Bug in Unix System V C compiler
Jay Lepreau
main()
{
int godamnid1;
int godamnid2;
}
This bug is, as usual, not present in the 4.2 BSD C compiler,
which I use in preference to the Sys V one.
gw...@brl-tgr.uucp
uses long identifiers so that porting it to non-BSD UNIXes
is much more work than it had to be. Very professional.
Clyde W. Hoover
> uses long identifiers so that porting it to non-BSD UNIXes
> is much more work than it had to be. Very professional.
Doug, I for one am SICK AND GODDAMN TIRED of your kneejerk Berkeley phobia!
You also seem to prefer personal insult (as I have learned firsthand) to
discussion of issues. I can no longer put up with such junk, so
I will no longer read your submissions, which is a shame
because you do make some useful contributions (especially in net.lang.c).
Net.unix-wizards should be for the discussion of UNIX problems and solutions
NOT for axe-grinding and personal insults. You are a respected computer
professional (thanks for the System V emulation package). I wish you
could behave in a more 'professional' manner with people on this net,
especially people whom you disagree with.
<<<<FLAME off>>>>
.... Flames to /dev/null (Forgive my flamage, fellow wizards)
--
Clyde W. Hoover @ Univ. of Texas Computation Center; Austin, Texas
(Shouter-To-Dead-Parrots)
"The ennui is overpowering" - Marvin
clyde@ut-ngp.{UUCP,ARPA} clyde@ut-sally.{UUCP,ARPA} ihnp4!ut-ngp!clyde
P McFerrin
main()
{
int godamnid1;
int godamnid2;
}
The above is not bug for system V C compiler. Only the 1st 8 characters
of internal names are significant. Thus the above example attempted to
declare the variable 'godamnid' twice.
John Pierce
Gee, I didn't realize any of you really extra super professional people working
on those really super professional non-BSD systems considered any of us poor
turkeys running BSD systems capable of writing anything you might *want*.
My goodness!!! Do you really mean that someday some really extra super non-BSD
superior professional type programmer might really ... God! This has me
so excited I can hardly type!! ... might really actually maybe want to port
something some slob BSD hacker wrote to their lovely pristinely professional
non-BSD system!!?? I can hardly believe it! I mean, like, man, the thought
just really boggles my (admittedly inferior nonprofessional) mind!!!
Golly Whiz!!! I gonna have to start giving this some thought though that's
probably gonna be hard for a poor nonprofessional like me. Let's see... I
guess I can start with cutting identifiers back to 8 characters (since that
seems to be what all the great professional people on the standards committee
think is best)... And I guess I could put in line numbers so they could find
things easier... WOW, MAN!! I just had a BRILLIANT idea (in my nonprofessional
judgement, of course)... I could write everything in FORTRAN (pure ANSII
standard, of course) since we all know that that's really easy to port...
John Pierce, Chemistry, UC San Diego
sdcsvax!sdchema!jwp
Dick Dunn
> > uses long identifiers so that porting it to non-BSD UNIXes
> > is much more work than it had to be. Very professional.
> <<<<FLAME ON>>>>
> Doug, I for one am SICK AND GODDAMN TIRED of your kneejerk Berkeley phobia!
>
Go back and read the whole sequence. Then just back off. Doug was
responding to some fool posting that claimed that the System V C compiler
had a bug because it didn't discriminate between two overly-long
identifiers. Since the poster of THAT article labeled his coding problem
as a compiler bug, and compounded it by using "godamnid1" as the
identifier which illustrated it, it had already gone from a technical
matter to a flame when Doug shot back.
I don't find the treatment of long identifiers to be particularly nice, but
that's the way the language works right now and it gets tired to see people
keep flaming about the same old problems. Moreover, I'm neither a
Berkeleyphobe nor a Berkeleyphile, but it's getting pretty old to see such
a cheap shot anytime someone like Doug tries to advocate software which is
portable to the extent of not using every little feature and featurette
that Berkeley tossed in.
--
Dick Dunn {hao,ucbvax,allegra}!nbires!rcd (303)444-5710 x3086
...Lately it occurs to me what a long, strange trip it's been.
