Mail Archives: djgpp-workers/2001/06/29/09:37:22
On Fri, 29 Jun 2001, Hans-Bernhard Broeker wrote:
> On Thu, 28 Jun 2001, JT Williams wrote:
>
> > I may well be overlooking something, but from this thread
> > it seems that the _only_ way to get into trouble is to install
> > stock djdev203 *after* installing gcc-3.0
>
> ... and answering "yes" when unzip asks whether you want to overwrite
> the linker script.
>
> The problem is that we can't stop users from doing exactly that.
> Experience has shown that there's no way short of an automatic
> splap-user-in-the-face-for-not-reading-READMEs demon that works.
>
> > If this is the case, it sure seems reasonable to me to require
> > that djdev 2.03 be installed _prior_ to installing gcc-3.0, i.e.,
>
> ... and that's exactly the heart of the problem: there's no foolproof way
> of requiring it --- fools are far to ingenous for that. Not unless we
> change the packaging to an RPM-like format, for all packages (!), which
> outright refuses to install a package if its preconditions aren't
> fulfilled.
>
> Neither does that cover the case of an installation from scratch, where
> some industrious user may well unzip gcc30b before djdev203.
>
I myself sometimes put all ZIPs in a same directory and unzip them all
with one line so it can be unpacked in any order. I don't think we can
stop users from doing that
ftp://ftp.delorie.com/incoming is not world readable. So currently
nobody except me can test these binaries which caused so much
discussions. So please go to
http://www.ltn.lv/~pavenis/gcctest.html
download them, test and then we can continue the discussion.
I tried to think how to make gcc-3.0 binary archives for DJGPP
more foolproof for users who doesn't read readme files.
There will be no more builds of gcc-3.0 (or even repacakaging)
from me at least up to July 10th. So go at URL mentioned above
and test them. (I'm thinking about posting this URL in
main DJGPP mailing list with corresponding warning that these
are not official binaries yet)
Andris
- Raw text -