terms, it is also permitted for
> Berkeley DB to be loaded dynamically by the GNU C Library to implement
> standard ISO/IEC 9945 and Unix interface functionality.
>
> Sleepycat Software, Inc.
Egcs-1.1.1 doesn't install over your old gcc. Your old gcc is still there. Look at what directories you have in /usr/lib/gcc-lib/ and use the appropriate -V and -b options.
I installed pgcc-1.1.1 (based on egcs-1.1.1), and used it to compile glibc-2.1. Everything works fine as before. My original installation was Red Hat 4.2.
Look, you can get glibc 2.1 somewhere else if you really want it. The FSF, however, is paranoid about copyright infringement, and cannot be distributing files that could put it in legal trouble.
Now, perhaps there are other motives involved, as some have suggested; I have no idea. I'm sick of all these whining newbs jumping on the FSF without a valid reason, though.
- RF (dfelker@cnu.edu)
I think that 2.0.7 is basically each individual distribution's set of patches against 2.0.6 (I may be wrong), which contains many bugs. Don't use 2.0.6! Grab debian or redhat's 2.0.7 and compile it from the srcs.
a) It compiles fine for me (silly excuse)
/opt/glibc21, compile with the stand-alone switch, and then link to the new glibc explicitly)
b) IIRC, binary compatibiliy (or lack thereof) is
one issue reguaring compiling with both EGCS and GCC. I personally run Stampede, with GCC 2.7 and 2.8 nowhere in sight, and only run PGCC 1.1.1.
c) you C library, despite any 'free software' ethos surrounding should NOT EVER be considered a 'user serviceable part' -- I will probably try compiling it, but I will not install it until I'm sure that I can compile the core of my system, and my C compiler under it. (i.e. initially install
it in
That said, I have better things to do with my time.
Get this: the dev_t data type is 64 bits.
Besides being slow on a 32-bit processor
and taking up too much space...
Major numbers are 56 bits
Minor numbers are still 8 bits
Wow. That is incredibly stupid. We can now have
up to 2**56 different device types, but we still
can't have more than 256 of each one. Morons!
they go up to 2.0.11something at alpha.gnu.org.
For whatever it's worth, I upgraded my Slackware 3.4 box from libc5/kernel 2.0.30/gcc 2.7.something. Sure, there were problems, but in every case simply recompiling the offending app against the new lib fixed the problem. Now I've got glibc2.1/kernel 2.2.1/egcs 1.1.1.
I've learned to take claims of "binary compatibility" with a big grain of salt. One should either be prepared to recompile virtually everything on one's system, or wait until someone does it for you, and buy their distribution.
Yeah, and you're probably the same guy who has a pirated copy of NT, because it's too expensive to purchase at your local microshit outlet, right?
..... think of us linux/UN*X (enlightened) folks who aren't bothered by silly little problems.
Next time your windows NT BSOD's
Learn to use it.
This system uses software developed by SuSE.
The Linux kernel was developed in part by Caldera.
This OS is based on the GNU system. Proprietary software is evil; intellectual property is an oxymoron.
This product contains software developed at the University of Whatever.
Yeah sure, you say give them a little acknowledgment. Well, it all starts with just a little acknowledgment, next thing you know, they will be wanting to change the name of GNU/Linux to GNU/BSD/Linux. That would rob the FSF of the recognition it deserves!
I disagree completely.
- This response has been provided as a public service by the Auto Troll Responder(r)
Time to stop reading slashdot until the technical problems are worked out. I've tried commenting on this story, only to not have my comments show up.
"signal() behavior is now incompatible with both libc 5 and UNIX(TM). It follows the non-standard BSD behavior now, gratuitously breaking old software in hard-to-detect ways."
Before you bash glibc, please check your history. According to APUE, in the beginning there was UNIX on the PDP-11. This split into AT&T UNIX (System III and System V), BSD UNIX, and a research version at Bell Labs. Both AT&T and BSD UNIX were very popular and each led to a number of Unix variants. The POSIX standard includes some components from each, and created some from scratch (e.g., the STANDARD sigaction). They still contain a number of differences in things not specified by the standard, including the signal function. While in some cases the behavior of one is clearly better than that of the other, in most cases, which is "better" is entirely a matter of personal taste.
(I am David Feuer: dfeuer@his.com)
My computer is trying to expand the GNU acronym.
I've debugged the program, and here's the expansion going on:
GNU
GNU is Not Unix
GNU is Not Unix is Not Unix
GNU is Not Unix is Not Unix is Not Unix
..
