Upgrade failures for asdf-3.1.7 and later

James M. Lawrence llmjjmll at gmail.com
Mon May 8 12:05:54 UTC 2017

On Thu, May 4, 2017 at 2:35 PM, Robert Goldman <rpgoldman at sift.net> wrote:
> On 5/4/17 May 4 -1:17 PM, James M. Lawrence wrote:
>> As I said, LispWorks PE provides an old ASDF. To verify this,
>> * download LispWorks Personal Edition
>> * launch it
>> * (require "asdf")
>> * (asdf:asdf-version)
>> Again, I am writing on behalf of users, not myself personally. I don't
>> use LispWorks PE or any LispWorks version for development. Others use
>> LispWorks PE. Some do so to evaluate not only LispWorks but Common
>> Lisp as a language. It seemed reasonable to prevent them from getting
>> into a borked state, if possible (and not too difficult or annoying).
> I'm not sure what is the proper solution.  Should we detect this kind of
> condition and raise an error if ASDF is not reconfigured after an upgrade?
> R

If you have consideration for the end-user experience in mind, then
you should emphasize, underline, asterisk, and caps-lock shout that
install-asdf.lisp should be run before doing anything else. Given the
circumstances, that seems like the best you can do.

Don't even suggest doing a live "upgrade", if you have consideration
for end-users, since no user will expect an upgrade feature to
purposefully bork things. That doesn't match what the word "upgrade"
generally means. To the general end-user who adds the quicklisp init
code to their init file and doesn't otherwise bother with asdf (I
suspect this is the vast majority of users), it's just a way to break
quickloading for some unknown reason. (And it's not clear that there
is *any* practical use case for it -- see my previous message starting
at "I don't know why the on-the-fly feature exists".)

Yes, I get that LispWorks shouldn't be bundling an old asdf. Yes, I
get that Quicklisp shouldn't be bundling an old asdf. Yes, I get that
Quicklisp shouldn't be using the central registry. Yes, I get that
LispWorks PE is not for serious use. And so forth. I am addressing how
the lisp ecosystem actually exists today, as opposed to how developers
wish it to be. I am looking at the experience of end-users, not me

More information about the asdf-devel mailing list