Mats Wichmann
asleep. I am not going to hope that this old argument can be put to rest,
but....could we please try? The use of long identifiers provided in the
BSD compiler is typical of nonportable language exentsions that we have
all suffered through with many other languages, and it would have been
nice if this could have been avoided with C. However.....now AT&T is
muddying the waters with flexnames in the Sys V compiler, while the
ANSI standard proposal does not call for this `feature'. It is hard to
flame someone for using the tools available to him locally, UNLESS that
person is specifically working on a commercial product designed to be
portable. For some reason we (and I don't claim innocence here myself)
seem very ready to pronounce the author of a program an idiot for using
some particular feature.
Mats Wichmann
Dual Systems Corp.
...{ucbvax,amd,ihnp4,cbosgd,decwrl,fortune}!dual!mats
gw...@brl-tgr.uucp
It is interesting (psychologically, not technically) that every time
I suggest that there has been considerable progress made in some
aspects of UNIX inside AT&T (in this case, a better VAX C compiler),
I get accused of raving Berkeleyphobia. Do you people who blindly
believe that everything Berkeley does is perfect whereas everything AT&T
does is stupid feel threatened when your (unjustified and unjustifiable)
beliefs are called into question?
Jay's silly example of a UNIX System V compiler "bug" (in not supporting
identifiers longer than 8 characters, which is not even true any more)
in response to my pointing out that one can get correct code generation
on 4.2BSD by using the System V compiler (which is what I do), seemed
to me to call for some sort of rebuttal. Since my team had not long ago
tried to port some of U. Utah's C code to a non-BSD UNIX system (by the
way, we helped fund the development of some of this code, so I feel I
have a right to complain), I have had first-hand experience with the
unnecessary porting difficulty that extreme reliance on long
identifiers can cause.
Clyde's "personal insult" from me consisted of my private response to
his net flame that the Teletype 5620 and the layers software were
"obsolete klunkers" (or words to that effect) in which I said that I did
not think he knew what he was talking about (as 5620 users can attest).
When he in turn responded that he was used to receiving personal abuse,
I suggested that there might be a reason for that (using approximately
the same number and tone of words that appear in my paraphrases here).
I wanted to keep this correspondence private, but Clyde has referred to
it publicly without explaining what transpired. One nice thing, though,
is that since he is no longer reading my postings I don't have to worry
about a flame back in response to this.
Sorry to tie up net bandwidth with this, but I felt that silence could
be misconstrued as acquiescence.
Henry Spencer
and then claiming that your software is "portable" is fraud.
--
Henry Spencer @ U of Toronto Zoology
{allegra,ihnp4,linus,decvax}!utzoo!henry
Donn Seeley
USG got a good one to Berkeley's left inode with an SCCS... Now
it's USG on top! Or maybe Berkeley, it's hard to tell... <Bing!>
If this keeps up, we'll start hearing from the 'They don't make 'em
like they did for the 6th Edition any more' crowd... Come on, folks,
if you have to complain about something, complain about VMS or MVS
or whatever it's called. The competition.
If half the effort spent bitching was spent in making the fixes that
negate the motivation for bitching, where would all the fun be?
Clean up your act (where have I heard that before?),
Donn Seeley University of Utah CS Dept do...@utah-cs.arpa
40 46' 6"N 111 50' 34"W (801) 581-5668 decvax!utah-cs!donn
P McFerrin
> Relay-Version: version B 2.10.1 6/24/83; site inuxc.UUCP
> Path: inuxc!ihnp4!zehntel!hplabs!sri-unix!r...@BRL-TGR.ARPA
> From: r...@BRL-TGR.ARPA
> Newsgroups: net.unix-wizards
> Subject: Re: Bug in Unix System V C compiler
> Message-ID: <7...@sri-arpa.UUCP>
> Article-I.D.: sri-arpa.739
> Posted: Sat Aug 4 07:08:15 1984
> From: Ron Natalie <r...@BRL-TGR.ARPA>
>
> The problem is that even cutting back to only eight letters is not
> sufficient because no AT&T C compiler seems to be able to handle even
> eight properly.
>
> -Ron
>
>
The limit of 8 characters for variable names are for 'internal' variables.
The ones that are NOT global and defined within local proceedures.
It is the external (globbal) variables that have additional characters added
to form the external symbol. Therefore, external variables usually have
a limit less than 8. I don't remember how much but I think that it varies
among different releases of the compiler.
Laura Creighton
interest only'' and ``but gee, I only wrote it as a 10 minute throw away
program'' hacks get pressed into service? And who has time to rewrite them?
(answer: the poor sucker who has to port them to another architecture
and/or another flavour of UNIX.)
