[asdf-devel] New OPERATION subclasses in ASDF 3

[Faré]

On Wed, Jan 1, 2014 at 2:31 PM, Robert P. Goldman <rpgoldman at sift.info> wrote:
> Recent experience shows that changes leading to ASDF 3 have changed the
> way one must specify new operation classes, in a way that is not
> backward-compatible.  Thanks to Faré for helping me diagnose problems in
> one such system.
>
> I will be looking into this, and hope to add at least one new test to
> ensure that we are probing this important aspect of ASDF going forward.
>
> Related to this, does anyone have a good sense of which quicklisp
> systems might contain new operation and component classes?
>
You could try to install everything in quicklisp, then grep the .asd
and .lisp files for defclass.*operation
See ql-test:install-all-quicklisp-provided-systems in
ssh://common-lisp.net/home/frideau/git/ql-test.git
less -p 'defclass.*operation' $(grep -il 'defclass.*operation' **/*.{asd,lisp})

Victims are:
arnesi and arnesi+ (clean-op, should use traverse, not recurse downward)
asdf-dependency-grovel (dependency-op - now fixed)
autoproject (revert-op, buggy like arnesi's clean-op)
paren-files (parenscript-compile-op)

I leave it to someone else to contact the maintainers (except for
a-d-g, which I hopefully fixed).


Non-Victims follow.

These only act on files, and downward or sideways recursion doesn't apply:
protobuf (proto-to-lisp)
cffi-grovel (process-op)

These don't want downward recursion, and take steps to make the
downward recursion implicit in ASDF 1&2 but an expensive NOP, sideways
recursion is probably a good thing not to apply:
weblocks (wop::test-op, doc-op, make-app-op)
autoproject (simple-system-wide-operation and subclasses, but not revert-op)
deoxybyte-systems (doc-op)
hu.dwim.asdf (develop-op)
qbook (publish-op)

These have clever component-depends-on method to recurse properly:
def-symbol-macro (stat-op)

These have asdf3 semantics, possibly in #+asdf3:
lisp-executable (create-executables-op, sticky-beak-op, ugly)

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
By modern standards, Common Lisp's syntax is rather verbose at the micro level.
But it more than makes up for it at the macro level.