If we can have only one of these changes in, I'd like !59, because it fixes bundles on ECL on platforms that can't link libraries from libraries (is that what makes it fail on Mac?). It's not a regression, though, so unless Daniel K says it's urgent we can wait for the next release.

!60 is a bug fix for people using logical pathnames, i.e. asking for trouble. I would be nice, but not indispensible.

!61 is still a WIP pending review by Daniel K. It's me dropping the sponge on trying to make load-bundle-op an implicit default at all on ECL — but maybe that's not what Daniel K wants, and he wants it to be a non-default implicit setting, rather than either a default implicit setting or something you only get explicitly. I don't think it's urgent.

PS: For our readers, these numbers refer to entries on gitlab: https://gitlab.common-lisp.net/asdf/asdf/merge_requests/

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org Democracy is the theory that the common people know what they want, and deserve to get it good and hard. — H. L. Mencken

On Mon, Dec 12, 2016 at 1:58 PM, Robert Goldman rpgoldman@sift.net wrote:

I would very much like to get 3.2.0 released, but I am starting to feel like we have a Zeno process here. Merge requests are appearing as fast as they are merged.

Currently pending for 3.2.0:

!59 Link objects only

!60 Fix bad-system-name for "logical" .asd pathnames

!61 Remove *load-system-operation*

Which, if any, of these must be in our 3.2.0 release candidate? I would like to freeze a release candidate, marking it as version 3.2.0, for people to play with for at least a week, before we push it into the world.

I'd very much like to do this this week, since the holidays are coming up when people's minds will turn to other things than testing new ASDF versions.

I can pretty much guarantee that if the answer to the above is "we need all three," we are talking a 2017 release, instead of finishing this year, because of limitations on my available testing cycles in the remainder of 2016.

Robert Goldman writes:

Currently pending for 3.2.0:

!59 Link objects only

This is a long-standing bug in ASDF misusing using ECL's builder. ECL now will sanitize it in builder, so if its not merged, building in ASDF may fail with an error in some cases (instead of producing not working binary).

!61 Remove *load-system-operation*

This bug is introduced by enabling load-bundle-op by default on ECL. Effectively it prevents from switching to bytecodes compiler at runtime (i.e after loading ASDF) - ASDF uses wrong operation in that case for loading.

MR removes the whole load operation switch interface (what fixes this issue). @Fare: I see nothing controversial with it (i.e nothing to add with the peer review).