It seems that the original developer of this acronym neglected to put in a base case for the recursion.
Careless, if you ask me.
Does anyone have a fix for this bug?
I do have some temporary fixes to the problem.
a. Equate Gnu only with an animal, and not an Acronym.
b. Equate G with something else (Generic, Gag, Good, Gimmick, Greedy) or nothing.
...then you must be running it on doze, in which case I don't have a lot of sympathy... :p
- RF (dfelker@cnu.edu)
I totally agree... Slackware had it's reasons to sticking with Libc5 and these were the reasons!
I cannot install Linux so it is shit ... ahh, logic. Okay, read again ...
... gee, there are machines where it varies dependend in which mode you compile, that's the whole idea of hiding the real type behind a typedef. Doing it the other way round is the best way to get yourself into trouble.
ncurses, slang and glibc must be compiled against the 2.1. Perl too if you've got loadable modules, apache 1.2.4 is fine for me (almost all loadable modules).
... and sorry, I could not resist here. I've made the upgrade myself too (like others) and over my dead body I would go back to an older libc. Things like socklen_t is not size of an int
Wait for a distribution, I've none though, and let them do the job and sort out the problems.
BTW, it's the first final release and not only was kernel 0.98 the end of all developments either, newer kernels broke older stuff too, so (if it is true he said that) Linus might cool down a bit too.
Actually, you don't, all it needs from libc6 is some of the defines from net/ethernet.h
Just grab that file out of the glibc, and it compiles and runs fine (or you can just hack up everything to use the standard definitions).
I've got it running on my box here without any problems - you can grap a binary from http://weedwacker.highrise.ca/dhcp/
a.
Lot's of C++ related bugs and it's years away from support for modern C++ features. egcs is the way to go for C++.
is fine. use it.
Oh darn. I must have bought (*) the wrong version of Windoze NT 4.0: mine crashes regularly, requires a couple of reboots each week, and doesn't support half the hardware I have at home.
Surely if I bought the right one, it would've beat the cr*p out of my Linux box, which runs as a {telnet,ssh,ftp,http,ip masquerading,postgres} server/ {ada95, c++, Erlang} development environment _without a single reboot_ since I installed it (3 months ago).
Must... buy... new... NT... release...
Chris.
meunierc@netscape.net
--
(*) Actually, they forced a copy down my throat, which I'd be glad to prove. Especially to the DOJ.
Defend Linux against this Microsoft sponsored CRAP. Linux is cool. Linux is FREE.
The GPL has saved us. Go GPL! Go!
Ya know. Sometimes I could do without the poitical attachment. I mean, How free is software if it has to be pulled because the legalese isn't kosher. sigh. maybe, I'm too utopian, why can't things just "be".
Funny how all the BSD'ers scream bloody murder about the "GPL Virus" and how it isn't "truely free". Seems as if the BSD'ers want to exclude everything RMS . . .
Get over it. To maintain the purity of either license requires exclusion of the other. Plain and simple.
Rules must exist to maintain one's freedom. No rules leaves anarchy and destroys freedom. Each does their best to maintain the rules that create freedom as they see it.
I prefer GPL/LGPL.
You prefer BSD.
So what?
It would indeed be stupid to use KDE. C++ is
just a toy for now, unless you think that one
particular version of Visual C++ defines C++.
The library, standards, and compiler people
still don't know what to do about compatibility.
How would _you_ do name mangling? Hmmm...
Now which code fork is FreeBSD, It is hard to keep track.
The claim that EGCS must be used because of thread exceptions problems is false.
For patches to support per thread exception stacks in gcc 2.8.x, see the FreeBSD GCC 2.8.x port. The patches were done by Jeremy Allison of the SAMBA team while he was employed by Whistle Communications.
Jeremy didn't submit the code back directly to the GCC maintainers because he didn't want to widen the implementation chasm between GCC and EGCS.
They can have the code, if they want it, since the FreeBSD ports collection is freely available.
In general, Jeremy's implementation is *better* than the EGCS implementation, since you can make the threads/nothreads decision at application compilation time instead of at compiler compilation time.
This is actually very important for threads implementations with user space components, since it means that non-threaded applications don't have to eat the overhead of the user space components (usually including system call wrappers).
This is not a user space vs. kernel space threads argument; the most efficient threads implementation available at this time, that of Solaris, uses a cooperative scheduling component in user space to minimize context switch overhead. It is a hybrid kernel/user thread implementation, and EGCS on Solaris results in extra overhead for non-threaded applications, just like it does on FreeBSD.
