home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.sun.misc
- Path: sparky!uunet!decwrl!csus.edu!netcom.com!rfg
- From: rfg@netcom.com (Ronald F. Guilmette)
- Subject: Re: GCC, Solaris 2.0, Cygnus Support, and a lesson in corporate ethics
- Message-ID: <3v7m2cr.rfg@netcom.com>
- Date: Sat, 15 Aug 92 15:52:38 GMT
- Organization: Netcom - Online Communication Services (408 241-9760 guest)
- References: <g1xm!yb.rfg@netcom.com> <WBE.92Aug13012313@crystal.bbn.com>
- Lines: 111
-
- In article <WBE.92Aug13012313@crystal.bbn.com> wbe@bbn.com (Winston Edmond) writes:
- >That's certainly an interesting piece of history about Cygnus, gcc, and the
- >port to Solaris 2.0 for SPARCs. There are some things you didn't discuss,
- >though.
- >
- >First, the $2000 Cygnus solicited was advertised as having two parts:
- >(1) (which you mentioned) it would help get gcc ported to Solaris 2.0 and
- > placed on the Catalyst CD-ROM by the shipping date for Solaris 2.0, so no
- > one would have to be without a C compiler;
- >(2) (not mentioned) it purchased one year of Cygnus's support at a price 2/3
- > of the usual price for such support.
-
- So you are saying that the "standard" price for a Cygnus "support" contract
- is $3000, right? I'm sure that the folks that are paying cygnus $25,000
- and/or $100,000 for the same service will be interested to hear that.
-
- >Second, part of Cygnus's appeal to potential contributors was that they were
- >promising a working version by a specific date -- the Solaris 2.0 ship date.
-
- Considering that at the time of their "offer", GCC had already been fully
- hosted and targeted to Sparcs (running SunOS 4.1.2) and that it was also
- running really well on a variety of other SVR4 systems (e.g. i386, i860,
- m88k) due to my work, I don't think that the boys at Cygnus were sticking
- their corporate necks out very far in making this "promise".
-
- >Sure, we all figured it was likely that gcc would someday be ported to
- >Solaris 2.0 for SPARCstations, but, in the absence of knowing with reasonable
- >certainty when, the trade-off is a $2000 one-time cost versus loss of
- >developers' time and development delays if appropriate tools aren't available
- >when needed. That's easy to decide.
-
- I have nothing against buying "insurance", but if one is going to buy
- insurance against the possibility of a very unlikely event (e.g. a metorite
- hitting your house) I would think that any intelligent person would at
- least ask what the probability of such an event is before spending money.
-
- >Had you made your porting successes
- >more widely known, perhaps some contributors might have chosen not to
- >contribute.
-
- My porting successes (of GCC to various SVR4 systems) have been widely
- known to users of GCC on SVR4/x86, SVR4/i860, and SVR4/m88k systems for
- some time. Anyone who took the time to fetch and inspect a copy of the
- (freely available) GCC sources for any version since 2.0 (released quite
- some time ago) would have easily been able to see that I think.
-
- I'm sorry if you (and others) didn't know about my SVR4 porting work,
- but unlike Cygnus, I spend almost all of my time actually writing code,
- and almost no time writing press releases which trumpet my successes on
- the net and in other forums.
-
- >In the end, the one year's support and the guarantee were worth the risk that
- >it might all be a waste of money.
-
- A "waste of money", eh? (You said it. I didn't.)
-
- >From: rfg@netcom.com (Ronald F. Guilmette)
- >>Something else that struck me as being quite humorous at the time was
- >>Cygnus' apparently deliberate reinforcement of the widely held belief
- >>(based purely on the uninformed speculation of certain unrelated netnews
- >>posters) that Solaris 2.0 would *not* be bundled with an assembler, a
- >>linker, standard libraries, or header files. ... even though *anyone* ...
- >>could have easily determined ... that Solaris 2.0
- >>would in fact be "bundled" with an assembler, a linker, a set of standard
- >>libraries, and a nearly full complement of standard system include files.
- >
- >Yes, it was common knowledge, posted to various newsgroups at the time, that
- >these *would* be in Solaris 2.0. However, gcc doesn't use the Sun assembler,
- >it uses GAS, and its output had better be compatible with whatever the linker
- >required. I had the impression that this was not entirely a solved problem.
-
- To put it bluntly, you are confused. You do not need GAS to use GCC. Nor
- do you need to use the GNU linker. And contrary to your assertion, running
- either GCC or g++ on an SVR4 system is an entirely solved problem. If you
- don't believe that, you should just ask the folks over in comp.unix.sysv386.
- I'm sure they will be glad to tell you that many of them have been using
- GCC under SVR4 on i386/i486 machines for quite some time.
-
- >>They [Cygnus] also ... broke a couple of minor things, and totally disabled
- >>any possibility of using g++ on Solaris 2.0 on Sparcs.
- >
- >Could you elaborate on what's broken? Also, why would anything Cygnus did
- >prevent g++ from the FSF from ever working on Solaris 2.0 on SPARCs? I would
- >normally assume whatever was broken would eventually get fixed.
-
- Richard Stallman and I worked out a very clean scheme for getting C++
- file-scope object constructors/destcructors to execute at the right
- times via the use of some nice features available as a standard part of
- *all* SVR4 linkers (i.e. the .init and .fini sections). Up until Cygnus
- got ahold of it, this scheme was working just fine on all SVR4 systems
- *including* ones based on Sparc processors. The scheme relied on an
- assumption that the `gcc' driver program would link your programs
- with certain additional .o files (i.e. crtbegin.o and crtend.o) which
- are both derived from a file I wrote called crtstuff.c.
-
- For most GCC confiugurations which target some SVR4 environment, the gcc
- driver program *does* cause these files to be linked into your program.
- For reasons which are not apparent however, Cygnus changed the solaris
- configuration such that these critical files DO NOT get linked in. Their
- only explanation for doing this was a one line comment which they added
- to the `spc-sol2.h' file (which I originally wrote) saying that "we do not
- yet support g++ on solaris" (or words to that effect).
-
- >On behalf of hackers, developers, and project managers everywhere, I'd like
- >to thank you (and everyone else that's been involved, whoever they may be)
- >for any and all contributions toward getting gcc to compile SPARC code for
- >Solaris 2.0.
-
- OK. I appreciate your thanks, but how about sending me a $2000 check also?
- It's the least you can do.
-
-