Re: [clisp-list] what's proc for then?

[Peter W. <pet...@wo...>]

Hi,

On Fri, May 17, 2002 at 05:04:32PM -0400, Sam Steingold wrote:
> > * In message <20020517224124.A173@localhost.localdomain>
> > * On the subject of "[clisp-list] what's proc for then?"
> > * Sent on Fri, 17 May 2002 22:41:24 +0200
> > * Honorable Peter Wood <pet...@wo...> writes:
> >
> > On Thu, May 16, 2002 at 02:18:21PM -0700, Sam Steingold wrote:
> > > Update of /cvsroot/clisp/clisp/doc
> > > In directory usw-pr-cvs1:/tmp/cvs-serv14407/doc
> > > 
> > > Modified Files:
> > > 	impbody.xml 
> > > Log Message:
> > > (make_file_stream) [UNIX, RISCOS]: file streams for the /proc filesystem
> > > default to unbuffered (it's better not to mess with those files anyway!)
> 
> I consider it rude that you send a follow-up to a message in
> <clisp-devel> to <clisp-list>.
> Please follow-up to the same mailing list.
>

I have previously asked you not to cc mail to me as I get whatever is
sent to the list, and find it annoying getting duplicates.  You chose
to ignore that request.  I find that rude.  I sent the mail in
question to clisp-list because it relates to the recent thread (in
clisp-list) regarding the buggy i/o under /proc.  I find that
perfectly reasonable.  I also changed the subject of the message so
that it would be obvious that it was not simply a follow-up to a post
on clisp-devel.

Making unbuffered the default under /proc is not neccesarily going to
help this problem.  Do you think it's a good idea to make changes
without knowing the cause of the problem?  Do you think it's a good
idea to make changes to buggy functionality without first
understanding what is going wrong?
 
> > A better solution would be to find out _why_ Clisp is having problems
> > with some files in /proc when other systems running under Linux are
> > not. (Bash, Python, Perl, Ruby, CMUCL)
> > No - I won't go digging around with gdb.
> 
> if you are not willing to help, why do you expect me to spend my time
> here?  do others think that this attitude is reasonable?

I am not willing to help in this specific case.  Last time this bug
came up I wasted many hours on it with gdb.  I feel that file i/o is
such a basic area that you can't expect users to do this for you.  If
this issue is not important to you, fine.  I also have to justify the
time I spend on Clisp, and many hours on a debugger (again) is losing
out to 'it just works' with half a dozen other programming
languages/systems.

> 
> > Writing to files under /proc does not qualify as 'messing with them'
> > (unless you think "foo" is suitable for testing)
> 
> I would have understood your frustration better if you actually _did_
> something.

Why do you think I'm frustrated?  I am a bit shocked by the basic bugs
which keep turning up every time I try to use Clisp other than
trivially.

> 
> If you do not understand log messages, you better ask what they mean.
> [okay, your outburst qualifies as a question just once: here by
> "messing" I meant doing fancy stuff with them, like buffering].
> 
> regarding your previous reports, note that (FORMAT FOO "..." ...) is not
> the same as (WRITE-STRING (FORMAT NIL "..." ...) FOO) wrt atomicity of
> write.
> 

I did not say it was.  I was deliberately experimenting with
_different_ functionality in an attempt to get some idea of the shape
of the problem.

Please just forget my reports on this bug.  I won't be bothering you
with any more.

Regards,
Peter