Laura Creighton
utzoo!laura
The WITNESS
I have been working with Unix for only 8 months and the system manager here
expects me to take every program we get with BSD stuff in it and get it running
on here -- while Rich thinks I'm a wizard, apprentice sorceror comes a little
closer to the mark. (What the heck is a SIGTSTP, anyway?)
The past few weeks or so have seen the coming of wars between System V users
and Berkeleyites. Personally, I like what I hear of 4.2 more (and am eagerly
awaiting an upgrade to it (on a new system) here), but then I have yet to hear
SysV people come out and tell us what their system really does, as opposed to
LOTS of 4.2 articles all over the net.
Will someone please tell me why it matters which is "better", anyway? And why
we compare them at all, given that 4.2 is a programming environment and SysV
is a runtime business environment? After all, there are not that many com-
panies that really need the C language, make, and so on (not to mention f77,
which we don't have, and lex and yacc, which we do). Why would Widget Mfg. Co.
which has purchased a full business package (Micro Manufacturing Systems' MCS
package, requiring RM-COBOL runtime but nothing else, for example), need to
play with /usr/include/sys/callo.h? Or "symbolic links"? (I speak from
experience here; I do some business programming... mainly because the mfg.
package we use is not compatible with the customer order entry package; we
don't have the room on our current system to run the compatible one.)
If System V people want to do development work, they probably can, with AT&T
"add-on" packages. Bugs? System V is still fairly new; let them get worked
out by being reported; with enough pressure brought to bear, even AT&T will
bend. And 4.2, from all accounts, has a few bugs of its own.
If you guys have time to burn (by flaming each other), why don't you come up
with a standardized way of easily making 4.2 and SysV source compatible? Maybe
some kind of alternate -lc with simulations for those functions that can be
simulated, such things as the Berkeley compatibility library to present the
"old-style" Unix directory in the "new" Berkeley format, and such? Those
without source licenses might not be able to use them, but there could be a way
to get object code for "standard" systems, and source-licensed sites for the
others could provide object versions on their own; maybe a hex-type protocol
could be used to post them to the net for VAXen, etc. (Look out: Intel hex
format rides again! :-) Not only will it give you something to do, it would
benefit all of us (including both Berkeley and AT&T).
<<<<flame off>>>>
--
Brandon Allbery: decvax!cwruecmp{!atvax}!bsafw
6504 Chestnut Road, Independence, OH 44131
Witness, n. To watch and learn, joyously.
G.TOMASEVICH
>sufficient because no AT&T C compiler seems to be able to handle even
>eight properly.
The compiler prefixes the name with underscore in the translation to assembly
code, so if the assembler takes only 8 characters, then only 7 of the
original characters get through.
George Tomasevich, AT&T-BL
Doug Gwyn <gwyn>
compatibility between 4.2BSD and UNIX System V. Happens he was
one of the persons most deeply embroiled in the recent flaming.
Directory access routines? Yes. Alternate C library? Yes.
Hex protocol? No way. Send for more details:
gw...@brl-vld.arpa
{esquire,decvax,research}!brl-bmd!gwyn (UUCP)
John Woods
"Ah, let's dump this losing 2.9BSD stuff and go back to Version 6, when
everything WORKED!" ;-)
With tongue firmly protruding through the cheek,
--
----------------------------------------------------------------
John Woods, Charles River Data Systems, Framingham MA, (617) 626-1114
...!decvax!frog!john, ...!mit-eddie!jfw, J...@MIT-XX.ARPA
I have absolutely nothing clever to say in this signature.
r...@brl-tgr.arpa
It slices, it dices, it makes julienne fries, but you've got to hit
that sucker just right.
Sounds like "an automatic vegetable slicer that works on television but
not when you get it home" (Apologies to Alan Sherman).
Enter John Little, inventor of the vegamatic operating system.
-Ron
The WITNESS
--bsa
--
Brandon Allbery: decvax!cwruecmp{!atvax}!bsa: R0...@CSUOHIO.BITNET
^ Note name change!
6504 Chestnut Road, Independence, OH 44131 <> (216) 524-1416
"The more they overthink the plumbin', the easier 'tis tae stop up the drain."
Shub
sounds more like a SLEDGEomatic to me :-)>>> (bearded smile)
...!druxv!cdash {charlie shub AT&T-IS Denver (303) 538-3922}