search

digikar99 / extensible-optimizing-coerce

2 stars 0 forks source link

Some systems failed to build for Quicklisp dist #1

Closed quicklisp closed 1 year ago

quicklisp commented 3 years ago

Building with SBCL 2.1.6 / ASDF 3.3.5 for quicklisp dist creation.

Trying to build commit id fb365b62affeb5e84aebd8a7a338e39a1684fa09

trivial-coerce fails to build with the following error:

Unhandled ASDF/FIND-COMPONENT:MISSING-DEPENDENCY in thread #<SB-THREAD:THREAD "main thread" RUNNING {1001BE8103}>: Component "polymorphic-functions" not found, required by #<SYSTEM "trivial-coerce">

trivial-coerce/tests fails to build with the following error:

Unhandled ASDF/FIND-COMPONENT:MISSING-DEPENDENCY in thread #<SB-THREAD:THREAD "main thread" RUNNING {1001BE8103}>: Component "polymorphic-functions" not found, required by #<SYSTEM "trivial-coerce">

Full log here

digikar99 commented 3 years ago

Erm, complicated story I guess.

It was originally requested that trivial-coerce be fetched from the releases. However, the release v0.0.1 depended on my version of trivial-types, as the original author of trivial-types had archived m2ym/trivial-types; the original version is still incomplete in accordance with CLHS terminologies, and a bit misleading in a few cases. And IIUC, quicklisp is still relying on m2ym/trivial-types.

Quicklisp had failed to build using my version of trivial-types back then; the breakage was perhaps primarily due to cl-syntax. But, a fix was committed to lisp-maintainers/cl-syntax and it was requested that this be added in place of the now archived m2ym/cl-syntax.

digikar99 commented 3 years ago

Possible solutions I see include:

  1. I rename digikar99/trivial-types to trivial-types2 and create/update the trivial-coerce release.

But I don't want to do this; I'd rather prefer some methodology/protocol to transfer/update any archived repositories. Me renaming this to trivial-types2 feels indicative of the community getting a trivial-types3 10 years down the line, which feels ugly.

  1. quicklisp update's trivial-types to the updated version; and I revert the probably cl-syntax breaking commits to trivial-types to and leave those fixes/changes as wont-fix.

No changes to cl-syntax would be inherently required; but cl-syntax would still be without a maintainer, and there is a pending bug/fix independent of trivial-types; so cl-syntax will need to be updated independent of the issues related to trivial-coerce and trivial-types.

  1. quicklisp update's both of trivial-types and cl-syntax to the new URLs, and update trivial-coerce so that it gets fetched from the releases.

I just tested in portacle, using the latest quicklisp release, with the following changes: (i) my version of trivial-types / latest-commit (ii) trivial-coerce tagged v0.0.1 (iii) lisp-maintainers version of cl-syntax / latest commit. The first two allow trivial-coerce to compile/load, but breaks systems that depend on cl-syntax. Doing the third allows systems that depend on cl-syntax to compile/load. I specifically tested jonathan and clack-middleware-clsql. It's possible there could be breakages other than this; but I'm not sure how to locate them.

I'd be glad to know if there are other solutions.

My personal preference is 3; but if there are other breakages which are not simple to fix, 2 seems better, to avoid involving more repositories into this mess.

digikar99 commented 1 year ago

Closing this because new quicklisp version has arrived.