Memory corruption errors!?
Stephen J. Turnbull
stephen at xemacs.org
Wed Mar 22 22:25:27 EST 2006
>>>>> "APA" == Adrian Aichner <adrian at xemacs.org> writes:
APA> Jan Rychter <jan at rychter.com> writes:
>> I was just making a suggestion that would make life easier for
>> some people. Yes, I know it's a beta, yes, I know there are no
>> guarantees. I wasn't complaining, as you might have noticed,
>> just suggesting.
I realized that. I was reminding you that there are reasons why there
are no guarantees; I'll spell them out below.
Executive summary: I doubt anybody but Marcus could properly document
the current bugginess (i.e., explain who should use which combination
of options under what circumstances); isn't it preferable that he
spend the time on *fixing* it?
>> Does the Beta-ness of the beta series mean we should not aim
>> for well-documented config options?
APA> What should we do about this?
IMO, nothing in the options documentation, at least not along the
lines of what Jan is suggesting. The documentation of the options is
correct as far as I can see. If you want to know what those options
actually do when enabled, see Marcus's home page. There's a pointer
at http://www.xemacs.org/People/marcus.crestani/, IIRC the URL.
The code pertaining to the --enable-kkcc setup should be fixed, if it
can be done with minimal effort, or if no workarounds can be found
for some class of testers. Marcus is actively working on the code,
although I think he's currently concentrating on the newgc side rather
than the KKCC side.
(0) "Ordinary users" can always back up to the last usable version.
If they do use features that are basically only available in CVS HEAD,
they're "testers" (whether they want to be or not ;-). But most of
the important features, e.g., Mule on Windows, are available in
reasonably stable, reasonably fresh beta releases.
(1) --with-kkcc SHOULD be backward-compatible with the old collector.
Experience has shown it to be so. And Marcus went to some trouble to
ensure that in his recent refactoring, from my brief review of the
patch. Perhaps he should have tried harder, but AFAIK only Marcus can
judge that very well. :-)
(2) --with-mc-alloc was designed to be incompatible, as was --with-newgc.
Both have had some bugs. Marcus made the decision to refactor those
into a single set of #ifdef's, and IMHO did a pretty good job of it,
along the right lines (ie, leave --with-kkcc as a separate option and
on by default).
This means that the garbage collector code is unstable; this was
mentioned in the announcement of 21.5.25. But Marcus is patching it
on a nearly daily basis; any documentation we put into the configure
script will probably be obsolete within a week or two.
That is the nature of the mainline.
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
More information about the XEmacs-Beta