In any case, GCC 2.8.x *can* compile the code, if you just bother to apply patches that have been available for more than six months.
My NT is so damned nice. Unlike Linux, it's also usable.
I run a kernel (2.2.1-ac5) and a glibc 2.1 that were compiled with egcs 1.1.1-release. I find no problems with either them or the compiler.
Some of the late 2.0 series wouldnt compile with egcs -- BUT THAT WAS THE KERNEL's PROBLEM! Egcs fixes are highlighting coding flaws, particularly in the inline assembly and inlined C code in the kernel. The old gcc let it slide, the newer egcs are being much stricter in this sense.
Actually, the latest the newest egcs snapshots arent compiling the kernel due to an inlining problem, which I'm told is fixed in the current CVS tree, and will be with next week's snapshot posted to the FTP sites.
At the moment, i'd say Egcs 1.1.1 is the compiler to use.
Oh, and as I tell so many people, if you compile gcc/egcs from sources (not hard), you dont have to install it AT ALL in order to use it. I didnt install it, until I was sure it compiled everything that I typically use. I've hand-converted my slackware 2.x install to an all-glibc-2.1/all-2.2.X/all-egcs state.
BSD UNIX never existed. BSD is less UNIX than
OS/390 and VMS, both of which have rights to
the name. BSD does not have the source either.
(they did, but not after the lawsuit)
UNIX signal() could cause a race condition if
you use it for something that it was not designed
to do. Don't do that. UNIX signal() has always
behaved in the same way, from before version 7
to the latest SysV. BSD fucked up, and we should
not follow them in their stupidity. The Linux
kernel and libc 5 both get this right; but GNU
weenies decided to break that compatibility.
According to Andreas Jaeger (one of the glibc maintainers) in a post on the glibc-linux mailing list, as far as he knows the real problem is the advise to use egcs instead of gcc-2.8.
So this issue is 100% political. There's no licensing problem whatsoever since glibc has obtained special permissions from sleepycat. It's the unwillingness of the FSF to acknowledge egcs' superiority over gcc-2.8. And they're using microsoft-like methods in their fight.
Anyway, glibc-2.1 is still available and can be found at for example ftp.funet.fi in pub/gnu/funet.
Which is why I have to f**king upgrade our server to NetBSD.
2.0.7 is actually a pile of patches for 2.0.6, it was never actually released.
I think some asshole wants to claim that everything is GNU source, since we must specify -D_GNU_SOURCE to make anything interesting work.
"Products like glibc have too many requirements. forget it".
:-).
That's nice. May be you want to re-install a.out libc4. Converting to ELF in libc5 had even more requirements
Except for GNU libc, fputs() commonly has a useful return value. Why return a junk value?
UNIX has always had "SysV" semantics. I don't care if BSD or NT or VMS gets it wrong. Portable UNIX code requires UNIX behavior... and why would you want code to break anyway? Oh, GNU is Not UNIX; therefore UNIX code should not be able to run correctly. That sucks dude. Maybe I could agree if I had invoked the compiler as "c89", but I didn't do that.This crud is only there to satisfy standards. We need it, but not as the default for gcc. The default should support UNIX software.
Read the license before you start rambling about it.
It says:
*mentioning features or use of this software*
Rawhide is ALPHA release software. It is for people who want to live on the edge. If you want a rock-solid system, stick with the distro that came on your CD, and only apply vendor-approved hotfixes.
Putting together a solid operating system takes a lot of work, even for something like Unix. Unix doesn't magically get stable, people have to work on it. This means beta testing and release engineering. And RedHat has been careful to distinguish its "beta" software from its "release-quality" software.
Why is this so hard for people to understand?
I had the school's web server running on a 486SX/33 with 12 MB ram and a 210 MB HD (Slackware 3.0). This included some Perl CGI and a MySQL database. Also ran appletalk stuff to allow teachers who only know how to run a MAC to update the web page.
Another 20 MB was swap partition.
Still had 50 MB Free. Worked better than the old server that had better hardware (this was a proof of concept, ran this way for three months, no problems) I will admit, CGIs were a little slow.
What the hell do you want, anyway? This is your C library you're talking about.
