Re: [Sheflug] LINUX on a P75, 16 Mb ram, 814 Mb HD, No CDROM,

["Stephen J. Turnbull" <nospam [at] sk.tsukuba.ac.jp>]

>>>>> "Al" == Al Hudson <eah106 [at] york.ac.uk> writes:

    Al> Just as an aside, anyone using asserts in their code ought to
    Al> be shot.  They're horrific.

?  Dunno about projects you work on, but asserts in XEmacs generally
(1) guarantee a bug report (sure, this depends on pissing users off,
but experience shows that most are happy to be of that much help), and
(2) assure a bare minimum of useful information in that report.  Even
if the mere fact of an assert doesn't diagnose the bug (but often it
does), it also leaves very distinctive spoor for correlating widely
separated reports of intermittents.  Most XEmacs asserts come in one
of two forms, the default case in a switch(), or preceded by a comment
like

    /* Al said this can't happen, but it does.  Al, will you
       fix this !#$%? code?  You get the PRs until you do!   */

Ie, you don't put asserts in your own code, you put it in someone
else's!  :=)

Of course the switch case should be handled whenever possible by
turning the #define constants into an enum, and we're heading in that
direction---but with 1.5 million lines of code or something like that,
it takes time.  And it's not always clear what should be done in cases
where the evil value isn't predictable; better to assert and crash in
a semi-controlled way than have your heap trompled and crash god knows
where with no useful stack trace.

    Al> Fix what? Every installer I've used has pretty much worked
    Al> completely or not at all, fairly rare that it's partially
    Al> working.

Lucky fellow!  I've had three setup deadlocks in Red Hat installers,
two in Debian, and heaven knows how many when I was beta testing
Turbolinux.  All of them could have been solved (except for the
earliest Red Hat case, which was pre-modules and involved a Japanese
network card unsupported in the install kernel series) with access to
a shell and insmod and/or a simple text editor to munge a file or two
in /etc.  But lacking shell access in the installer, I had to do
things like edit the initrd or reboot to single-user and edit config
files before continuing the standard install process.

True most of those instances were with beta versions of the
installers, but a few were with release versions.

    Al> Flexibility for powerfreaks begins and ends with setting up a
    Al> shell, for normal people it means putting more features in
    Al> easy-to-reach places ;)

But what I objected to was "reassuring."  Flexibility for me is a
means to an end == reassurance.

All I ask is that you don't get me wrong; it is precisely the
convenience and efficiency of installers that leaves me with the free
time to flame you.  I just find that more buttons tends to mean more
things to break, and more policy decisions made behind my back that I
have to search out and check.  That's not "reassuring."

-- 
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
_________________  _________________  _________________  _________________
What are those straight lines for?  "XEmacs rules."
---------------------------------------------------------------------
Sheffield Linux User's Group - http://www.sheflug.co.uk
To unsubscribe from this list send mail to
- <sheflug-request [at] vuw.ac.nz> - with the word 
 "unsubscribe" in the body of the message. 

  GNU the choice of a complete generation.