I can't beleive the number of people who don't realize that EVERY app on your system depends on this. They also don't realize that the people who do the development on glibc use it everyday themselves. They compile it several times a day, run it on their systems and deal with it every day. I am subscribed to libc-alpha, and have seen VERY FEW bug reports come through. Fair enough, it's not necessary to submit bug reports but feel free to stuff a sock in it if you can't be bothered.
Tks,
Jeff Bailey
Glibc 2.1 is HARDLY backwards compatible. I realize that *some* things may break, but in my experience, the majority of things (perl, many CURRENT and ESSENTIAL GNU apps, etc, etc) on my system did NOT work after I upgraded. I tried many, many things to resolve the problem, including recompiling everything that broke. This lead to another dead end, as most everything I recompiled seemed to segfault, generally due to glibc 2.1's buffered_vfprintf(). It's a shame, because it did seem to be faster. In the end, though, I had to restore my previous glibc 2.0.7 system.
This may not be the case on systems that are *COMPLETELY* up-to-date, but I certainly would not recommend attempting to install the current glibc 2.1 on a DYI system of any kind.
- llseek() does not have a prototype; this intentional bug destroyed some people's filesystems when fsck ran
- fputs() does not return a useful value. Both Digital UNIX and Solaris return the number of characters printed.
- SysV shared memory IPC is messed up. The struct ipc_perm members key and seq were renamed to __key and __seq This is incompatible with libc 5, AIX, and Solaris, Digital UNIX, etc. Some apps won't even compile anymore because glibc doesn't define SEMMSL and other SysV IPC related constants. (and they left the UID as a 16-bit value too)
- signal() behavior is now incompatible with both libc 5 and UNIX(TM). It follows the non-standard BSD behavior now, gratuitously breaking old software in hard-to-detect ways.
- socklen_t is not the same size as an int. According to Linus, "shoot the library maintainer". I agree!
- sigsend() is missing
- Almost everything must be compiled with -D_GNU_SOURCE, including non-GNU software. The namespace logic is backwards; one should specify POSIX to get a plain, bland, vanilla (nearly useless) namespace.
That is 7 ways. For every bug I find though, there are likely to be thousands of undiscovered little surprises.Everyone with a different opinion or belief is a zealot.
Have you ever thought about the problems that this clause of the BSD license causes? Imagine for a moment that I use little bits of code from 50 BSD licence projects to put my own project together? Pretty damn soon, the LICENSE file will be bigger than the damn code.
This product includes software developed by this guy.
This product includes software developed by that guy.
This product includes software developed by another guy.
This product includes software developed by someone else.
This product includes software developed by the NSA.
This product includes software developed by someone blinded by preconceptions.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
Yet another example of RMS excluding anyone who is not using GPL. It cant be "free" if it's using the BSD license, according to the GNU zealots, (read gnu.misc.discuss for more info)
Almost every GPL'd SW announcement repeats this hynagogic mantra and it seems to have put 98% of the OSS movement to sleep to the reality that GPL'd software IS NOT FREE SOFTWARE. It may or may not be a good licence, but FREE SOFTWARE can't be licenced. The two concepts are antithetical.
I do not quite understand the problem. Does the SleepyCat license actually require users of the library to mention it or does it only apply to glibc?
:) I personally would have no problem with mentioning in the documentation that a program of mine uses someone's library. Always remember to read the license before linking to a library.
If it is only glibc, then I see no problem with having mention in a file about the software. All GNU programs come with the LICENSE file. What is the problem with having a LICENSE.db file?
If it applies to all users of the glibc (I doubt this), I can see problems with users not knowing they are required to comply to the ad clause. Hey, even the LGPL has requirements you must go along with--I disagree with these but that is another issue.
This last point begs a question: why did GNU only just now decide there was a conflict? I am sure this is a good question whether or not you dislike GNU licenses.
Sean Farley
Anyone fancy porting whatever libc {Free,Net,Open}BSD use to work on Linux?
Just my EUR0.02.
So really, the problem is not the credits file.
---
Sigh....
.h file which was included for some odd reason in my distribution which didn't match the real prototypes... giving me a bunch of undefined errors.
I have Egcs 1.1.1, Glibc 2.1, and Linux 2.2.1 installed and working on my SMP box right now. Really the only problem I had was some old libraries which were in my linker search path being linked instead of the new glibc ones and a
After fixing those minor problems, everything throught went quite smoothly.
The world is neither black nor white nor good nor evil, only many shades of CowboyNeal.
The whole purpose of the GPL is to push RMS' political agenda that all software should be free (speech).
:^)
If you want to use GPLed software, you have to pay the price: all software based on it must also be GPLed.
If you don't like that, make your own programs, and license them however you wish
It's as easy as that. The BSD guys deserve a pat on the back - a few words during bootup or something isn't too much to ask is it?
Codifex Maximus ~ In search of... a shorter sig.
It might also be a good idea to actually read the advertising clause -- it has nothing to do with adding acknowledgement to the source files. What it says is that you have to mention them when you advertise.
Please alter my pants as fashion dictates.
This is the entire reason why the *BSD distributions are entirely upgradeable from src using "make world". 1 command, several hours later, you get a completely recompiled OS. The ports (add on packages) don't get done automatically, but it's fairly trivial to add them in if needed (usually not the case).
This is the reason I switched from using Linux.
The heedless fascism of the GPL has caused so much grief to the rest of the free software community, though, that I'm somewhat amused to see the FSF having similar problems with a much freer license....
Craig
and i was *THIS* close to upgrading this afternoon... :)
-herb
your mom!
TedC
If country is free (people in it have freedom), the country does not have laws. If a country has any kind of law, the country is NOT free. All the democrats that repeat this hynagogic mantra put 98% of the democratic movement to sleep. (of course for you americans, I don't mean the democratic party here)
Since it broke every single binary on my system and I had to revert back to glibc-2.0.7 let's hope they keep it locked up until it really works.
with the problems with glibc2.1.
It seems to me to be a nasty nasty upgrade - tons of things broke, dependencies on other packages are just horrendous, the db.so.2 seems to be broke, StarOffice doesn't work (yes, I know it's StarDivision's fault). Downgrading didn't seem to put my system back in order.
What a nightmare. Next time it should come with a big warning: GLIBC 2.1 CAN BE HAZARDOUS TO YOUR HEALTH.
It almost makes me wish Rawhide was still glibc 2.0 based - I was looking forward to Red Hat 6.0 - not any more...
--
Matt. Want XML + Apache + Stylesheets? Get AxKit.
Haha! This is why I use Slackware. I have seen problems with glibc systems from day 1.
"In true sound..." -Agents of Good Root
Umm...how could missing a prototype possibly cause problems at run-time?
"In true sound..." -Agents of Good Root
Yes, but this will not affect an existing binary, as the comment suggested.
"In true sound..." -Agents of Good Root
Negative.
What does it matter what the next version of Slackware uses. I'm using Slackware 3.1. Of course, I don't think there is anything left on my workstation that was from the original 3.1.
My workstation will more than likely remain libc5 based and up to date.
I need a stable libc for development I am involved with, and glibc just doesn't cut it.
"In true sound..." -Agents of Good Root
but at least i got mine...
and unlike everyine else posting here, it seems to work fine for me so far.
nothing broke except licq. so i had to go download the new (and unfortunately slightly more buggy) version and recompile. no biggie.
although maybe that's because i keep my system fairly up to date, and have been using pgcc (based of egcs-1.1) for some time now.
If I don't put anything here, will anyone recognize me anymore?
How is this license different than the lista incarnation of glibc ?
How did it change ?
Anybody know ?
Read the license, the Sleepycat license is *not* a BSD license although it looks like one at first glance. The principal issue is that they require source code redistribtuion, soemthing which is anethema to the typical BSD license.
You could try compiling from the Debian or Red Hat sources (e.g. ftp://ftp.debian.org/dists/debian/{frozen,unstable }/main/source/*/glibc*.
You may want to read the FSF's philosophy section instead, as you seem to have some misunderstandings about the issues.
The FSF recognises the existance of non-(L)GPLed free software; see Categories of Free and Non-Free Software and has chosen to adopt e.g. X11 for GNU.
There is however a difference between "free" and "(L)GPL-compatible". In the case of traditional BSD-style licensed code, like the sleepycat db, the advertisement clause makes it incompatible with the GPL (I've not studied the LGPL in detail, but I suspect it's like the GPL in this regard); see The BSD License Problem.
Go to ftp.**.kernel.org. That's where I've found all the newer glibcs.
--
Aaron Gaudio
"The fool finds ignorance all around him.
"Every man is a mob, a chain gang of idiots." - Jonathan Nolan, Memento Mori
You'd have a point, if the commercial world wasn't just as bogged down with lawsuits and battles over intellectual property. I'd say the FSF frustrated over a BSD-ish advertising clause is a little less severe than Microsoft being told to pull all Java products until the resolution of the Sun/MS trial. Go figure.
--
Aaron Gaudio
"The fool finds ignorance all around him.
"Every man is a mob, a chain gang of idiots." - Jonathan Nolan, Memento Mori
There is a solution to your problem, as highlighted by Linus:
"We all know Linux is great...it does infinite loops in 5 seconds."
Hope that helps.
--
Aaron Gaudio
"The fool finds ignorance all around him.
"Every man is a mob, a chain gang of idiots." - Jonathan Nolan, Memento Mori
The database is broken, articles appear and disappear at random. Whether this message will appear or not is impossible to predict.
slashdot.org == a couple of interesting articles + many stupid flamewars + lots of bad spelling + the odd intelligent comment.
No offense, but what do you expect? It's the core library that -everything- links against, and competes with the kernel for being the most complex piece of software you'll run.
I'd expect it to have a few dependancies on other software, frankly, and when you're upgrading something as fundamental as that, I'd expect that you'd think the upgrade through a little more carefully, or wait for your OS vendor to provide an upgrade if you're not up to the task.
-Ed Felix qui potuit rerum cognoscere causas.
The only reason the default configuration of egcs would install over gcc is when gcc was in /usr/local. If gcc is in /usr/local for whatever reason, configure --prefix=/usr/local/egcs and add simlinks to put stuff in the path as egcc or whatever.
Install it someplace else and name it 'egcs'. The linux kernel will not build with egcs right now, among other things. Read the linux kernel docs.
Hmmm... I can only seem to find 2.0.6 on the ftp sites. Anywhere know where to get it?
Isn't it up to the GCC maintainers to decide whether applying Jeremy's patches are a good idea or not? Instead, now we have Linux gcc 2.8, FreeBSD gcc 2.8, and EGCS. Isn't this situation worse? Sorry if I'm questioning the motives.
Um, what has the GPL to do with the BSD license?
Nevermind the fact that many BSD people have given
up the credit clause.
Please, come again?
I forget what 8 was for.
Not that he is truly evil, and yes, he is talented, but his attitude sucks big time and cause trouble.
Just IMO.
<^>_<(ô ô)>_<^>
So what!!! You have a 10 gig drive and you can't spare a thousand bytes for credits?? There is nothing wrong with wanting a little credit for your work. RMS is an ass.
I found it at: ftp.funet.fi/pub/gnu/funet
"I disagree, specifying a preprocessor symbol should not _remove"
"features, it should add them when necessary"
No. When you define _POSIX_SOURCE this will
limit the features to the POSIX subset. This is
according to POSIX and it makes sense. How could you
you write portable code if you couldn't define an exact
environment? If the "default" feature set is bigger
than any feature set implied by a particular symbol
(e.g. _POSIX_SOURCE) then defining the symbol must
remove features. Think about it.
OK, we have here two operating systems. One gives you choice; the other fetters you with restrictions. For instance, a Linux webserver can run on as little as 16M (maybe less) of ram, on a 486, with 400M of disk space. NT, just sitting there, takes 34M of ram. Another example--the licensing. You laugh because the glibc 2.1 had a case of mistaken licensing. Look, however, at the results that emerge from previous glibc's (amongst other gpl'd code): success by fixability. While NT users rely on sendings from the mothership at Redmond (bugfix packages, euphemized as "service packs"), each user of gpl'd software is empowered to enact their own fix, if they can. It's called freedom to use you machine any damn way you want, not the way Microsoft tells you to.
Fighting the War on the War on Drugs.
http://smokedot.org/
So where is 2.0.7? I only up through 2.0.6 at prep.ai.mit.edu.
I downloaded it and found it required egcs.
The documentation make nasty comments about
gcc. I downloaded egcs-1.1.1 and compiled/built
it. Installing it, it installed over gcc.
glibc now compiled but I did not install it.
egcs broke some things. I had problems
with ncurses and linux-2.2.1 stopped compiling.
Maybe that was my problem. maybe that was egcs'
problem. I got rid of egcs. Products like
glibc have too many requirements. forget it.
I use use the old gcc-2.7.2.3 to compile linux kernel and it works just fine. /usr/bin(if you used --prefix=/usr) but the /usr/lib/gcc-lib/ still has the old ones there.
If you installed egcs-1.1.1 it just overwritten your files in
I had a problem in compiling glibc-2.1 with the latest egcs snapshots(19990208 and 19990131). It seems that it breaks something in memcmp and in glob. Does anyone has idea about this?